feat(explorer): add Markdown editor write path (#1349 )

* feat(explorer): add Markdown editing write path * fix(explorer): address markdown editor review findings * feat(explorer): add Markdown editor write path --------- Co-authored-by: Sameer Kadam <sskadam6305@gmail.com>
docs(ontology): document quality gate threshold semantics (#1450 )
2026-09-04 04:01:07 +00:00 · 2026-09-04 01:38:17 +05:30 · 2026-09-03 23:40:00 +05:30 · 2026-09-03 23:20:39 +05:30 · 2026-09-03 23:05:32 +05:30 · 2026-09-03 22:29:50 +05:00
55 changed files with 6140 additions and 242 deletions
@@ -1461,6 +1461,22 @@ semantica-explorer --graph my_graph.json

 For contributor / dev-server setup: **[explorer/README.md: Local Setup Guide](explorer/README.md)**

+The CLI exposes the loaded `ContextGraph`. To also browse and edit an existing
+`AgentMemory`, create the ASGI app programmatically with both live objects:
+
+```python
+from semantica.context import AgentMemory, ContextGraph
+from semantica.explorer.app import create_app
+from semantica.explorer.session import GraphSession
+
+graph = ContextGraph()
+memory = AgentMemory()
+app = create_app(session=GraphSession(graph), agent_memory=memory)
+```
+
+The Memories workspace is shown only when `agent_memory` is provided. Apply
+updates the supplied runtime object; it does not add disk persistence.
+
 ---

 ## What's New in v0.6.7
@@ -1469,7 +1485,7 @@ For contributor / dev-server setup: **[explorer/README.md: Local Setup Guide](ex

 - **First-class LangChain integration** (`semantica[langchain]`): a `BaseRetriever` and `VectorStore` over `HybridSearch`, plus graph/decision-query tools
 - **SAP OData ingestor** (`semantica[ingest-sap]`): OAuth2/Basic-auth, SSRF-guarded ingestion for Business Partners and Sales Orders, following the existing Snowflake/Databricks connector pattern
- **`ContextGraph` gains deterministic, human-editable Markdown round-trip persistence** alongside the existing JSON API, and the Explorer graph inspector gains a read-only Markdown content viewer
+- **`ContextGraph` gains deterministic, human-editable Markdown round-trip persistence** alongside the existing JSON API, and Explorer can validate and apply Markdown edits to individual graph nodes and AgentMemory items supplied by the hosting application
 - **`reasoning` gains a structured Action layer**: rule-driven `Assert`/`Retract`/`Call`/`EmitEvent` actions with optional provenance, turning the reasoner into a production-rule system
 - **`run_shacl_validation` is now a public, documented API**, and a dozen ontology/RDF export correctness fixes land: OWL property/class export, SHACL target-namespace resolution, one canonical confidence datatype across all four RDF formats, reachable OWL-Time reification, JSON-LD default-graph and content-derived document identity, and full metadata passthrough on every RDF serializer
 - **Security**: Agno's `AgnoKnowledgeGraph.load_urls()` and OpenClaw's MCP tool now route outbound requests through the shared SSRF guard
@@ -195,7 +195,7 @@ apt29_intel = context.retrieve(
 ```python
 from semantica.llms import LiteLLM

-llm = LiteLLM(model="anthropic/claude-sonnet-4-20250514")
+llm = LiteLLM(model="anthropic/claude-sonnet-5")

 result = context.query_with_reasoning(
    "What are APT29's known TTPs against healthcare infrastructure, "
@@ -281,7 +281,7 @@ context.store(
    link_entities=True,
 )

-llm    = LiteLLM(model="anthropic/claude-sonnet-4-20250514")
+llm    = LiteLLM(model="anthropic/claude-sonnet-5")
 result = context.query_with_reasoning(
    "Trace the C2 infrastructure chain for APT29 operations targeting "
    "ITAR-controlled contractors in 2025. Include IP ranges, ASNs, and TTPs.",
@@ -351,7 +351,7 @@ Parent: wmiprvse.exe
 Sigma match: T1053.005 Scheduled Task/Job
 """

-llm    = LiteLLM(model="anthropic/claude-sonnet-4-20250514")
+llm    = LiteLLM(model="anthropic/claude-sonnet-5")
 triage = soc_context.query_with_reasoning(
    "Triage this SIEM alert and identify the correct response runbook:\n{}".format(alert_text),
    llm_provider=llm,
@@ -425,7 +425,7 @@ Patient: 68F, AF, CKD stage 3b (eGFR 32). On warfarin (INR target 2.0–3.0).
 Presenting for elective hip replacement. Concurrent: amiodarone 200mg, atorvastatin 40mg.
 """

-llm    = LiteLLM(model="anthropic/claude-sonnet-4-20250514")
+llm    = LiteLLM(model="anthropic/claude-sonnet-5")
 answer = clinical_context.query_with_reasoning(
    "What is the evidence-based warfarin bridging protocol for this patient "
    "given CKD and amiodarone interaction risk?\n\n{}".format(patient_context),
@@ -495,7 +495,7 @@ compliance_context.store(
    extract_relationships=True,
 )

-llm    = LiteLLM(model="anthropic/claude-sonnet-4-20250514")
+llm    = LiteLLM(model="anthropic/claude-sonnet-5")
 answer = compliance_context.query_with_reasoning(
    "Under Basel III CRE20, what are the RWA calculation requirements for "
    "commercial real estate exposures with LTV > 80%? "
@@ -275,20 +275,20 @@ print(data)

 **LiteLLM** is a universal adapter that provides a single interface to over 100 different LLM providers, including Anthropic Claude, Azure OpenAI, AWS Bedrock, Google Vertex AI, and local Ollama instances. It acts as a translation layer, converting your unified API calls into provider-specific requests, enabling easy switching between providers without code changes.

-`LiteLLM` is the Swiss Army knife. It wraps the `litellm` library, which speaks to every major provider using a unified completion API. The model string encodes both provider and model name: `"anthropic/claude-sonnet-4-20250514"`, `"azure/gpt-4o"`, `"bedrock/anthropic.claude-3-5-sonnet-20241022-v2:0"`, `"ollama/llama3.2"`. Change the string, change the provider — no other code changes needed.
+`LiteLLM` is the Swiss Army knife. It wraps the `litellm` library, which speaks to every major provider using a unified completion API. The model string encodes both provider and model name: `"anthropic/claude-sonnet-5"`, `"azure/gpt-4o"`, `"bedrock/anthropic.claude-sonnet-4-5-20250929-v1:0"`, `"ollama/llama3.2"`. Change the string, change the provider — no other code changes needed.

 ```python
 from semantica.llms import LiteLLM

 # Anthropic Claude — highest accuracy for complex reasoning
-llm = LiteLLM(model="anthropic/claude-sonnet-4-20250514")
+llm = LiteLLM(model="anthropic/claude-sonnet-5")
 # Reads ANTHROPIC_API_KEY from environment

 # Azure OpenAI — compliance and data-residency requirements
 llm = LiteLLM(model="azure/gpt-4o", api_key="YOUR_AZURE_KEY")

 # AWS Bedrock — existing cloud agreement, no new vendor
-llm = LiteLLM(model="bedrock/anthropic.claude-3-5-sonnet-20241022-v2:0")
+llm = LiteLLM(model="bedrock/anthropic.claude-sonnet-4-5-20250929-v1:0")

 # Google Vertex AI
 llm = LiteLLM(model="vertex_ai/gemini-1.5-pro")
@@ -306,7 +306,7 @@ The environment-variable convention for each provider: `ANTHROPIC_API_KEY`, `AZU
 import os

 PROVIDER_MAP = {
-    "prod":    "anthropic/claude-sonnet-4-20250514",
+    "prod":    "anthropic/claude-sonnet-5",
    "staging": "openai/gpt-4o-mini",
    "local":   "ollama/llama3.2",
    "azure":   "azure/gpt-4o",
@@ -378,7 +378,7 @@ print("FAST: {}  (conf={:.0%})".format(fast_result["response"], fast_result["con

 # Tier 2: deep answer with Claude if confidence is below threshold
 if fast_result["confidence"] < 0.85:
-    deep_llm = LiteLLM(model="anthropic/claude-sonnet-4-20250514")
+    deep_llm = LiteLLM(model="anthropic/claude-sonnet-5")
    deep_result = context.query_with_reasoning(
        query, llm_provider=deep_llm, max_results=15, max_hops=3
    )
@@ -574,7 +574,7 @@ print("TRIAGE: {} (conf={:.0%})".format(triage["response"], triage["confidence"]

 # Tier 2: escalate to Claude for deep analysis if Tier 1 is uncertain
 if triage["confidence"] < 0.88:
-    deep_llm = LiteLLM(model="anthropic/claude-sonnet-4-20250514")
+    deep_llm = LiteLLM(model="anthropic/claude-sonnet-5")
    deep = context.query_with_reasoning(
        "Full MITRE ATT&CK analysis of this alert: identify the attack chain, "
        "blast radius, affected systems, and recommended containment steps.",
@@ -630,7 +630,7 @@ for d in drugs:
 # trastuzumab (conf=0.98), pertuzumab (conf=0.97), docetaxel (conf=0.96)

 # Report synthesis with Claude — switch to azure/gpt-4o for HIPAA by changing one string
-report_llm = LiteLLM(model="anthropic/claude-sonnet-4-20250514")
+report_llm = LiteLLM(model="anthropic/claude-sonnet-5")
 # For HIPAA-constrained Azure deployment:
 # report_llm = LiteLLM(model="azure/gpt-4o", api_key="YOUR_AZURE_KEY")

@@ -682,7 +682,7 @@ question = (

 # Two-provider consensus — same query, same graph, different LLMs
 gpt4o  = OpenAI(model="gpt-4o", api_key="YOUR_OAI_KEY")
-claude = LiteLLM(model="anthropic/claude-sonnet-4-20250514")
+claude = LiteLLM(model="anthropic/claude-sonnet-5")

 answer_a = context.query_with_reasoning(question, llm_provider=gpt4o,  max_results=10)
 answer_b = context.query_with_reasoning(question, llm_provider=claude, max_results=10)
@@ -197,7 +197,7 @@ reasoning_agent.load("./pipeline/enriched_intel/")
 # All memories, graph nodes, and vector embeddings from both ingestion agents are now available.

 # Use a high-capability model for the synthesis step
-llm = LiteLLM(model="anthropic/claude-sonnet-4-20250514")
+llm = LiteLLM(model="anthropic/claude-sonnet-5")

 synthesis = reasoning_agent.query_with_reasoning(
    "Summarize the APT29 exploitation of CVE-2024-3400: affected products, "
@@ -428,7 +428,7 @@ tier1.store(

 # --- Tier 2: deep investigation when Tier 1 confidence is low ---
 if triage["confidence"] < 0.90:
-    deep_llm = LiteLLM(model="anthropic/claude-sonnet-4-20250514")
+    deep_llm = LiteLLM(model="anthropic/claude-sonnet-5")

    investigation = tier2.query_with_reasoning(
        "Full MITRE ATT&CK analysis of incident {}. "
@@ -533,7 +533,7 @@ t1.start(); t2.start()
 t1.join();  t2.join()

 # Chief agent synthesizes across literature and experimental data
-llm = LiteLLM(model="anthropic/claude-sonnet-4-20250514")
+llm = LiteLLM(model="anthropic/claude-sonnet-5")

 synthesis = chief.query_with_reasoning(
    "Identify the top two candidate compounds for KRAS G12C NSCLC that show "
@@ -576,7 +576,7 @@ credit_officer  = make_desk_agent()
 committee_chair = make_desk_agent()

 app_id = "LOAN-2025-88421"
-llm    = LiteLLM(model="anthropic/claude-sonnet-4-20250514")
+llm    = LiteLLM(model="anthropic/claude-sonnet-5")

 # --- Risk Desk: PD/LGD/EL analysis ---
 risk_desk.store(
@@ -477,7 +477,7 @@ regs = [
 ]

 # Use an LLM to extract the conceptual model from regulatory prose
-llm_gen  = LLMOntologyGenerator(provider="anthropic", model="claude-sonnet-4-20250514")
+llm_gen  = LLMOntologyGenerator(provider="anthropic", model="claude-sonnet-5")
 ontology = llm_gen.generate_ontology_from_text(
    "\n\n".join(r.text[:8000] for r in regs)  # token-safe excerpt per document
 )
@@ -129,7 +129,7 @@ from semantica.llms import Groq, OpenAI, LiteLLM, HuggingFaceLLM
    from semantica.llms import LiteLLM

    llm = LiteLLM(
-        model="anthropic/claude-sonnet-4-20250514",
+        model="anthropic/claude-sonnet-5",
        api_key=os.getenv("ANTHROPIC_API_KEY"),
        temperature=0.0,
    )
@@ -198,7 +198,7 @@ llm = Groq(api_key=os.getenv("GROQ_API_KEY"), model="llama-3.1-8b-instant")
 # Method 3: Multiple providers via LiteLLM
 providers = {
    "fast": LiteLLM(model="groq/llama-3.1-8b-instant", api_key=os.getenv("GROQ_API_KEY")),
-    "smart": LiteLLM(model="anthropic/claude-sonnet-4-20250514", api_key=os.getenv("ANTHROPIC_API_KEY"))
+    "smart": LiteLLM(model="anthropic/claude-sonnet-5", api_key=os.getenv("ANTHROPIC_API_KEY"))
 }
 ```

@@ -252,7 +252,7 @@ from semantica.llms import LiteLLM
 # pip install "semantica[llm-litellm]"

 # Anthropic Claude
-llm = LiteLLM(model="anthropic/claude-opus-4-5",         api_key=os.getenv("ANTHROPIC_API_KEY"))
+llm = LiteLLM(model="anthropic/claude-opus-4-7",         api_key=os.getenv("ANTHROPIC_API_KEY"))

 # Google Gemini
 llm = LiteLLM(model="gemini/gemini-1.5-pro",             api_key=os.getenv("GOOGLE_API_KEY"))
@@ -267,7 +267,7 @@ llm = LiteLLM(model="deepseek/deepseek-chat",            api_key=os.getenv("DEEP
 llm = LiteLLM(model="azure/gpt-4o",                      api_key=os.getenv("AZURE_API_KEY"))

 # AWS Bedrock
-llm = LiteLLM(model="bedrock/anthropic.claude-3-5-sonnet-20241022-v2:0")
+llm = LiteLLM(model="bedrock/anthropic.claude-sonnet-4-5-20250929-v1:0")

 # Novita AI
 llm = LiteLLM(model="novita/deepseek/deepseek-v3.2",     api_key=os.getenv("NOVITA_API_KEY"))
@@ -297,12 +297,12 @@ from semantica.llms import LiteLLM

 # Pattern: LiteLLM(model="<provider>/<model-name>")
 providers = {
-    "Anthropic":  LiteLLM(model="anthropic/claude-opus-4-5",       api_key=os.getenv("ANTHROPIC_API_KEY")),
+    "Anthropic":  LiteLLM(model="anthropic/claude-opus-4-7",       api_key=os.getenv("ANTHROPIC_API_KEY")),
    "Gemini":     LiteLLM(model="gemini/gemini-1.5-pro",            api_key=os.getenv("GOOGLE_API_KEY")),
    "Ollama":     LiteLLM(model="ollama/llama3.2:3b",               api_base="http://localhost:11434"),
    "DeepSeek":   LiteLLM(model="deepseek/deepseek-chat",           api_key=os.getenv("DEEPSEEK_API_KEY")),
    "Azure":      LiteLLM(model="azure/gpt-4o",                     api_key=os.getenv("AZURE_API_KEY")),
-    "Bedrock":    LiteLLM(model="bedrock/anthropic.claude-3-5-sonnet-20241022-v2:0"),
+    "Bedrock":    LiteLLM(model="bedrock/anthropic.claude-sonnet-4-5-20250929-v1:0"),
    "Cohere":     LiteLLM(model="cohere/command-r-plus",            api_key=os.getenv("COHERE_API_KEY")),
    "Novita AI":  LiteLLM(model="novita/deepseek/deepseek-v3.2",    api_key=os.getenv("NOVITA_API_KEY")),
 }
@@ -416,7 +416,7 @@ for text in texts:
 | :---------- | :--------------------------- | :----------- |
 | **Entity Extraction** | `Groq("llama-3.3-70b-versatile")` | Fast, good accuracy for structured tasks |
 | **Relation Extraction** | `OpenAI("gpt-4o")` | Best at complex relationship reasoning |
-| **Complex Analysis** | `LiteLLM("anthropic/claude-sonnet-4-20250514")` | Highest reasoning capability |
+| **Complex Analysis** | `LiteLLM("anthropic/claude-sonnet-5")` | Highest reasoning capability |
 | **High Volume/Cost** | `LiteLLM("deepseek/deepseek-chat")` | Lowest cost per token |

 ### Error Handling
@@ -22,6 +22,7 @@ icon: "sitemap"
 | `LLMOntologyGenerator` | LLM-powered ontology generation for complex domains |
 | `SHACLGenerator` | Generate SHACL shapes from an ontology or KG schema |
 | `OntologyValidator` | Validate any graph against SHACL shapes: returns `SHACLValidationReport` |
+| `OntologyQualityGate` | Run deterministic ontology/KG quality checks for CI |
 | `OWLGenerator` | Serialize ontologies to Turtle, RDF/XML, JSON-LD |
 | `NamespaceManager` | IRI generation, prefix management, and namespace binding |
 | `OntologyEvaluator` | Coverage, completeness, and granularity quality metrics |
@@ -81,9 +82,38 @@ engine.export_owl(ontology, "ontology.ttl", format="turtle")
 | :------ | :----------- |
 | `from_data(data)` | Run the 5-stage pipeline on entity/relationship data |
 | `validate_graph(kg, ontology=...)` | Check a knowledge graph against generated SHACL shapes |
+| `quality_check(ontology, graph_data=...)` | Return a deterministic quality report and CI-friendly pass/fail result |
 | `export_owl(ontology, path, format)` | Serialize to `"turtle"`, `"xml"`, or `"json-ld"` |
 | `evaluate(ontology, kg)` | Compute coverage, completeness, and granularity metrics |

+### Ontology Quality Gate
+
+Use the quality gate before export or deployment to catch structural issues
+without adding a runtime dependency:
+
+```python
+from semantica.ontology import ontology_quality_check
+
+report = ontology_quality_check(
+    ontology,
+    graph_data=kg,
+    thresholds={"min_coverage": 0.8},
+)
+
+if not report.passed:
+    for issue in report.issues:
+        print(issue.code, issue.message)
+```
+
+The report checks class/property coverage, orphan schema elements, domain and
+range references, and unresolved KG relationship endpoints. It includes
+machine-readable issue codes, severity, counts, metrics, and threshold
+failures. The first version reports findings only; it does not auto-fix data.
+
+### Thresholds
+
+`min_coverage` (default `0.0`) sets the minimum required `coverage` score, the average of class and property coverage from `0.0` to `1.0`; the gate fails below it. `max_errors` (default `0.0`) caps how many `error`/`critical` issues are allowed before the gate fails. `max_warnings` (default `None`) caps `warning` issues the same way, and `None` means warnings alone never fail the gate. `fail_on_warnings` is a separate parameter, not a `thresholds` key, passed to `OntologyQualityGate(...)` or `.check(...)` directly; when `True`, a single warning fails the gate regardless of `max_warnings`.
+
 ## OntologyGenerator (5-Stage Pipeline)

 **`OntologyGenerator`** auto-generates a formal ontology from your knowledge graph entities and relationships:
@@ -33,7 +33,9 @@
      "devDependencies": {
        "@babel/core": "^7.29.6",
        "@eslint/js": "^9.39.4",
+        "@testing-library/react": "^16.3.3",
        "@types/babel__core": "^7.20.5",
+        "@types/jsdom": "^21.1.7",
        "@types/node": "^24.12.0",
        "@types/react": "^19.2.14",
        "@types/react-dom": "^19.2.3",
@@ -43,12 +45,34 @@
        "eslint-plugin-react-hooks": "^7.0.1",
        "eslint-plugin-react-refresh": "^0.5.2",
        "globals": "^17.4.0",
+        "jsdom": "^26.1.0",
        "tsx": "^4.21.0",
        "typescript": "~5.9.3",
        "typescript-eslint": "^8.57.0",
        "vite": "^6.4.2"
      }
    },
+    "node_modules/@asamuzakjp/css-color": {
+      "version": "3.2.0",
+      "resolved": "https://registry.npmmirror.com/@asamuzakjp/css-color/-/css-color-3.2.0.tgz",
+      "integrity": "sha512-K1A6z8tS3XsmCMM86xoWdn7Fkdn9m6RSVtocUrJYIwZnFVkng/PvkEoWtOWmP+Scc6saYWHWZYbndEEXxl24jw==",
+      "dev": true,
+      "license": "MIT",
+      "dependencies": {
+        "@csstools/css-calc": "^2.1.3",
+        "@csstools/css-color-parser": "^3.0.9",
+        "@csstools/css-parser-algorithms": "^3.0.4",
+        "@csstools/css-tokenizer": "^3.0.3",
+        "lru-cache": "^10.4.3"
+      }
+    },
+    "node_modules/@asamuzakjp/css-color/node_modules/lru-cache": {
+      "version": "10.4.3",
+      "resolved": "https://registry.npmmirror.com/lru-cache/-/lru-cache-10.4.3.tgz",
+      "integrity": "sha512-JNAzZcXrCt42VGLuYz0zfAzDfAvJWW6AfYlDBQyDV5DClI2m5sAmK+OIO7s59XfsRsWHp02jAJrRadPRGTt6SQ==",
+      "dev": true,
+      "license": "ISC"
+    },
    "node_modules/@babel/code-frame": {
      "version": "7.29.0",
      "resolved": "https://registry.npmjs.org/@babel/code-frame/-/code-frame-7.29.0.tgz",
@@ -340,6 +364,121 @@
        "node": ">=6.9.0"
      }
    },
+    "node_modules/@csstools/color-helpers": {
+      "version": "5.1.0",
+      "resolved": "https://registry.npmmirror.com/@csstools/color-helpers/-/color-helpers-5.1.0.tgz",
+      "integrity": "sha512-S11EXWJyy0Mz5SYvRmY8nJYTFFd1LCNV+7cXyAgQtOOuzb4EsgfqDufL+9esx72/eLhsRdGZwaldu/h+E4t4BA==",
+      "dev": true,
+      "funding": [
+        {
+          "type": "github",
+          "url": "https://github.com/sponsors/csstools"
+        },
+        {
+          "type": "opencollective",
+          "url": "https://opencollective.com/csstools"
+        }
+      ],
+      "license": "MIT-0",
+      "engines": {
+        "node": ">=18"
+      }
+    },
+    "node_modules/@csstools/css-calc": {
+      "version": "2.1.4",
+      "resolved": "https://registry.npmmirror.com/@csstools/css-calc/-/css-calc-2.1.4.tgz",
+      "integrity": "sha512-3N8oaj+0juUw/1H3YwmDDJXCgTB1gKU6Hc/bB502u9zR0q2vd786XJH9QfrKIEgFlZmhZiq6epXl4rHqhzsIgQ==",
+      "dev": true,
+      "funding": [
+        {
+          "type": "github",
+          "url": "https://github.com/sponsors/csstools"
+        },
+        {
+          "type": "opencollective",
+          "url": "https://opencollective.com/csstools"
+        }
+      ],
+      "license": "MIT",
+      "engines": {
+        "node": ">=18"
+      },
+      "peerDependencies": {
+        "@csstools/css-parser-algorithms": "^3.0.5",
+        "@csstools/css-tokenizer": "^3.0.4"
+      }
+    },
+    "node_modules/@csstools/css-color-parser": {
+      "version": "3.1.0",
+      "resolved": "https://registry.npmmirror.com/@csstools/css-color-parser/-/css-color-parser-3.1.0.tgz",
+      "integrity": "sha512-nbtKwh3a6xNVIp/VRuXV64yTKnb1IjTAEEh3irzS+HkKjAOYLTGNb9pmVNntZ8iVBHcWDA2Dof0QtPgFI1BaTA==",
+      "dev": true,
+      "funding": [
+        {
+          "type": "github",
+          "url": "https://github.com/sponsors/csstools"
+        },
+        {
+          "type": "opencollective",
+          "url": "https://opencollective.com/csstools"
+        }
+      ],
+      "license": "MIT",
+      "dependencies": {
+        "@csstools/color-helpers": "^5.1.0",
+        "@csstools/css-calc": "^2.1.4"
+      },
+      "engines": {
+        "node": ">=18"
+      },
+      "peerDependencies": {
+        "@csstools/css-parser-algorithms": "^3.0.5",
+        "@csstools/css-tokenizer": "^3.0.4"
+      }
+    },
+    "node_modules/@csstools/css-parser-algorithms": {
+      "version": "3.0.5",
+      "resolved": "https://registry.npmmirror.com/@csstools/css-parser-algorithms/-/css-parser-algorithms-3.0.5.tgz",
+      "integrity": "sha512-DaDeUkXZKjdGhgYaHNJTV9pV7Y9B3b644jCLs9Upc3VeNGg6LWARAT6O+Q+/COo+2gg/bM5rhpMAtf70WqfBdQ==",
+      "dev": true,
+      "funding": [
+        {
+          "type": "github",
+          "url": "https://github.com/sponsors/csstools"
+        },
+        {
+          "type": "opencollective",
+          "url": "https://opencollective.com/csstools"
+        }
+      ],
+      "license": "MIT",
+      "engines": {
+        "node": ">=18"
+      },
+      "peerDependencies": {
+        "@csstools/css-tokenizer": "^3.0.4"
+      }
+    },
+    "node_modules/@csstools/css-tokenizer": {
+      "version": "3.0.4",
+      "resolved": "https://registry.npmmirror.com/@csstools/css-tokenizer/-/css-tokenizer-3.0.4.tgz",
+      "integrity": "sha512-Vd/9EVDiu6PPJt9yAh6roZP6El1xHrdvIVGjyBsHR0RYwNHgL7FJPyIIW4fANJNG6FtyZfvlRPpFI4ZM/lubvw==",
+      "dev": true,
+      "funding": [
+        {
+          "type": "github",
+          "url": "https://github.com/sponsors/csstools"
+        },
+        {
+          "type": "opencollective",
+          "url": "https://opencollective.com/csstools"
+        }
+      ],
+      "license": "MIT",
+      "engines": {
+        "node": ">=18"
+      }
+    },
    "node_modules/@egjs/hammerjs": {
      "version": "2.0.17",
      "resolved": "https://registry.npmjs.org/@egjs/hammerjs/-/hammerjs-2.0.17.tgz",
@@ -1473,6 +1612,63 @@
        "react": "^18 || ^19"
      }
    },
+    "node_modules/@testing-library/dom": {
+      "version": "10.4.1",
+      "resolved": "https://registry.npmmirror.com/@testing-library/dom/-/dom-10.4.1.tgz",
+      "integrity": "sha512-o4PXJQidqJl82ckFaXUeoAW+XysPLauYI43Abki5hABd853iMhitooc6znOnczgbTYmEP6U6/y1ZyKAIsvMKGg==",
+      "dev": true,
+      "license": "MIT",
+      "peer": true,
+      "dependencies": {
+        "@babel/code-frame": "^7.10.4",
+        "@babel/runtime": "^7.12.5",
+        "@types/aria-query": "^5.0.1",
+        "aria-query": "5.3.0",
+        "dom-accessibility-api": "^0.5.9",
+        "lz-string": "^1.5.0",
+        "picocolors": "1.1.1",
+        "pretty-format": "^27.0.2"
+      },
+      "engines": {
+        "node": ">=18"
+      }
+    },
+    "node_modules/@testing-library/react": {
+      "version": "16.3.3",
+      "resolved": "https://registry.npmmirror.com/@testing-library/react/-/react-16.3.3.tgz",
+      "integrity": "sha512-Uo193NgQbPMz6lrrhtRQQFcMC6Re/ELLFbbuVL30WDlZxlpZf9/lMHTAVxPRLw1q1iu9OJmR1c2BLiENRstdBg==",
+      "dev": true,
+      "license": "MIT",
+      "dependencies": {
+        "@babel/runtime": "^7.12.5"
+      },
+      "engines": {
+        "node": ">=18"
+      },
+      "peerDependencies": {
+        "@testing-library/dom": "^10.0.0",
+        "@types/react": "^18.0.0 || ^19.0.0",
+        "@types/react-dom": "^18.0.0 || ^19.0.0",
+        "react": "^18.0.0 || ^19.0.0",
+        "react-dom": "^18.0.0 || ^19.0.0"
+      },
+      "peerDependenciesMeta": {
+        "@types/react": {
+          "optional": true
+        },
+        "@types/react-dom": {
+          "optional": true
+        }
+      }
+    },
+    "node_modules/@types/aria-query": {
+      "version": "5.0.4",
+      "resolved": "https://registry.npmmirror.com/@types/aria-query/-/aria-query-5.0.4.tgz",
+      "integrity": "sha512-rfT93uj5s0PRL7EzccGMs3brplhcrghnDoV26NqKhCAS1hVo+WdNsPvE/yb6ilfr5hi2MEk6d5EWJTKdxg8jVw==",
+      "dev": true,
+      "license": "MIT",
+      "peer": true
+    },
    "node_modules/@types/babel__core": {
      "version": "7.20.5",
      "resolved": "https://registry.npmjs.org/@types/babel__core/-/babel__core-7.20.5.tgz",
@@ -1614,6 +1810,18 @@
        "@types/unist": "*"
      }
    },
+    "node_modules/@types/jsdom": {
+      "version": "21.1.7",
+      "resolved": "https://registry.npmmirror.com/@types/jsdom/-/jsdom-21.1.7.tgz",
+      "integrity": "sha512-yOriVnggzrnQ3a9OKOCxaVuSug3w3/SbOj5i7VwXWZEyUNl3bLF9V3MfxGbZKuwqJOQyRfqXyROBB1CoZLFWzA==",
+      "dev": true,
+      "license": "MIT",
+      "dependencies": {
+        "@types/node": "*",
+        "@types/tough-cookie": "*",
+        "parse5": "^7.0.0"
+      }
+    },
    "node_modules/@types/json-schema": {
      "version": "7.0.15",
      "resolved": "https://registry.npmjs.org/@types/json-schema/-/json-schema-7.0.15.tgz",
@@ -1665,6 +1873,13 @@
        "@types/react": "^19.2.0"
      }
    },
+    "node_modules/@types/tough-cookie": {
+      "version": "4.0.5",
+      "resolved": "https://registry.npmmirror.com/@types/tough-cookie/-/tough-cookie-4.0.5.tgz",
+      "integrity": "sha512-/Ad8+nIOV7Rl++6f1BdKxFSMgmoqEoYbHRpPcx3JEfv8VRsQe9Z4mCXeJBzxs7mbHY/XOZZuXlRNfhpVPbs6ZA==",
+      "dev": true,
+      "license": "MIT"
+    },
    "node_modules/@types/trusted-types": {
      "version": "2.0.7",
      "resolved": "https://registry.npmjs.org/@types/trusted-types/-/trusted-types-2.0.7.tgz",
@@ -2010,6 +2225,16 @@
        "acorn": "^6.0.0 || ^7.0.0 || ^8.0.0"
      }
    },
+    "node_modules/agent-base": {
+      "version": "7.1.4",
+      "resolved": "https://registry.npmmirror.com/agent-base/-/agent-base-7.1.4.tgz",
+      "integrity": "sha512-MnA+YT8fwfJPgBx3m60MNqakm30XOkyIoH1y6huTQvC0PwZG7ki8NacLBcrPbNoo8vEZy7Jpuk7+jMO+CUovTQ==",
+      "dev": true,
+      "license": "MIT",
+      "engines": {
+        "node": ">= 14"
+      }
+    },
    "node_modules/ajv": {
      "version": "6.15.0",
      "resolved": "https://registry.npmjs.org/ajv/-/ajv-6.15.0.tgz",
@@ -2027,6 +2252,42 @@
        "url": "https://github.com/sponsors/epoberezkin"
      }
    },
+    "node_modules/ansi-regex": {
+      "version": "5.0.1",
+      "resolved": "https://registry.npmmirror.com/ansi-regex/-/ansi-regex-5.0.1.tgz",
+      "integrity": "sha512-quJQXlTSUGL2LH9SUXo8VwsY4soanhgo6LNSm84E1LBcE8s3O0wpdiRzyR9z/ZZJMlMWv37qOOb9pdJlMUEKFQ==",
+      "dev": true,
+      "license": "MIT",
+      "peer": true,
+      "engines": {
+        "node": ">=8"
+      }
+    },
+    "node_modules/ansi-styles": {
+      "version": "5.2.0",
+      "resolved": "https://registry.npmmirror.com/ansi-styles/-/ansi-styles-5.2.0.tgz",
+      "integrity": "sha512-Cxwpt2SfTzTtXcfOlzGEee8O+c+MmUgGrNiBcXnuWxuFJHe6a5Hz7qwhwe5OgaSYI0IJvkLqWX1ASG+cJOkEiA==",
+      "dev": true,
+      "license": "MIT",
+      "peer": true,
+      "engines": {
+        "node": ">=10"
+      },
+      "funding": {
+        "url": "https://github.com/chalk/ansi-styles?sponsor=1"
+      }
+    },
+    "node_modules/aria-query": {
+      "version": "5.3.0",
+      "resolved": "https://registry.npmmirror.com/aria-query/-/aria-query-5.3.0.tgz",
+      "integrity": "sha512-b0P0sZPKtyu8HkeRAfCq0IfURZK+SuwMjY1UXGBU27wpAiTwQAIlq56IbIO+ytk/JjS1fMR14ee5WBBfKi5J6A==",
+      "dev": true,
+      "license": "Apache-2.0",
+      "peer": true,
+      "dependencies": {
+        "dequal": "^2.0.3"
+      }
+    },
    "node_modules/attr-accept": {
      "version": "2.2.5",
      "resolved": "https://registry.npmjs.org/attr-accept/-/attr-accept-2.2.5.tgz",
@@ -2259,6 +2520,20 @@
      "license": "MIT",
      "peer": true
    },
+    "node_modules/cssstyle": {
+      "version": "4.6.0",
+      "resolved": "https://registry.npmmirror.com/cssstyle/-/cssstyle-4.6.0.tgz",
+      "integrity": "sha512-2z+rWdzbbSZv6/rhtvzvqeZQHrBaqgogqt85sqFNbabZOuFbCVFb8kPeEtZjiKkbrm395irpNKiYeFeLiQnFPg==",
+      "dev": true,
+      "license": "MIT",
+      "dependencies": {
+        "@asamuzakjp/css-color": "^3.2.0",
+        "rrweb-cssom": "^0.8.0"
+      },
+      "engines": {
+        "node": ">=18"
+      }
+    },
    "node_modules/csstype": {
      "version": "3.2.3",
      "resolved": "https://registry.npmjs.org/csstype/-/csstype-3.2.3.tgz",
@@ -2370,6 +2645,20 @@
        "node": ">=12"
      }
    },
+    "node_modules/data-urls": {
+      "version": "5.0.0",
+      "resolved": "https://registry.npmmirror.com/data-urls/-/data-urls-5.0.0.tgz",
+      "integrity": "sha512-ZYP5VBHshaDAiVZxjbRVcFJpc+4xGgT0bK3vzy1HLN8jTO975HEbuYzZJcHoQEY5K1a0z8YayJkyVETa08eNTg==",
+      "dev": true,
+      "license": "MIT",
+      "dependencies": {
+        "whatwg-mimetype": "^4.0.0",
+        "whatwg-url": "^14.0.0"
+      },
+      "engines": {
+        "node": ">=18"
+      }
+    },
    "node_modules/debug": {
      "version": "4.4.3",
      "resolved": "https://registry.npmjs.org/debug/-/debug-4.4.3.tgz",
@@ -2387,6 +2676,13 @@
        }
      }
    },
+    "node_modules/decimal.js": {
+      "version": "10.6.0",
+      "resolved": "https://registry.npmmirror.com/decimal.js/-/decimal.js-10.6.0.tgz",
+      "integrity": "sha512-YpgQiITW3JXGntzdUmyUR1V812Hn8T1YVXhCu+wO3OpS4eU9l4YdD3qjyiKdV6mvV29zapkMeD390UVEf2lkUg==",
+      "dev": true,
+      "license": "MIT"
+    },
    "node_modules/decode-named-character-reference": {
      "version": "1.3.0",
      "resolved": "https://registry.npmjs.org/decode-named-character-reference/-/decode-named-character-reference-1.3.0.tgz",
@@ -2449,6 +2745,14 @@
        "@babel/runtime": "^7.9.2"
      }
    },
+    "node_modules/dom-accessibility-api": {
+      "version": "0.5.16",
+      "resolved": "https://registry.npmmirror.com/dom-accessibility-api/-/dom-accessibility-api-0.5.16.tgz",
+      "integrity": "sha512-X7BJ2yElsnOJ30pZF4uIIDfBEVgF4XEBxL9Bxhy6dnrm5hkzqmsWHGTiHqRiITNhMyFLyAiWndIJP7Z1NTteDg==",
+      "dev": true,
+      "license": "MIT",
+      "peer": true
+    },
    "node_modules/dompurify": {
      "version": "3.4.13",
      "resolved": "https://registry.npmjs.org/dompurify/-/dompurify-3.4.13.tgz",
@@ -2466,6 +2770,19 @@
      "dev": true,
      "license": "ISC"
    },
+    "node_modules/entities": {
+      "version": "6.0.1",
+      "resolved": "https://registry.npmmirror.com/entities/-/entities-6.0.1.tgz",
+      "integrity": "sha512-aN97NXWF6AWBTahfVOIrB/NShkzi5H7F9r1s9mD3cDj4Ko5f2qhhVoYMibXF7GlLveb/D2ioWay8lxI97Ven3g==",
+      "dev": true,
+      "license": "BSD-2-Clause",
+      "engines": {
+        "node": ">=0.12"
+      },
+      "funding": {
+        "url": "https://github.com/fb55/entities?sponsor=1"
+      }
+    },
    "node_modules/esbuild": {
      "version": "0.28.1",
      "resolved": "https://registry.npmjs.org/esbuild/-/esbuild-0.28.1.tgz",
@@ -3058,6 +3375,19 @@
        "react-is": "^16.7.0"
      }
    },
+    "node_modules/html-encoding-sniffer": {
+      "version": "4.0.0",
+      "resolved": "https://registry.npmmirror.com/html-encoding-sniffer/-/html-encoding-sniffer-4.0.0.tgz",
+      "integrity": "sha512-Y22oTqIU4uuPgEemfz7NDJz6OeKf12Lsu+QC+s3BVpda64lTiMYCyGwg5ki4vFxkMwQdeZDl2adZoqUgdFuTgQ==",
+      "dev": true,
+      "license": "MIT",
+      "dependencies": {
+        "whatwg-encoding": "^3.1.1"
+      },
+      "engines": {
+        "node": ">=18"
+      }
+    },
    "node_modules/html-url-attributes": {
      "version": "3.0.1",
      "resolved": "https://registry.npmjs.org/html-url-attributes/-/html-url-attributes-3.0.1.tgz",
@@ -3068,6 +3398,47 @@
        "url": "https://opencollective.com/unified"
      }
    },
+    "node_modules/http-proxy-agent": {
+      "version": "7.0.2",
+      "resolved": "https://registry.npmmirror.com/http-proxy-agent/-/http-proxy-agent-7.0.2.tgz",
+      "integrity": "sha512-T1gkAiYYDWYx3V5Bmyu7HcfcvL7mUrTWiM6yOfa3PIphViJ/gFPbvidQ+veqSOHci/PxBcDabeUNCzpOODJZig==",
+      "dev": true,
+      "license": "MIT",
+      "dependencies": {
+        "agent-base": "^7.1.0",
+        "debug": "^4.3.4"
+      },
+      "engines": {
+        "node": ">= 14"
+      }
+    },
+    "node_modules/https-proxy-agent": {
+      "version": "7.0.6",
+      "resolved": "https://registry.npmmirror.com/https-proxy-agent/-/https-proxy-agent-7.0.6.tgz",
+      "integrity": "sha512-vK9P5/iUfdl95AI+JVyUuIcVtd4ofvtrOr3HNtM2yxC9bnMbEdp3x01OhQNnjb8IJYi38VlTE3mBXwcfvywuSw==",
+      "dev": true,
+      "license": "MIT",
+      "dependencies": {
+        "agent-base": "^7.1.2",
+        "debug": "4"
+      },
+      "engines": {
+        "node": ">= 14"
+      }
+    },
+    "node_modules/iconv-lite": {
+      "version": "0.6.3",
+      "resolved": "https://registry.npmmirror.com/iconv-lite/-/iconv-lite-0.6.3.tgz",
+      "integrity": "sha512-4fCk79wshMdzMp2rH06qWrJE4iolqLhCUH+OiuIgU++RB0+94NlDL81atO7GX55uUKueo0txHNtvEyI6D7WdMw==",
+      "dev": true,
+      "license": "MIT",
+      "dependencies": {
+        "safer-buffer": ">= 2.1.2 < 3.0.0"
+      },
+      "engines": {
+        "node": ">=0.10.0"
+      }
+    },
    "node_modules/ignore": {
      "version": "5.3.2",
      "resolved": "https://registry.npmjs.org/ignore/-/ignore-5.3.2.tgz",
@@ -3173,6 +3544,13 @@
        "url": "https://github.com/sponsors/sindresorhus"
      }
    },
+    "node_modules/is-potential-custom-element-name": {
+      "version": "1.0.1",
+      "resolved": "https://registry.npmmirror.com/is-potential-custom-element-name/-/is-potential-custom-element-name-1.0.1.tgz",
+      "integrity": "sha512-bCYeRA2rVibKZd+s2625gGnGF/t7DSqDs4dP7CrLA1m7jKWz6pps0LpYLJN8Q64HtmPKJ1hrN3nzPNKFEKOUiQ==",
+      "dev": true,
+      "license": "MIT"
+    },
    "node_modules/isexe": {
      "version": "2.0.0",
      "resolved": "https://registry.npmjs.org/isexe/-/isexe-2.0.0.tgz",
@@ -3186,6 +3564,46 @@
      "integrity": "sha512-RdJUflcE3cUzKiMqQgsCu06FPu9UdIJO0beYbPhHN4k6apgJtifcoCtT9bcxOpYBtpD2kCM6Sbzg4CausW/PKQ==",
      "license": "MIT"
    },
+    "node_modules/jsdom": {
+      "version": "26.1.0",
+      "resolved": "https://registry.npmmirror.com/jsdom/-/jsdom-26.1.0.tgz",
+      "integrity": "sha512-Cvc9WUhxSMEo4McES3P7oK3QaXldCfNWp7pl2NNeiIFlCoLr3kfq9kb1fxftiwk1FLV7CvpvDfonxtzUDeSOPg==",
+      "dev": true,
+      "license": "MIT",
+      "dependencies": {
+        "cssstyle": "^4.2.1",
+        "data-urls": "^5.0.0",
+        "decimal.js": "^10.5.0",
+        "html-encoding-sniffer": "^4.0.0",
+        "http-proxy-agent": "^7.0.2",
+        "https-proxy-agent": "^7.0.6",
+        "is-potential-custom-element-name": "^1.0.1",
+        "nwsapi": "^2.2.16",
+        "parse5": "^7.2.1",
+        "rrweb-cssom": "^0.8.0",
+        "saxes": "^6.0.0",
+        "symbol-tree": "^3.2.4",
+        "tough-cookie": "^5.1.1",
+        "w3c-xmlserializer": "^5.0.0",
+        "webidl-conversions": "^7.0.0",
+        "whatwg-encoding": "^3.1.1",
+        "whatwg-mimetype": "^4.0.0",
+        "whatwg-url": "^14.1.1",
+        "ws": "^8.18.0",
+        "xml-name-validator": "^5.0.0"
+      },
+      "engines": {
+        "node": ">=18"
+      },
+      "peerDependencies": {
+        "canvas": "^3.0.0"
+      },
+      "peerDependenciesMeta": {
+        "canvas": {
+          "optional": true
+        }
+      }
+    },
    "node_modules/jsesc": {
      "version": "3.1.0",
      "resolved": "https://registry.npmjs.org/jsesc/-/jsesc-3.1.0.tgz",
@@ -3321,6 +3739,17 @@
        "react": "^16.5.1 || ^17.0.0 || ^18.0.0 || ^19.0.0"
      }
    },
+    "node_modules/lz-string": {
+      "version": "1.5.0",
+      "resolved": "https://registry.npmmirror.com/lz-string/-/lz-string-1.5.0.tgz",
+      "integrity": "sha512-h5bgJWpxJNswbU7qCrV0tIKQCaS3blPDrqKWx+QxzuzL1zGUzij9XCWLrSLsJPu5t+eWA/ycetzYAO5IOMcWAQ==",
+      "dev": true,
+      "license": "MIT",
+      "peer": true,
+      "bin": {
+        "lz-string": "bin/bin.js"
+      }
+    },
    "node_modules/markdown-table": {
      "version": "3.0.4",
      "resolved": "https://registry.npmjs.org/markdown-table/-/markdown-table-3.0.4.tgz",
@@ -4283,6 +4712,13 @@
        "node": ">=18"
      }
    },
+    "node_modules/nwsapi": {
+      "version": "2.2.27",
+      "resolved": "https://registry.npmmirror.com/nwsapi/-/nwsapi-2.2.27.tgz",
+      "integrity": "sha512-gQPNF78qebCQ6tvVFBYrvJdBNOrYZm90ZlXgpIFm06p6qHDHq/XC4TnJftN6OMbxVE0UTBAoRgcsDeJBBooITw==",
+      "dev": true,
+      "license": "MIT"
+    },
    "node_modules/object-assign": {
      "version": "4.1.1",
      "resolved": "https://registry.npmjs.org/object-assign/-/object-assign-4.1.1.tgz",
@@ -4382,6 +4818,19 @@
      "integrity": "sha512-CmBKiL6NNo/OqgmMn95Fk9Whlp2mtvIv+KNpQKN2F4SjvrEesubTRWGYSg+BnWZOnlCaSTU1sMpsBOzgbYhnsA==",
      "license": "MIT"
    },
+    "node_modules/parse5": {
+      "version": "7.3.0",
+      "resolved": "https://registry.npmmirror.com/parse5/-/parse5-7.3.0.tgz",
+      "integrity": "sha512-IInvU7fabl34qmi9gY8XOVxhYyMyuH2xUNpb2q8/Y+7552KlejkRvqvD19nMoUW/uQGGbqNpA6Tufu5FL5BZgw==",
+      "dev": true,
+      "license": "MIT",
+      "dependencies": {
+        "entities": "^6.0.0"
+      },
+      "funding": {
+        "url": "https://github.com/inikulin/parse5?sponsor=1"
+      }
+    },
    "node_modules/path-exists": {
      "version": "4.0.0",
      "resolved": "https://registry.npmjs.org/path-exists/-/path-exists-4.0.0.tgz",
@@ -4505,6 +4954,30 @@
        "node": ">= 0.8.0"
      }
    },
+    "node_modules/pretty-format": {
+      "version": "27.5.1",
+      "resolved": "https://registry.npmmirror.com/pretty-format/-/pretty-format-27.5.1.tgz",
+      "integrity": "sha512-Qb1gy5OrP5+zDf2Bvnzdl3jsTf1qXVMazbvCoKhtKqVs4/YK4ozX4gKQJJVyNe+cajNPn0KoC0MC3FUmaHWEmQ==",
+      "dev": true,
+      "license": "MIT",
+      "peer": true,
+      "dependencies": {
+        "ansi-regex": "^5.0.1",
+        "ansi-styles": "^5.0.0",
+        "react-is": "^17.0.1"
+      },
+      "engines": {
+        "node": "^10.13.0 || ^12.13.0 || ^14.15.0 || >=15.0.0"
+      }
+    },
+    "node_modules/pretty-format/node_modules/react-is": {
+      "version": "17.0.2",
+      "resolved": "https://registry.npmmirror.com/react-is/-/react-is-17.0.2.tgz",
+      "integrity": "sha512-w2GsyukL62IJnlaff/nRegPQR94C/XXamvMWmSHRJ4y7Ts/4ocGRmTHvOs8PSE6pB3dWOrD/nueuU5sduBsQ4w==",
+      "dev": true,
+      "license": "MIT",
+      "peer": true
+    },
    "node_modules/prop-types": {
      "version": "15.8.1",
      "resolved": "https://registry.npmjs.org/prop-types/-/prop-types-15.8.1.tgz",
@@ -4817,6 +5290,33 @@
        "fsevents": "~2.3.2"
      }
    },
+    "node_modules/rrweb-cssom": {
+      "version": "0.8.0",
+      "resolved": "https://registry.npmmirror.com/rrweb-cssom/-/rrweb-cssom-0.8.0.tgz",
+      "integrity": "sha512-guoltQEx+9aMf2gDZ0s62EcV8lsXR+0w8915TC3ITdn2YueuNjdAYh/levpU9nFaoChh9RUS5ZdQMrKfVEN9tw==",
+      "dev": true,
+      "license": "MIT"
+    },
+    "node_modules/safer-buffer": {
+      "version": "2.1.2",
+      "resolved": "https://registry.npmmirror.com/safer-buffer/-/safer-buffer-2.1.2.tgz",
+      "integrity": "sha512-YZo3K82SD7Riyi0E1EQPojLz7kpepnSQI9IyPbHHg1XXXevb5dJI7tpyN2ADxGcQbHG7vcyRHk0cbwqcQriUtg==",
+      "dev": true,
+      "license": "MIT"
+    },
+    "node_modules/saxes": {
+      "version": "6.0.0",
+      "resolved": "https://registry.npmmirror.com/saxes/-/saxes-6.0.0.tgz",
+      "integrity": "sha512-xAg7SOnEhrm5zI3puOOKyy1OMcMlIJZYNJY7xLBwSze0UjhPLnWfj2GF2EpT0jmzaJKIWKHLsaSSajf35bcYnA==",
+      "dev": true,
+      "license": "ISC",
+      "dependencies": {
+        "xmlchars": "^2.2.0"
+      },
+      "engines": {
+        "node": ">=v12.22.7"
+      }
+    },
    "node_modules/scheduler": {
      "version": "0.27.0",
      "resolved": "https://registry.npmjs.org/scheduler/-/scheduler-0.27.0.tgz",
@@ -4924,6 +5424,13 @@
        "inline-style-parser": "0.2.7"
      }
    },
+    "node_modules/symbol-tree": {
+      "version": "3.2.4",
+      "resolved": "https://registry.npmmirror.com/symbol-tree/-/symbol-tree-3.2.4.tgz",
+      "integrity": "sha512-9QNk5KwDF+Bvz+PyObkmSYjI5ksVUYtjW7AU22r2NKcfLJcXp96hkDWU3+XndOsUb+AQ9QhfzfCT2O+CNWT5Tw==",
+      "dev": true,
+      "license": "MIT"
+    },
    "node_modules/tinyglobby": {
      "version": "0.2.16",
      "resolved": "https://registry.npmjs.org/tinyglobby/-/tinyglobby-0.2.16.tgz",
@@ -4941,6 +5448,52 @@
        "url": "https://github.com/sponsors/SuperchupuDev"
      }
    },
+    "node_modules/tldts": {
+      "version": "6.1.86",
+      "resolved": "https://registry.npmmirror.com/tldts/-/tldts-6.1.86.tgz",
+      "integrity": "sha512-WMi/OQ2axVTf/ykqCQgXiIct+mSQDFdH2fkwhPwgEwvJ1kSzZRiinb0zF2Xb8u4+OqPChmyI6MEu4EezNJz+FQ==",
+      "dev": true,
+      "license": "MIT",
+      "dependencies": {
+        "tldts-core": "^6.1.86"
+      },
+      "bin": {
+        "tldts": "bin/cli.js"
+      }
+    },
+    "node_modules/tldts-core": {
+      "version": "6.1.86",
+      "resolved": "https://registry.npmmirror.com/tldts-core/-/tldts-core-6.1.86.tgz",
+      "integrity": "sha512-Je6p7pkk+KMzMv2XXKmAE3McmolOQFdxkKw0R8EYNr7sELW46JqnNeTX8ybPiQgvg1ymCoF8LXs5fzFaZvJPTA==",
+      "dev": true,
+      "license": "MIT"
+    },
+    "node_modules/tough-cookie": {
+      "version": "5.1.2",
+      "resolved": "https://registry.npmmirror.com/tough-cookie/-/tough-cookie-5.1.2.tgz",
+      "integrity": "sha512-FVDYdxtnj0G6Qm/DhNPSb8Ju59ULcup3tuJxkFb5K8Bv2pUXILbf0xZWU8PX8Ov19OXljbUyveOFwRMwkXzO+A==",
+      "dev": true,
+      "license": "BSD-3-Clause",
+      "dependencies": {
+        "tldts": "^6.1.32"
+      },
+      "engines": {
+        "node": ">=16"
+      }
+    },
+    "node_modules/tr46": {
+      "version": "5.1.1",
+      "resolved": "https://registry.npmmirror.com/tr46/-/tr46-5.1.1.tgz",
+      "integrity": "sha512-hdF5ZgjTqgAntKkklYw0R03MG2x/bSzTtkxmIRw/sTNV8YXsCJ1tfLAX23lhxhHJlEf3CRCOCGGWw3vI3GaSPw==",
+      "dev": true,
+      "license": "MIT",
+      "dependencies": {
+        "punycode": "^2.3.1"
+      },
+      "engines": {
+        "node": ">=18"
+      }
+    },
    "node_modules/trim-lines": {
      "version": "3.0.1",
      "resolved": "https://registry.npmjs.org/trim-lines/-/trim-lines-3.0.1.tgz",
@@ -5364,6 +5917,67 @@
        }
      }
    },
+    "node_modules/w3c-xmlserializer": {
+      "version": "5.0.0",
+      "resolved": "https://registry.npmmirror.com/w3c-xmlserializer/-/w3c-xmlserializer-5.0.0.tgz",
+      "integrity": "sha512-o8qghlI8NZHU1lLPrpi2+Uq7abh4GGPpYANlalzWxyWteJOCsr/P+oPBA49TOLu5FTZO4d3F9MnWJfiMo4BkmA==",
+      "dev": true,
+      "license": "MIT",
+      "dependencies": {
+        "xml-name-validator": "^5.0.0"
+      },
+      "engines": {
+        "node": ">=18"
+      }
+    },
+    "node_modules/webidl-conversions": {
+      "version": "7.0.0",
+      "resolved": "https://registry.npmmirror.com/webidl-conversions/-/webidl-conversions-7.0.0.tgz",
+      "integrity": "sha512-VwddBukDzu71offAQR975unBIGqfKZpM+8ZX6ySk8nYhVoo5CYaZyzt3YBvYtRtO+aoGlqxPg/B87NGVZ/fu6g==",
+      "dev": true,
+      "license": "BSD-2-Clause",
+      "engines": {
+        "node": ">=12"
+      }
+    },
+    "node_modules/whatwg-encoding": {
+      "version": "3.1.1",
+      "resolved": "https://registry.npmmirror.com/whatwg-encoding/-/whatwg-encoding-3.1.1.tgz",
+      "integrity": "sha512-6qN4hJdMwfYBtE3YBTTHhoeuUrDBPZmbQaxWAqSALV/MeEnR5z1xd8UKud2RAkFoPkmB+hli1TZSnyi84xz1vQ==",
+      "deprecated": "Use @exodus/bytes instead for a more spec-conformant and faster implementation",
+      "dev": true,
+      "license": "MIT",
+      "dependencies": {
+        "iconv-lite": "0.6.3"
+      },
+      "engines": {
+        "node": ">=18"
+      }
+    },
+    "node_modules/whatwg-mimetype": {
+      "version": "4.0.0",
+      "resolved": "https://registry.npmmirror.com/whatwg-mimetype/-/whatwg-mimetype-4.0.0.tgz",
+      "integrity": "sha512-QaKxh0eNIi2mE9p2vEdzfagOKHCcj1pJ56EEHGQOVxp8r9/iszLUUV7v89x9O1p/T+NlTM5W7jW6+cz4Fq1YVg==",
+      "dev": true,
+      "license": "MIT",
+      "engines": {
+        "node": ">=18"
+      }
+    },
+    "node_modules/whatwg-url": {
+      "version": "14.2.0",
+      "resolved": "https://registry.npmmirror.com/whatwg-url/-/whatwg-url-14.2.0.tgz",
+      "integrity": "sha512-De72GdQZzNTUBBChsXueQUnPKDkg/5A5zp7pFDuQAj5UFoENpiACU0wlCvzpAGnTkj++ihpKwKyYewn/XNUbKw==",
+      "dev": true,
+      "license": "MIT",
+      "dependencies": {
+        "tr46": "^5.1.0",
+        "webidl-conversions": "^7.0.0"
+      },
+      "engines": {
+        "node": ">=18"
+      }
+    },
    "node_modules/which": {
      "version": "2.0.2",
      "resolved": "https://registry.npmjs.org/which/-/which-2.0.2.tgz",
@@ -5390,6 +6004,45 @@
        "node": ">=0.10.0"
      }
    },
+    "node_modules/ws": {
+      "version": "8.21.3",
+      "resolved": "https://registry.npmmirror.com/ws/-/ws-8.21.3.tgz",
+      "integrity": "sha512-201TZ/kPWxoPr/OKWjquZR1SWKXcvxdH+e1xrx89b3YbmzLMFCLfnaG1HFIgWzJOEWZ7MvpK++odZufgYR50Rw==",
+      "dev": true,
+      "license": "MIT",
+      "engines": {
+        "node": ">=10.0.0"
+      },
+      "peerDependencies": {
+        "bufferutil": "^4.0.1",
+        "utf-8-validate": ">=5.0.2"
+      },
+      "peerDependenciesMeta": {
+        "bufferutil": {
+          "optional": true
+        },
+        "utf-8-validate": {
+          "optional": true
+        }
+      }
+    },
+    "node_modules/xml-name-validator": {
+      "version": "5.0.0",
+      "resolved": "https://registry.npmmirror.com/xml-name-validator/-/xml-name-validator-5.0.0.tgz",
+      "integrity": "sha512-EvGK8EJ3DhaHfbRlETOWAS5pO9MZITeauHKJyb8wyajUfQUenkIg2MvLDTZ4T/TgIcm3HU0TFBgWWboAZ30UHg==",
+      "dev": true,
+      "license": "Apache-2.0",
+      "engines": {
+        "node": ">=18"
+      }
+    },
+    "node_modules/xmlchars": {
+      "version": "2.2.0",
+      "resolved": "https://registry.npmmirror.com/xmlchars/-/xmlchars-2.2.0.tgz",
+      "integrity": "sha512-JZnDKK8B0RCDw84FNdDAIpZK+JuJw+s7Lz8nksI7SIuU3UXJJslUthsi+uWBUYOwPFwW7W7PRLRfUKpxjtjFCw==",
+      "dev": true,
+      "license": "MIT"
+    },
    "node_modules/xss": {
      "version": "1.0.15",
      "resolved": "https://registry.npmjs.org/xss/-/xss-1.0.15.tgz",
@@ -9,6 +9,7 @@
    "lint": "eslint .",
    "preview": "vite preview",
    "test:graph-store": "node --test tests/graphStore.multi-edge.test.mjs",
+    "test:graph-workspace": "node --import tsx --test tests/markdownContentViewer.test.ts tests/markdownEditorInteraction.test.tsx tests/markdownEditorState.test.ts tests/nodeMarkdownSync.test.ts tests/graphSceneState.display.test.ts tests/temporalLifecycle.test.ts tests/deterministicExplorerRendering.test.ts tests/explorerCapabilities.test.tsx tests/smallGraphLayout.test.ts tests/realtimeGraphAttributes.test.ts",
    "test:graph-workspace": "node --import tsx --test tests/markdownContentViewer.test.ts tests/graphSceneState.display.test.ts tests/temporalLifecycle.test.ts tests/deterministicExplorerRendering.test.ts tests/smallGraphLayout.test.ts tests/realtimeGraphAttributes.test.ts tests/ontologyEditorModel.test.ts",
    "test:deterministic-e2e": "node --import tsx --test tests/deterministicExplorerRendering.e2e.ts",
    "test:plugin-registry": "node --import tsx --test tests/pluginRegistry.temporal.test.mjs"
@@ -39,7 +40,9 @@
  "devDependencies": {
    "@babel/core": "^7.29.6",
    "@eslint/js": "^9.39.4",
+    "@testing-library/react": "^16.3.3",
    "@types/babel__core": "^7.20.5",
+    "@types/jsdom": "^21.1.7",
    "@types/node": "^24.12.0",
    "@types/react": "^19.2.14",
    "@types/react-dom": "^19.2.3",
@@ -49,6 +52,7 @@
    "eslint-plugin-react-hooks": "^7.0.1",
    "eslint-plugin-react-refresh": "^0.5.2",
    "globals": "^17.4.0",
+    "jsdom": "^26.1.0",
    "tsx": "^4.21.0",
    "typescript": "~5.9.3",
    "typescript-eslint": "^8.57.0",
@@ -17,10 +17,13 @@ import {
  type LucideIcon,
 } from 'lucide-react';
 import { ErrorBoundary } from './ErrorBoundary';
+import { ExploreWorkspaceTabs, type ExploreView } from './ExploreWorkspaceTabs';
+import { fetchAgentMemoryAvailability } from './explorerCapabilities';

 const DecisionWorkspace = lazy(() => import('./workspaces/DecisionWorkspace/DecisionWorkspace').then((module) => ({ default: module.DecisionWorkspace })));
 const DiffMergeWorkspace = lazy(() => import('./workspaces/DiffMergeWorkspace/DiffMergeWorkspace').then((module) => ({ default: module.DiffMergeWorkspace })));
 const GraphWorkspace = lazy(() => import('./workspaces/GraphWorkspace/GraphWorkspace').then((module) => ({ default: module.GraphWorkspace })));
+const MemoryWorkspace = lazy(() => import('./workspaces/MemoryWorkspace').then((module) => ({ default: module.MemoryWorkspace })));
 const ImportExportWorkspace = lazy(() => import('./workspaces/ImportExportWorkspace/ImportExportWorkspace').then((module) => ({ default: module.ImportExportWorkspace })));
 const LineageDiagram = lazy(() => import('./workspaces/LineageWorkspace/LineageDiagram').then((module) => ({ default: module.LineageDiagram })));
 const ReasoningWorkspace = lazy(() => import('./workspaces/ReasoningWorkspace').then((module) => ({ default: module.ReasoningWorkspace })));
@@ -33,7 +36,6 @@ const OntologySummaryTab = lazy(() => import('./workspaces/ManageWorkspace/Ontol
 const OntologyWorkspace = lazy(() => import('./workspaces/OntologyWorkspace').then((module) => ({ default: module.OntologyWorkspace })));

 type WorkspaceId = 'welcome' | 'explore' | 'analyze' | 'decisions' | 'enrich' | 'manage' | 'ontology-hub';
-type ExploreView = 'graph' | 'vocabulary';
 type AnalyzeView = 'sparql' | 'reasoning';
 type EnrichView = 'import' | 'merge' | 'registry' | 'resolve';
 type ManageView = 'lineage' | 'kg-overview' | 'ontology';
@@ -1791,6 +1793,37 @@ export default function App() {
  const [enrichView, setEnrichView] = useState<EnrichView>('import');
  const [manageView, setManageView] = useState<ManageView>('lineage');
  const [graphFocusRequest, setGraphFocusRequest] = useState<{ nodeId: string; token: number } | null>(null);
+  const [exploreDraftDirty, setExploreDraftDirty] = useState(false);
+  const [agentMemoryAvailable, setAgentMemoryAvailable] = useState(false);
+
+  useEffect(() => {
+    let active = true;
+    void fetchAgentMemoryAvailability().then((available) => {
+      if (active) setAgentMemoryAvailable(available);
+    });
+    return () => {
+      active = false;
+    };
+  }, []);
+
+  const confirmDiscardExploreDraft = () => (
+    !exploreDraftDirty
+    || window.confirm("Discard the unapplied Markdown draft and leave this resource?")
+  );
+
+  const switchExploreView = (nextView: ExploreView) => {
+    if (nextView === exploreView) return;
+    if (!confirmDiscardExploreDraft()) return;
+    setExploreDraftDirty(false);
+    setExploreView(nextView);
+  };
+
+  const switchWorkspace = (nextWorkspace: WorkspaceId) => {
+    if (nextWorkspace === activeWorkspace) return;
+    if (activeWorkspace === "explore" && !confirmDiscardExploreDraft()) return;
+    setExploreDraftDirty(false);
+    setActiveWorkspace(nextWorkspace);
+  };


  const renderWorkspace = () => {
@@ -1823,18 +1856,15 @@ export default function App() {
      return (
        <WorkspaceShell
          title="Explore"
-          subtitle={exploreView === 'graph' ? undefined : "Browse the graph and switch views without leaving the workspace."}
-          kicker={exploreView === 'graph' ? 'Graph Studio' : 'Vocabulary Browser'}
+          subtitle={exploreView === 'graph' ? undefined : exploreView === 'memories' ? "Browse and edit canonical AgentMemory documents." : "Browse the graph and switch views without leaving the workspace."}
+          kicker={exploreView === 'graph' ? 'Graph Studio' : exploreView === 'memories' ? 'Memory Browser' : 'Vocabulary Browser'}
          compact
          tabs={
-            <>
-              <button className="workspace-tab" data-active={exploreView === 'graph'} onClick={() => setExploreView('graph')}>
-                Semantica Explorer
-              </button>
-              <button className="workspace-tab" data-active={exploreView === 'vocabulary'} onClick={() => setExploreView('vocabulary')}>
-                Vocabulary Browser
-              </button>
-            </>
+            <ExploreWorkspaceTabs
+              activeView={exploreView}
+              agentMemoryAvailable={agentMemoryAvailable}
+              onSelect={switchExploreView}
+            />
          }
        >
          <ErrorBoundary key={`explore-${exploreView}`}>
@@ -1843,8 +1873,9 @@ export default function App() {
                <GraphWorkspace
                  externalFocusNodeId={graphFocusRequest?.nodeId}
                  externalFocusToken={graphFocusRequest?.token}
+                  onDirtyChange={setExploreDraftDirty}
                />
-              ) : <VocabularyWorkspace />}
+              ) : exploreView === 'memories' ? <MemoryWorkspace onDirtyChange={setExploreDraftDirty} /> : <VocabularyWorkspace />}
            </Suspense>
          </ErrorBoundary>
        </WorkspaceShell>
@@ -1989,13 +2020,13 @@ export default function App() {
      <style>{shellStyles}</style>
      <div className="app-shell">
        <aside className="app-rail">
-          <button className="brand-pill" title="Semantica Knowledge Explorer" onClick={() => setActiveWorkspace('welcome')} style={{ cursor: 'pointer', border: '1px solid rgba(127,208,255,0.18)' }}>SKE</button>
+          <button className="brand-pill" title="Semantica Knowledge Explorer" onClick={() => switchWorkspace('welcome')} style={{ cursor: 'pointer', border: '1px solid rgba(127,208,255,0.18)' }}>SKE</button>
          {navItems.map(({ id, label, hint, icon: Icon }) => (
            <button
              key={id}
              className="nav-button"
              data-active={activeWorkspace === id}
-              onClick={() => setActiveWorkspace(id)}
+              onClick={() => switchWorkspace(id)}
              title={hint}
            >
              <Icon size={20} />
@@ -0,0 +1,29 @@
+export type ExploreView = 'graph' | 'memories' | 'vocabulary';
+
+type ExploreWorkspaceTabsProps = {
+  activeView: ExploreView;
+  agentMemoryAvailable: boolean;
+  onSelect: (view: ExploreView) => void;
+};
+
+export function ExploreWorkspaceTabs({
+  activeView,
+  agentMemoryAvailable,
+  onSelect,
+}: ExploreWorkspaceTabsProps) {
+  return (
+    <>
+      <button className="workspace-tab" data-active={activeView === 'graph'} onClick={() => onSelect('graph')}>
+        Semantica Explorer
+      </button>
+      {agentMemoryAvailable ? (
+        <button className="workspace-tab" data-active={activeView === 'memories'} onClick={() => onSelect('memories')}>
+          Memories
+        </button>
+      ) : null}
+      <button className="workspace-tab" data-active={activeView === 'vocabulary'} onClick={() => onSelect('vocabulary')}>
+        Vocabulary Browser
+      </button>
+    </>
+  );
+}
@@ -0,0 +1,24 @@
+type Fetcher = (
+  input: RequestInfo | URL,
+  init?: RequestInit,
+) => Promise<Response>;
+
+type ExplorerInfo = {
+  capabilities?: {
+    agent_memory?: boolean;
+  };
+};
+
+export async function fetchAgentMemoryAvailability(
+  fetcher: Fetcher = fetch,
+): Promise<boolean> {
+  try {
+    const response = await fetcher('/api/info');
+    if (!response.ok) return false;
+
+    const info = await response.json() as ExplorerInfo;
+    return info.capabilities?.agent_memory === true;
+  } catch {
+    return false;
+  }
+}
@@ -15,6 +15,7 @@ export type RegistryEntryOp =
  | "export"
  | "merge"
  | "add-node"
+  | "update-node"
  | "add-edge"
  | "delete"
  | "infer"
@@ -16,6 +16,7 @@ const OP_META: Record<
  export:       { label: "EXPORT",       color: "#8fa8c6", bg: "rgba(143,168,198,0.08)", border: "rgba(143,168,198,0.18)" },
  merge:        { label: "MERGE",        color: "#f2b66d", bg: "rgba(242,182,109,0.12)", border: "rgba(242,182,109,0.28)" },
  "add-node":   { label: "ADD NODE",     color: "#4cc38a", bg: "rgba(76,195,138,0.12)",  border: "rgba(76,195,138,0.28)" },
+  "update-node": { label: "UPDATE NODE",  color: "#79c0ff", bg: "rgba(121,192,255,0.10)", border: "rgba(121,192,255,0.24)" },
  "add-edge":   { label: "ADD EDGE",     color: "#4cc38a", bg: "rgba(76,195,138,0.10)",  border: "rgba(76,195,138,0.22)" },
  delete:       { label: "DELETE",       color: "#ff7b72", bg: "rgba(255,123,114,0.12)", border: "rgba(255,123,114,0.28)" },
  infer:        { label: "INFER",        color: "#d2a8ff", bg: "rgba(210,168,255,0.12)", border: "rgba(210,168,255,0.28)" },
@@ -23,7 +24,7 @@ const OP_META: Record<
 };

 const ALL_OPS: (RegistryEntryOp | "all")[] = [
-  "all", "import", "export", "merge", "add-node", "add-edge", "infer", "delete", "vocab-import",
+  "all", "import", "export", "merge", "add-node", "update-node", "add-edge", "infer", "delete", "vocab-import",
 ];

 function formatTimestamp(date: Date): string {
@@ -4,6 +4,7 @@ import { graph } from "../../store/graphStore";
 import { GRAPH_THEME, withAlpha } from "./graphTheme";
 import type { GraphSelectedNodeKind } from "./types";
 import { MarkdownContentViewer } from "./MarkdownContentViewer";
+import type { MarkdownApplyResult } from "./markdownResourceClient";

 export type LinkPrediction = {
  target: string;
@@ -44,6 +45,8 @@ export interface GraphInspectorPanelProps {
  pathResult: PathResponse | null;
  onDownloadProvenance: (format: "json" | "markdown") => void;
  onFocusNode?: (nodeId: string) => void;
+  onMarkdownApplied?: (result: MarkdownApplyResult) => void;
+  onMarkdownDirtyChange?: (dirty: boolean) => void;
 }

 const PROVENANCE_KEYS = ["source", "source_url", "pmid", "pmids", "evidence", "provenance", "confidence"] as const;
@@ -304,6 +307,8 @@ export function GraphInspectorPanel({
  pathResult,
  onDownloadProvenance,
  onFocusNode,
+  onMarkdownApplied,
+  onMarkdownDirtyChange,
 }: GraphInspectorPanelProps) {
  if (!nodeId) {
    return (
@@ -414,19 +419,18 @@ export function GraphInspectorPanel({
        </div>
      ) : null}

-      {/* Content Section — only rendered when the node carries actual content.
-           This matches the existing inspector convention: sections that have no
-           data for the current node are either hidden (temporal bounds) or closed
-           by default (Source Attribution, Properties).  Always showing an open
-           empty panel would add noise for every relationship/predicate node. */}
-      {nodeContent && (
-        <details className="node-panel-collapse" open>
-          <summary className="node-panel-summary">Content</summary>
-          <div className="node-panel-body" style={{ marginTop: 8 }}>
-            <MarkdownContentViewer content={nodeContent} />
-          </div>
-        </details>
-      )}
+      {/* Canonical nodes remain editable even when their current body is empty. */}
+      <details className="node-panel-collapse" open>
+        <summary className="node-panel-summary">Content</summary>
+        <div className="node-panel-body" style={{ marginTop: 8 }}>
+          <MarkdownContentViewer
+            content={nodeContent}
+            resource={{ kind: "context-node", id: effectiveNodeId }}
+            onApplied={onMarkdownApplied}
+            onDirtyChange={onMarkdownDirtyChange}
+          />
+        </div>
+      </details>

      {/* Actions */}
      <section style={sectionStyle}>
@@ -44,6 +44,12 @@ import { createTemporalSnapshotGuards, type TemporalSnapshotResponse } from "./t
 import { SMALL_GRAPH_MAX_NODES } from "./smallGraphLayout";
 import { buildRealtimeEdgeAttributes } from "./realtimeGraphAttributes";
 import type { LinkPrediction, PathResponse } from "./GraphInspectorPanel";
+import type { MarkdownApplyResult } from "./markdownResourceClient";
+import {
+  NodeMarkdownRefreshGuard,
+  buildNodeMarkdownAttributeUpdate,
+  readNodeMarkdownAttributeUpdate,
+} from "./nodeMarkdownSync";
 import type { GraphSceneHandle, GraphSceneRuntime } from "./scene";
 import type {
  GraphAnalyticsSnapshot,
@@ -1241,9 +1247,10 @@ function collectPluginOverlays(
 interface GraphWorkspaceProps {
  externalFocusNodeId?: string;
  externalFocusToken?: number;
+  onDirtyChange?: (dirty: boolean) => void;
 }

-export function GraphWorkspace({ externalFocusNodeId, externalFocusToken }: GraphWorkspaceProps = {}) {
+export function GraphWorkspace({ externalFocusNodeId, externalFocusToken, onDirtyChange }: GraphWorkspaceProps = {}) {
  const [selectedNodeId, setSelectedNodeId] = useState("");
  const [focusedNodeId, setFocusedNodeId] = useState("");
  const [lastGroupedSelectedNodeId, setLastGroupedSelectedNodeId] = useState("");
@@ -1251,6 +1258,12 @@ export function GraphWorkspace({ externalFocusNodeId, externalFocusToken }: Grap
  const [isLayoutRunning, setIsLayoutRunning] = useState(false);
  const [graphReady, setGraphReady] = useState(false);
  const [graphVersion, setGraphVersion] = useState(0);
+  const [markdownDraftDirty, setMarkdownDraftDirty] = useState(false);
+  const markdownRefreshGuard = useMemo(() => new NodeMarkdownRefreshGuard(), []);
+  const handleMarkdownDirtyChange = useCallback((dirty: boolean) => {
+    setMarkdownDraftDirty(dirty);
+    onDirtyChange?.(dirty);
+  }, [onDirtyChange]);
  const [viewMode, setViewMode] = useState<GraphViewMode>("full");
  const [aggregationEnabled] = useState(true);
  const [collapsedNeighborhoodNodeIds, setCollapsedNeighborhoodNodeIds] = useState<string[]>([]);
@@ -1628,8 +1641,17 @@ export function GraphWorkspace({ externalFocusNodeId, externalFocusToken }: Grap
      : null,
    [viewMode, aggregationEnabled, collapsedNeighborhoodNodeIds, graphVersion],
  );
+  const confirmDiscardMarkdownDraft = useCallback(() => {
+    if (!markdownDraftDirty) return true;
+    const discard = window.confirm(
+      "Discard the unapplied Markdown draft and leave this node?",
+    );
+    return discard;
+  }, [markdownDraftDirty]);
+

  const requestViewMode = useCallback((nextViewMode: GraphViewMode) => {
+    if (nextViewMode !== viewMode && !confirmDiscardMarkdownDraft()) return;
    if (nextViewMode === "focused") {
      const resolution = resolveNodeIdForFocusedMode(selectedNodeId, pluginRuntimeRef.current?.displayGraph);
      if (!resolution.resolvedNodeId) {
@@ -1682,6 +1704,7 @@ export function GraphWorkspace({ externalFocusNodeId, externalFocusToken }: Grap
    ));
    setViewMode("full");
  }, [
+    confirmDiscardMarkdownDraft,
    aggregationEnabled,
    collapsedNeighborhoodNodeIds,
    focusedNodeId,
@@ -1692,9 +1715,11 @@ export function GraphWorkspace({ externalFocusNodeId, externalFocusToken }: Grap
    lastGroupedSelectedNodeId,
    resolveNodeIdForFocusedMode,
    selectedNodeId,
+    viewMode,
  ]);

  const focusNode = useCallback((nodeId: string) => {
+    if (nodeId !== selectedNodeId && !confirmDiscardMarkdownDraft()) return;
    if (!nodeId) {
      setSelectedNodeId("");
      setSelectedEdgeId("");
@@ -1720,14 +1745,18 @@ export function GraphWorkspace({ externalFocusNodeId, externalFocusToken }: Grap
      setFocusedNodeId(nextSelectedNodeId);
      setIsLayoutRunning(false);
    }
-  }, [viewMode]);  // Note: ego/heatmap/distanceMode effects re-run automatically when selectedNodeId changes
+  }, [confirmDiscardMarkdownDraft, selectedNodeId, viewMode]);  // Note: ego/heatmap/distanceMode effects re-run automatically when selectedNodeId changes

  useEffect(() => {
    if (!externalFocusNodeId || externalFocusToken == null) return;
    if (lastExternalFocusTokenRef.current === externalFocusToken) return;
    if (!graphReady || !graph.hasNode(externalFocusNodeId)) return;
-
+    if (
+      externalFocusNodeId !== selectedNodeId
+      && !confirmDiscardMarkdownDraft()
+    ) return;
    lastExternalFocusTokenRef.current = externalFocusToken;
+
    // Set state directly instead of going through focusNode(), which captures
    // a stale viewMode in its closure. setViewMode is called first so the node
    // is visible in the full graph before the scene pans to it.
@@ -1737,7 +1766,13 @@ export function GraphWorkspace({ externalFocusNodeId, externalFocusToken }: Grap
    window.setTimeout(() => {
      sceneRef.current?.focusNode(externalFocusNodeId);
    }, 0);
-  }, [externalFocusNodeId, externalFocusToken, graphReady]);
+  }, [
+    confirmDiscardMarkdownDraft,
+    externalFocusNodeId,
+    externalFocusToken,
+    graphReady,
+    selectedNodeId,
+  ]);

  const handleEdgeSelect = useCallback((edgeId: string) => {
    setSelectedEdgeId(edgeId);
@@ -1843,6 +1878,37 @@ export function GraphWorkspace({ externalFocusNodeId, externalFocusToken }: Grap
    document.body.removeChild(anchor);
  }, [inspectableNodeId]);

+  const handleMarkdownApplied = useCallback((result: MarkdownApplyResult) => {
+    if (result.resource.kind !== "context-node") return;
+    if (!graph.hasNode(result.resource.id)) return;
+    const syncGeneration = markdownRefreshGuard.begin(result.resource.id);
+    const attributes = graph.getNodeAttributes(result.resource.id) as NodeAttributes;
+    graph.mergeNodeAttributes(
+      result.resource.id,
+      buildNodeMarkdownAttributeUpdate(
+        result.resource.id,
+        result.body,
+        attributes.properties ?? {},
+      ),
+    );
+    setGraphVersion((current) => current + 1);
+    sceneRef.current?.getRuntime()?.requestRender();
+
+    void readNodeMarkdownAttributeUpdate(result.resource.id)
+      .then((savedAttributes) => {
+        if (
+          !markdownRefreshGuard.isCurrent(result.resource.id, syncGeneration)
+          || !graph.hasNode(result.resource.id)
+        ) return;
+        graph.mergeNodeAttributes(result.resource.id, savedAttributes);
+        setGraphVersion((current) => current + 1);
+        sceneRef.current?.getRuntime()?.requestRender();
+      })
+      .catch((syncError) => {
+        console.error("[GraphWorkspace] applied node refresh failed", syncError);
+      });
+  }, [markdownRefreshGuard]);
+
  useEffect(() => {
    const protocol = window.location.protocol === "https:" ? "wss:" : "ws:";
    const socket = new WebSocket(`${protocol}//${window.location.host}/ws/graph-updates`);
@@ -1873,6 +1939,25 @@ export function GraphWorkspace({ externalFocusNodeId, externalFocusToken }: Grap
          setGraphVersion((current) => current + 1);
          sceneRef.current?.getRuntime()?.requestRender();
        }
+        if (eventType === "UPDATE_NODE" && payload?.id && graph.hasNode(payload.id)) {
+          markdownRefreshGuard.invalidate(payload.id);
+          const properties = payload.properties ?? {};
+          const current = graph.getNodeAttributes(payload.id) as NodeAttributes;
+          const content = typeof properties.content === "string" ? properties.content : "";
+          graph.mergeNodeAttributes(payload.id, {
+            ...buildNodeMarkdownAttributeUpdate(payload.id, content, properties),
+            nodeType: payload.type ?? current.nodeType,
+            valid_from: properties.valid_from ?? null,
+            valid_until: properties.valid_until ?? null,
+          });
+          logEvent(
+            "update-node",
+            `Updated node ${payload.id} via realtime ws`,
+            { nodeId: payload.id, nodeType: payload.type },
+          );
+          setGraphVersion((version) => version + 1);
+          sceneRef.current?.getRuntime()?.requestRender();
+        }
        if (eventType === "ADD_EDGE") {
          const isSmallGraph = smallGraphModeRef.current;
          batchMergeEdges([
@@ -1899,7 +1984,7 @@ export function GraphWorkspace({ externalFocusNodeId, externalFocusToken }: Grap
    return () => {
      socket.close();
    };
-  }, []);
+  }, [markdownRefreshGuard]);

  useEffect(() => {
    setCollapsedNeighborhoodNodeIds([]);
@@ -2172,28 +2257,29 @@ export function GraphWorkspace({ externalFocusNodeId, externalFocusToken }: Grap
  }, [collapsedNeighborhoodNodeIds, focusedNodeId, selectedNodeId, viewMode]);
  const structuralActivePath = structuralSelectedNodeId ? activePath : EMPTY_PATH;
  const structuralActivePathEdgeIds = structuralSelectedNodeId ? activePathEdgeIds : EMPTY_PATH;
-  const displayResult = useMemo(
-    () => (
-      viewMode === "grouped"
-        ? (groupedDisplayCandidate ?? resolveDisplayGraph("", EMPTY_PATH, EMPTY_PATH, "grouped", {
-            aggregationEnabled,
-            collapsedNeighborhoodNodeIds,
-          }))
-        : resolveDisplayGraph(structuralSelectedNodeId, structuralActivePath, structuralActivePathEdgeIds, viewMode, {
-            aggregationEnabled,
-            collapsedNeighborhoodNodeIds,
-          })
-    ),
-    [
-      aggregationEnabled,
-      collapsedNeighborhoodNodeIds,
-      groupedDisplayCandidate,
-      structuralActivePath,
-      structuralActivePathEdgeIds,
-      structuralSelectedNodeId,
-      viewMode,
-    ],
-  );
+  const displayResult = useMemo(() => {
+    // The displayed graph is an aggregated clone. Rebuild it after domain
+    // mutations so applied Markdown labels do not remain stale on the canvas.
+    void graphVersion;
+    return viewMode === "grouped"
+      ? (groupedDisplayCandidate ?? resolveDisplayGraph("", EMPTY_PATH, EMPTY_PATH, "grouped", {
+          aggregationEnabled,
+          collapsedNeighborhoodNodeIds,
+        }))
+      : resolveDisplayGraph(structuralSelectedNodeId, structuralActivePath, structuralActivePathEdgeIds, viewMode, {
+          aggregationEnabled,
+          collapsedNeighborhoodNodeIds,
+        });
+  }, [
+    aggregationEnabled,
+    collapsedNeighborhoodNodeIds,
+    graphVersion,
+    groupedDisplayCandidate,
+    structuralActivePath,
+    structuralActivePathEdgeIds,
+    structuralSelectedNodeId,
+    viewMode,
+  ]);
  const displayState = useMemo(
    () => (
      viewMode === "grouped"
@@ -3295,6 +3381,8 @@ export function GraphWorkspace({ externalFocusNodeId, externalFocusToken }: Grap
                      pathResult={pathResult}
                      onDownloadProvenance={(format) => void handleDownloadProvenance(format)}
                      onFocusNode={focusNode}
+                      onMarkdownApplied={handleMarkdownApplied}
+                      onMarkdownDirtyChange={handleMarkdownDirtyChange}
                    />
                  </Suspense>
                </div>
@@ -1,125 +1,243 @@
-import { useState, useRef, useEffect, useMemo, type CSSProperties } from "react";
+import {
+  useEffect,
+  useMemo,
+  useRef,
+  useState,
+  type CSSProperties,
+} from "react";
 import ReactMarkdown, { type Components } from "react-markdown";
 import remarkGfm from "remark-gfm";
-import { Check, Copy, Code2, Eye, ExternalLink, Image as ImageIcon } from "lucide-react";
+import {
+  Check,
+  Code2,
+  Copy,
+  Eye,
+  ExternalLink,
+  Image as ImageIcon,
+  Loader2,
+  Pencil,
+  RefreshCw,
+  X,
+} from "lucide-react";
 import { GRAPH_THEME } from "./graphTheme";
+import type { MarkdownApplyResult } from "./markdownResourceClient";
+import type { MarkdownResourceRef } from "./markdownEditorState";
 import { isSafeUrl } from "./markdownUrlSafety";
+import { useMarkdownEditor } from "./useMarkdownEditor";

 export interface MarkdownContentViewerProps {
  content?: string | null;
+  resource?: MarkdownResourceRef;
+  onApplied?: (result: MarkdownApplyResult) => void;
+  onDirtyChange?: (dirty: boolean) => void;
  className?: string;
  defaultMode?: "preview" | "source";
 }

 export function MarkdownContentViewer({
  content,
+  resource,
+  onApplied,
+  onDirtyChange,
  className,
  defaultMode = "preview",
 }: MarkdownContentViewerProps) {
  const [activeMode, setActiveMode] = useState<"preview" | "source">(defaultMode);
  const [copied, setCopied] = useState(false);
+  const modeBeforeEditRef = useRef<"preview" | "source">(defaultMode);
+  const resourceKey = resource ? `${resource.kind}:${resource.id}` : "";
+  const [activeResourceKey, setActiveResourceKey] = useState(resourceKey);
+  const editor = useMarkdownEditor({ resource, onApplied, onDirtyChange });
+  const {
+    session,
+    error,
+    dirty,
+    editing,
+    saving,
+    loading,
+  } = editor;
+
+  if (activeResourceKey !== resourceKey) {
+    setActiveResourceKey(resourceKey);
+    setCopied(false);
+    setActiveMode(defaultMode);
+  }
+
  // Track the content value for which the copied indicator is valid.
-  // When content changes (i.e. the user selects a different node), reset the
-  // copied indicator inline during render rather than in a useEffect — this
-  // avoids a cascading-render lint error and is the React-recommended pattern
-  // for resetting derived visual state on prop changes.
  const [copiedForContent, setCopiedForContent] = useState<string | null | undefined>(content);
  if (copiedForContent !== content) {
    setCopiedForContent(content);
-    if (copied) {
-      // Clear the stale indicator synchronously so the new node's copy button
-      // never shows "Copied" from the previous selection.
-      setCopied(false);
-    }
+    if (copied) setCopied(false);
  }

-  const copyTimeoutRef = useRef<ReturnType<typeof setTimeout> | null>(null);
-
-  // Clean up any outstanding timeout on unmount.
+  const copyTimeoutRef = useRef<number | undefined>(undefined);
  useEffect(() => {
    return () => {
-      if (copyTimeoutRef.current) {
-        clearTimeout(copyTimeoutRef.current);
-      }
+      clearTimeout(copyTimeoutRef.current);
    };
  }, []);

-  const rawContent = typeof content === "string" ? content : "";
+  const rawContent = editor.editing
+    ? editor.session?.draft ?? ""
+    : (typeof content === "string" ? content : "");
+  const previewContent = useMemo(() => {
+    if (!editor.editing) return rawContent;
+    const lines = rawContent.split(/\r?\n/);
+    if (lines[0] !== "---") return rawContent;
+    const closingIndex = lines.findIndex((line, index) => index > 0 && line === "---");
+    return closingIndex < 0 ? rawContent : lines.slice(closingIndex + 1).join("\n").replace(/^\n/, "");
+  }, [editor.editing, rawContent]);
  const hasContent = rawContent.trim().length > 0;
-
-  // react-markdown runs the whole remark pipeline synchronously inside its own
-  // render, so without this memo every unrelated re-render of this component --
-  // clicking Copy, toggling Preview/Source -- re-parses the entire document.
-  // Measured at ~364ms per re-render for a 1000-row GFM table (issue #1118).
-  // Keyed on rawContent so a genuine node change still re-parses exactly once.
  const renderedMarkdown = useMemo(
    () => (
      <ReactMarkdown remarkPlugins={REMARK_PLUGINS} components={MARKDOWN_COMPONENTS}>
-        {rawContent}
+        {previewContent}
      </ReactMarkdown>
    ),
-    [rawContent],
+    [previewContent],
  );
-
  const handleCopy = async () => {
    if (!hasContent) return;
    try {
      await navigator.clipboard.writeText(rawContent);
-      if (copyTimeoutRef.current) {
-        clearTimeout(copyTimeoutRef.current);
-      }
+      clearTimeout(copyTimeoutRef.current);
      setCopied(true);
-      copyTimeoutRef.current = setTimeout(() => setCopied(false), 1500);
+      copyTimeoutRef.current = window.setTimeout(() => setCopied(false), 1500);
    } catch {
-      // Clipboard write unavailable
+      // Clipboard write unavailable.
+    }
+  };
+
+  const handleEdit = async () => {
+    modeBeforeEditRef.current = activeMode;
+    setActiveMode("source");
+    if (!await editor.beginEdit()) {
+      setActiveMode(modeBeforeEditRef.current);
+    }
+  };
+
+  const handleCancel = () => {
+    editor.discard();
+    setActiveMode(modeBeforeEditRef.current);
+  };
+
+  const handleApply = async () => {
+    if (await editor.save()) {
+      setActiveMode("preview");
    }
  };

  return (
    <div className={className} style={viewerContainerStyle}>
      <div style={viewerHeaderStyle}>
-        <div style={{ display: "flex", gap: 4 }} role="tablist">
+        <div style={{ display: "flex", gap: 4 }} role="tablist" aria-label="Markdown view">
          <button
            type="button"
            role="tab"
            aria-selected={activeMode === "preview"}
+            aria-controls="markdown-viewer-panel"
            onClick={() => setActiveMode("preview")}
            style={{ ...tabBtnStyle, ...(activeMode === "preview" ? activeTabBtnStyle : {}) }}
          >
-            <Eye size={12} style={{ marginRight: 5 }} />
+            <Eye size={12} style={{ marginRight: 5 }} aria-hidden="true" />
            Preview
          </button>
          <button
            type="button"
            role="tab"
            aria-selected={activeMode === "source"}
+            aria-controls="markdown-viewer-panel"
            onClick={() => setActiveMode("source")}
            style={{ ...tabBtnStyle, ...(activeMode === "source" ? activeTabBtnStyle : {}) }}
          >
-            <Code2 size={12} style={{ marginRight: 5 }} />
+            <Code2 size={12} style={{ marginRight: 5 }} aria-hidden="true" />
            Source
          </button>
        </div>

-        {hasContent && (
-          <button type="button" onClick={() => void handleCopy()} style={copyBtnStyle} title="Copy raw content">
-            {copied ? (
-              <>
-                <Check size={12} color="#3fb950" style={{ marginRight: 4 }} />
-                <span style={{ color: "#3fb950", fontSize: 11 }}>Copied</span>
-              </>
-            ) : (
-              <>
-                <Copy size={12} style={{ marginRight: 4 }} />
-                <span style={{ fontSize: 11 }}>Copy</span>
-              </>
-            )}
-          </button>
-        )}
+        <div style={{ display: "flex", alignItems: "center", gap: 6 }}>
+          {hasContent && (
+            <button type="button" onClick={() => void handleCopy()} style={copyBtnStyle} title="Copy raw content">
+              {copied ? (
+                <>
+                  <Check size={12} color="#3fb950" style={{ marginRight: 4 }} aria-hidden="true" />
+                  <span style={{ color: "#3fb950", fontSize: 11 }}>Copied</span>
+                </>
+              ) : (
+                <>
+                  <Copy size={12} style={{ marginRight: 4 }} aria-hidden="true" />
+                  <span style={{ fontSize: 11 }}>Copy</span>
+                </>
+              )}
+            </button>
+          )}
+          {resource && !editing && !loading ? (
+            <button type="button" onClick={() => void handleEdit()} style={copyBtnStyle}>
+              <Pencil size={12} style={{ marginRight: 4 }} aria-hidden="true" />
+              Edit
+            </button>
+          ) : null}
+          {loading ? (
+            <button type="button" disabled style={{ ...copyBtnStyle, opacity: 0.65 }}>
+              <Loader2 size={12} className="animate-spin" style={{ marginRight: 4 }} aria-hidden="true" />
+              Loading…
+            </button>
+          ) : null}
+          {editing ? (
+            <>
+              <button type="button" onClick={handleCancel} disabled={saving} style={copyBtnStyle}>
+                <X size={12} style={{ marginRight: 4 }} aria-hidden="true" />
+                Cancel
+              </button>
+              <button
+                type="button"
+                onClick={() => void handleApply()}
+                disabled={saving || !dirty}
+                title={!dirty ? "Make a change before applying" : undefined}
+                style={{ ...saveBtnStyle, opacity: saving || !dirty ? 0.55 : 1 }}
+              >
+                {saving ? (
+                  <Loader2 size={12} className="animate-spin" style={{ marginRight: 4 }} aria-hidden="true" />
+                ) : (
+                  <Check size={12} style={{ marginRight: 4 }} aria-hidden="true" />
+                )}
+                {saving ? "Applying…" : "Apply"}
+              </button>
+            </>
+          ) : null}
+        </div>
      </div>

-      <div style={viewerBodyStyle}>
-        {!hasContent ? (
+      {error ? (
+        <div id="markdown-editor-error" role="alert" style={errorStyle}>
+          <span>{error.message}</span>
+          {error.kind === "conflict" ? (
+            <button type="button" onClick={() => void editor.reloadLatest()} style={errorActionStyle}>
+              <RefreshCw size={12} style={{ marginRight: 4 }} aria-hidden="true" />
+              Reload latest
+            </button>
+          ) : null}
+        </div>
+      ) : null}
+
+      <div
+        id="markdown-viewer-panel"
+        role="tabpanel"
+        aria-busy={saving || loading}
+        style={viewerBodyStyle}
+      >
+        {activeMode === "source" && editing ? (
+          <textarea
+            aria-label="Markdown source"
+            aria-describedby={error ? "markdown-editor-error" : undefined}
+            aria-invalid={error?.kind === "validation" || undefined}
+            value={session?.draft ?? ""}
+            onChange={(event) => editor.changeDraft(event.target.value)}
+            disabled={saving}
+            spellCheck={false}
+            style={editorStyle}
+          />
+        ) : !hasContent ? (
          <div style={emptyTextStyle}>No content available for this node.</div>
        ) : activeMode === "source" ? (
          <pre style={sourcePreStyle}>
@@ -238,6 +356,8 @@ const viewerHeaderStyle: CSSProperties = {
  display: "flex",
  alignItems: "center",
  justifyContent: "space-between",
+  gap: 8,
+  flexWrap: "wrap",
  padding: "6px 10px",
  background: "rgba(0, 0, 0, 0.2)",
  borderBottom: `1px solid ${GRAPH_THEME.ui.surface.panelBorder}`,
@@ -275,12 +395,56 @@ const copyBtnStyle: CSSProperties = {
  cursor: "pointer",
 };

+const saveBtnStyle: CSSProperties = {
+  ...copyBtnStyle,
+  background: GRAPH_THEME.ui.control.primaryBg,
+  border: `1px solid ${GRAPH_THEME.ui.control.primaryBorder}`,
+  color: GRAPH_THEME.ui.control.primaryText,
+  fontWeight: 700,
+};
+
+const errorStyle: CSSProperties = {
+  display: "flex",
+  alignItems: "center",
+  justifyContent: "space-between",
+  gap: 8,
+  padding: "8px 12px",
+  color: "#ffb4ad",
+  background: "rgba(248, 81, 73, 0.1)",
+  borderBottom: "1px solid rgba(248, 81, 73, 0.25)",
+  fontSize: 12,
+  lineHeight: 1.5,
+};
+
+const errorActionStyle: CSSProperties = {
+  ...copyBtnStyle,
+  flexShrink: 0,
+  color: "#ffb4ad",
+  border: "1px solid rgba(248, 81, 73, 0.32)",
+};
+
 const viewerBodyStyle: CSSProperties = {
  padding: 12,
  maxHeight: 380,
  overflowY: "auto",
 };

+const editorStyle: CSSProperties = {
+  display: "block",
+  boxSizing: "border-box",
+  width: "100%",
+  minHeight: 280,
+  resize: "vertical",
+  padding: 10,
+  borderRadius: 8,
+  border: `1px solid ${GRAPH_THEME.ui.control.activeBorder}`,
+  background: "rgba(0, 0, 0, 0.3)",
+  color: GRAPH_THEME.ui.text.strong,
+  fontFamily: "'JetBrains Mono', 'Fira Code', monospace",
+  fontSize: 12,
+  lineHeight: 1.6,
+};
+
 const emptyTextStyle: CSSProperties = {
  color: GRAPH_THEME.ui.text.muted,
  fontSize: 12,
@@ -0,0 +1,114 @@
+export type MarkdownResourceRef =
+  | { kind: "context-node"; id: string }
+  | { kind: "agent-memory"; id: string };
+
+export type EditorStatus =
+  | "viewing"
+  | "loading-document"
+  | "editing"
+  | "saving"
+  | "validation-error"
+  | "save-error"
+  | "conflict";
+
+export interface MarkdownEditorError {
+  kind: "validation" | "conflict" | "save" | "network";
+  message: string;
+  field?: string;
+  currentRevision?: string;
+}
+
+export interface MarkdownEditSession {
+  resource: MarkdownResourceRef;
+  baseSource: string;
+  baseRevision: string;
+  draft: string;
+  status: EditorStatus;
+  error: MarkdownEditorError | null;
+}
+
+export interface MarkdownSavedDocument {
+  source: string;
+  revision: string;
+}
+
+export function createLoadingSession(resource: MarkdownResourceRef): MarkdownEditSession {
+  return {
+    resource,
+    baseSource: "",
+    baseRevision: "",
+    draft: "",
+    status: "loading-document",
+    error: null,
+  };
+}
+
+export function createEditSession(
+  resource: MarkdownResourceRef,
+  document: MarkdownSavedDocument,
+): MarkdownEditSession {
+  return {
+    resource,
+    baseSource: document.source,
+    baseRevision: document.revision,
+    draft: document.source,
+    status: "editing",
+    error: null,
+  };
+}
+
+export function updateDraft(
+  session: MarkdownEditSession,
+  draft: string,
+): MarkdownEditSession {
+  return {
+    ...session,
+    draft,
+    status: "editing",
+    error: null,
+  };
+}
+
+export function isDirty(session: MarkdownEditSession | null): boolean {
+  return session !== null && session.draft !== session.baseSource;
+}
+
+export function saveStarted(session: MarkdownEditSession): MarkdownEditSession {
+  if (!isDirty(session)) return session;
+  return { ...session, status: "saving", error: null };
+}
+
+export function saveSucceeded(
+  session: MarkdownEditSession,
+  document: MarkdownSavedDocument,
+): MarkdownEditSession {
+  return {
+    ...session,
+    baseSource: document.source,
+    baseRevision: document.revision,
+    draft: document.source,
+    status: "viewing",
+    error: null,
+  };
+}
+
+export function saveFailed(
+  session: MarkdownEditSession,
+  error: MarkdownEditorError,
+): MarkdownEditSession {
+  const status: EditorStatus =
+    error.kind === "validation"
+      ? "validation-error"
+      : error.kind === "conflict"
+        ? "conflict"
+        : "save-error";
+  return { ...session, status, error };
+}
+
+export function cancelEdit(): null {
+  return null;
+}
+
+export function shouldConfirmDiscard(session: MarkdownEditSession | null): boolean {
+  return isDirty(session) && session?.status !== "saving";
+}
@@ -0,0 +1,103 @@
+import type {
+  MarkdownEditorError,
+  MarkdownResourceRef,
+} from "./markdownEditorState";
+
+export interface MarkdownDocument {
+  resource: MarkdownResourceRef;
+  source: string;
+  body: string;
+  revision: string;
+  editable: boolean;
+}
+
+export interface MarkdownApplyResult extends MarkdownDocument {
+  changed: boolean;
+}
+
+type ErrorDetail = {
+  code?: string;
+  message?: string;
+  field?: string;
+  current_revision?: string;
+};
+
+export class MarkdownClientError extends Error implements MarkdownEditorError {
+  readonly kind: MarkdownEditorError["kind"];
+  readonly field?: string;
+  readonly currentRevision?: string;
+
+  constructor(error: MarkdownEditorError) {
+    super(error.message);
+    this.name = "MarkdownClientError";
+    this.kind = error.kind;
+    this.field = error.field;
+    this.currentRevision = error.currentRevision;
+  }
+}
+
+function resourceUrl(ref: MarkdownResourceRef): string {
+  return `/api/markdown/${ref.kind}/${encodeURIComponent(ref.id)}`;
+}
+
+async function responseError(response: Response): Promise<MarkdownClientError> {
+  let detail: ErrorDetail = {};
+  try {
+    const payload = (await response.json()) as { detail?: ErrorDetail };
+    if (payload.detail && typeof payload.detail === "object") {
+      detail = payload.detail;
+    }
+  } catch {
+    // A non-JSON response is mapped from its status below.
+  }
+
+  const kind: MarkdownEditorError["kind"] =
+    response.status === 422
+      ? "validation"
+      : response.status === 409
+        ? "conflict"
+        : "save";
+  return new MarkdownClientError({
+    kind,
+    message: detail.message || `Markdown request failed (${response.status}).`,
+    field: detail.field,
+    currentRevision: detail.current_revision,
+  });
+}
+
+async function request<T>(input: RequestInfo | URL, init?: RequestInit): Promise<T> {
+  try {
+    const response = await fetch(input, init);
+    if (!response.ok) {
+      throw await responseError(response);
+    }
+    return (await response.json()) as T;
+  } catch (error) {
+    if (error instanceof MarkdownClientError) throw error;
+    throw new MarkdownClientError({
+      kind: "network",
+      message: "The Markdown service could not be reached. Your draft was kept.",
+    });
+  }
+}
+
+export function readMarkdownResource(
+  ref: MarkdownResourceRef,
+): Promise<MarkdownDocument> {
+  return request<MarkdownDocument>(resourceUrl(ref));
+}
+
+export function applyMarkdownResource(
+  ref: MarkdownResourceRef,
+  markdown: string,
+  expectedRevision: string,
+): Promise<MarkdownApplyResult> {
+  return request<MarkdownApplyResult>(resourceUrl(ref), {
+    method: "PUT",
+    headers: { "Content-Type": "application/json" },
+    body: JSON.stringify({
+      markdown,
+      expected_revision: expectedRevision,
+    }),
+  });
+}
@@ -0,0 +1,81 @@
+export interface NodeMarkdownAttributeUpdate {
+  label: string;
+  content: string;
+  properties: Record<string, unknown>;
+}
+
+export interface GraphNodeMarkdownSnapshot {
+  id: string;
+  type: string;
+  content: string;
+  properties: Record<string, unknown>;
+  valid_from: string | null;
+  valid_until: string | null;
+}
+
+export interface SavedNodeMarkdownAttributeUpdate extends NodeMarkdownAttributeUpdate {
+  nodeType: string;
+  valid_from: string | null;
+  valid_until: string | null;
+}
+
+export class NodeMarkdownRefreshGuard {
+  private readonly generations = new Map<string, number>();
+
+  begin(nodeId: string): number {
+    const generation = (this.generations.get(nodeId) ?? 0) + 1;
+    this.generations.set(nodeId, generation);
+    return generation;
+  }
+
+  invalidate(nodeId: string): void {
+    this.begin(nodeId);
+  }
+
+  isCurrent(nodeId: string, generation: number): boolean {
+    return this.generations.get(nodeId) === generation;
+  }
+}
+
+export function buildNodeMarkdownAttributeUpdate(
+  nodeId: string,
+  content: string,
+  properties: Record<string, unknown>,
+): NodeMarkdownAttributeUpdate {
+  return {
+    label: content || nodeId,
+    content,
+    properties: {
+      ...properties,
+      content,
+    },
+  };
+}
+
+export async function readNodeMarkdownAttributeUpdate(
+  nodeId: string,
+  fetcher: typeof fetch = fetch,
+): Promise<SavedNodeMarkdownAttributeUpdate> {
+  const response = await fetcher(
+    `/api/graph/node?node_id=${encodeURIComponent(nodeId)}`,
+  );
+  if (!response.ok) {
+    throw new Error(`Graph node refresh failed (${response.status}).`);
+  }
+
+  const node = await response.json() as GraphNodeMarkdownSnapshot;
+  if (node.id !== nodeId) {
+    throw new Error("Graph node refresh returned a different resource.");
+  }
+
+  return {
+    ...buildNodeMarkdownAttributeUpdate(
+      node.id,
+      node.content,
+      node.properties ?? {},
+    ),
+    nodeType: node.type,
+    valid_from: node.valid_from ?? null,
+    valid_until: node.valid_until ?? null,
+  };
+}
@@ -0,0 +1,177 @@
+import {
+  useCallback,
+  useEffect,
+  useLayoutEffect,
+  useRef,
+  useState,
+} from "react";
+import {
+  MarkdownClientError,
+  applyMarkdownResource,
+  readMarkdownResource,
+  type MarkdownApplyResult,
+} from "./markdownResourceClient";
+import {
+  cancelEdit,
+  createEditSession,
+  createLoadingSession,
+  isDirty,
+  saveFailed,
+  saveStarted,
+  saveSucceeded,
+  updateDraft,
+  type MarkdownEditorError,
+  type MarkdownEditSession,
+  type MarkdownResourceRef,
+} from "./markdownEditorState";
+
+interface MarkdownEditorOptions {
+  resource?: MarkdownResourceRef;
+  onApplied?: (result: MarkdownApplyResult) => void;
+  onDirtyChange?: (dirty: boolean) => void;
+}
+
+interface KeyedError {
+  resourceKey: string;
+  error: MarkdownEditorError;
+}
+
+function keyOf(resource?: MarkdownResourceRef): string {
+  return resource ? `${resource.kind}:${resource.id}` : "";
+}
+
+function normalizeError(failure: unknown): MarkdownEditorError {
+  if (failure instanceof MarkdownClientError) return failure;
+  return {
+    kind: "network",
+    message: "The Markdown service could not be reached. Your draft was kept.",
+  };
+}
+
+export function useMarkdownEditor({
+  resource,
+  onApplied,
+  onDirtyChange,
+}: MarkdownEditorOptions) {
+  const resourceKey = keyOf(resource);
+  const [session, setSession] = useState<MarkdownEditSession | null>(null);
+  const [viewError, setViewError] = useState<KeyedError | null>(null);
+  const [renderedResourceKey, setRenderedResourceKey] = useState(resourceKey);
+  const loadGenerationRef = useRef(0);
+
+  if (renderedResourceKey !== resourceKey) {
+    setRenderedResourceKey(resourceKey);
+    setSession(null);
+    setViewError(null);
+  }
+
+  useLayoutEffect(() => {
+    loadGenerationRef.current += 1;
+  }, [resourceKey]);
+
+  const activeSession = session && keyOf(session.resource) === resourceKey
+    ? session
+    : null;
+  const dirty = isDirty(activeSession);
+  const editing = activeSession !== null
+    && activeSession.status !== "viewing"
+    && activeSession.status !== "loading-document";
+  const saving = activeSession?.status === "saving";
+  const loading = activeSession?.status === "loading-document";
+  const error = activeSession?.error
+    ?? (viewError?.resourceKey === resourceKey ? viewError.error : null);
+
+  useEffect(() => {
+    onDirtyChange?.(dirty);
+    return () => {
+      if (dirty) onDirtyChange?.(false);
+    };
+  }, [dirty, onDirtyChange]);
+
+  useEffect(() => {
+    if (!dirty) return;
+    const protectDraft = (event: BeforeUnloadEvent) => {
+      event.preventDefault();
+      event.returnValue = "";
+    };
+    window.addEventListener("beforeunload", protectDraft);
+    return () => window.removeEventListener("beforeunload", protectDraft);
+  }, [dirty]);
+
+  const beginEdit = useCallback(async () => {
+    if (!resource) return false;
+    const loadGeneration = ++loadGenerationRef.current;
+    setViewError(null);
+    setSession(createLoadingSession(resource));
+    try {
+      const document = await readMarkdownResource(resource);
+      if (loadGeneration !== loadGenerationRef.current) return false;
+      setSession(createEditSession(resource, document));
+      return true;
+    } catch (failure) {
+      if (loadGeneration !== loadGenerationRef.current) return false;
+      setSession(null);
+      setViewError({ resourceKey, error: normalizeError(failure) });
+      return false;
+    }
+  }, [resource, resourceKey]);
+
+  const discard = useCallback(() => {
+    if (!activeSession || saving) return;
+    setSession(cancelEdit());
+    setViewError(null);
+  }, [activeSession, saving]);
+
+  const save = useCallback(async () => {
+    if (!activeSession || saving || !dirty) return false;
+    const resourceGeneration = loadGenerationRef.current;
+    const pending = saveStarted(activeSession);
+    setSession(pending);
+    try {
+      const result = await applyMarkdownResource(
+        pending.resource,
+        pending.draft,
+        pending.baseRevision,
+      );
+      if (resourceGeneration !== loadGenerationRef.current) return false;
+      setSession(saveSucceeded(pending, result));
+      onApplied?.(result);
+      return true;
+    } catch (failure) {
+      if (resourceGeneration !== loadGenerationRef.current) return false;
+      setSession(saveFailed(pending, normalizeError(failure)));
+      return false;
+    }
+  }, [activeSession, dirty, onApplied, saving]);
+
+  const reloadLatest = useCallback(async () => {
+    if (!resource || saving) return;
+    if (
+      dirty
+      && !window.confirm("Discard this draft and reload the latest applied version?")
+    ) return;
+    await beginEdit();
+  }, [beginEdit, dirty, resource, saving]);
+
+  const changeDraft = useCallback((draft: string) => {
+    setSession((current) => (
+      current && keyOf(current.resource) === resourceKey
+        ? updateDraft(current, draft)
+        : current
+    ));
+  }, [resourceKey]);
+
+  return {
+    session: activeSession,
+    error,
+    dirty,
+    editing,
+    saving,
+    loading,
+    beginEdit,
+    discard,
+    save,
+    reloadLatest,
+    changeDraft,
+  };
+}
@@ -0,0 +1,464 @@
+import { useCallback, useEffect, useRef, useState, type CSSProperties } from "react";
+import { Brain, RefreshCw } from "lucide-react";
+
+import { MarkdownContentViewer } from "./GraphWorkspace/MarkdownContentViewer";
+import {
+  readMarkdownResource,
+  type MarkdownApplyResult,
+} from "./GraphWorkspace/markdownResourceClient";
+import { GRAPH_THEME } from "./GraphWorkspace/graphTheme";
+
+interface MemorySummary {
+  id: string;
+  type: string;
+  excerpt: string;
+  updated_at: string | null;
+}
+
+interface MemoryListResponse {
+  items: MemorySummary[];
+  total: number;
+  skip: number;
+  limit: number;
+}
+interface MemoryWorkspaceProps {
+  onDirtyChange?: (dirty: boolean) => void;
+}
+
+const MEMORY_PAGE_SIZE = 100;
+
+
+function responseMessage(payload: unknown, fallback: string): string {
+  if (!payload || typeof payload !== "object" || !("detail" in payload)) {
+    return fallback;
+  }
+  return typeof payload.detail === "string" ? payload.detail : fallback;
+}
+
+
+async function fetchMemoryList(skip = 0): Promise<MemoryListResponse> {
+  const response = await fetch(`/api/memories?skip=${skip}&limit=${MEMORY_PAGE_SIZE}`);
+  if (!response.ok) {
+    const payload = await response.json().catch(() => null);
+    throw new Error(responseMessage(payload, `Memory list failed (${response.status}).`));
+  }
+  return response.json() as Promise<MemoryListResponse>;
+}
+
+async function fetchLoadedMemoryPages(endOffset: number): Promise<{
+  items: MemorySummary[];
+  total: number;
+  nextOffset: number;
+}> {
+  const items: MemorySummary[] = [];
+  let total = 0;
+  let nextOffset = 0;
+  const targetOffset = Math.max(endOffset, MEMORY_PAGE_SIZE);
+
+  while (nextOffset < targetOffset) {
+    const payload = await fetchMemoryList(nextOffset);
+    items.push(...payload.items);
+    total = payload.total;
+    nextOffset = payload.skip + payload.items.length;
+    if (payload.items.length === 0 || nextOffset >= total) break;
+  }
+
+  return { items, total, nextOffset };
+}
+
+export function MemoryWorkspace({ onDirtyChange }: MemoryWorkspaceProps = {}) {
+  const [items, setItems] = useState<MemorySummary[]>([]);
+  const [total, setTotal] = useState(0);
+  const [nextOffset, setNextOffset] = useState(0);
+  const [selectedId, setSelectedId] = useState("");
+  const [selectedBody, setSelectedBody] = useState("");
+  const [loading, setLoading] = useState(true);
+  const [loadingMore, setLoadingMore] = useState(false);
+  const [error, setError] = useState("");
+  const [dirty, setDirty] = useState(false);
+  const [reloadToken, setReloadToken] = useState(0);
+  const listGenerationRef = useRef(0);
+  const selectionGenerationRef = useRef(0);
+  const handleDirtyChange = useCallback((nextDirty: boolean) => {
+    setDirty(nextDirty);
+    onDirtyChange?.(nextDirty);
+  }, [onDirtyChange]);
+
+
+  useEffect(() => {
+    let cancelled = false;
+    const listGeneration = ++listGenerationRef.current;
+    const selectionGeneration = ++selectionGenerationRef.current;
+    const isCurrent = () => (
+      !cancelled
+      && listGeneration === listGenerationRef.current
+      && selectionGeneration === selectionGenerationRef.current
+    );
+    const load = async () => {
+      setLoading(true);
+      setLoadingMore(false);
+      setError("");
+      try {
+        const payload = await fetchMemoryList();
+        if (!isCurrent()) return;
+        setItems(payload.items);
+        setTotal(payload.total);
+        setNextOffset(payload.skip + payload.items.length);
+        const first = payload.items[0];
+        if (!first) {
+          setSelectedId("");
+          setSelectedBody("");
+          return;
+        }
+        const document = await readMarkdownResource({
+          kind: "agent-memory",
+          id: first.id,
+        });
+        if (!isCurrent()) return;
+        setSelectedId(first.id);
+        setSelectedBody(document.body);
+      } catch (failure) {
+        if (isCurrent()) {
+          setError(failure instanceof Error ? failure.message : "Memories could not be loaded.");
+        }
+      } finally {
+        if (isCurrent()) setLoading(false);
+      }
+    };
+    void load();
+    return () => {
+      cancelled = true;
+      listGenerationRef.current += 1;
+      selectionGenerationRef.current += 1;
+    };
+  }, [reloadToken]);
+
+  const selectMemory = async (memoryId: string) => {
+    if (memoryId === selectedId) return;
+    if (dirty && !window.confirm("Discard the unapplied Markdown draft and open another memory?")) return;
+    const selectionGeneration = ++selectionGenerationRef.current;
+    // Do NOT call handleDirtyChange(false) here: the editor's own onDirtyChange
+    // callback fires automatically when MarkdownContentViewer re-renders with
+    // the new resource prop and its session is cleared.
+    setLoading(true);
+    setError("");
+    try {
+      const document = await readMarkdownResource({
+        kind: "agent-memory",
+        id: memoryId,
+      });
+      if (selectionGeneration !== selectionGenerationRef.current) return;
+      setSelectedId(memoryId);
+      setSelectedBody(document.body);
+    } catch (failure) {
+      if (selectionGeneration === selectionGenerationRef.current) {
+        setError(failure instanceof Error ? failure.message : "The memory could not be loaded.");
+      }
+    } finally {
+      if (selectionGeneration === selectionGenerationRef.current) {
+        setLoading(false);
+      }
+    }
+  };
+
+  const loadMoreMemories = useCallback(async () => {
+    if (loadingMore || nextOffset >= total) return;
+    const listGeneration = ++listGenerationRef.current;
+    setLoadingMore(true);
+    setError("");
+    try {
+      const payload = await fetchMemoryList(nextOffset);
+      if (listGeneration !== listGenerationRef.current) return;
+      setItems((current) => {
+        const knownIds = new Set(current.map((item) => item.id));
+        return [
+          ...current,
+          ...payload.items.filter((item) => !knownIds.has(item.id)),
+        ];
+      });
+      setTotal(payload.total);
+      setNextOffset(payload.skip + payload.items.length);
+    } catch (failure) {
+      if (listGeneration === listGenerationRef.current) {
+        setError(failure instanceof Error ? failure.message : "More memories could not be loaded.");
+      }
+    } finally {
+      if (listGeneration === listGenerationRef.current) {
+        setLoadingMore(false);
+      }
+    }
+  }, [loadingMore, nextOffset, total]);
+
+  const refreshMemorySummaries = useCallback(async () => {
+    const listGeneration = ++listGenerationRef.current;
+    try {
+      const payload = await fetchLoadedMemoryPages(nextOffset);
+      if (listGeneration !== listGenerationRef.current) return;
+      setItems(payload.items);
+      setTotal(payload.total);
+      setNextOffset(payload.nextOffset);
+    } catch {
+      if (listGeneration === listGenerationRef.current) {
+        setError("Memory was applied, but its summary could not be refreshed.");
+      }
+    } finally {
+      if (listGeneration === listGenerationRef.current) {
+        setLoadingMore(false);
+      }
+    }
+  }, [nextOffset]);
+
+  const applyMemory = useCallback((result: MarkdownApplyResult) => {
+    setSelectedBody(result.body);
+    handleDirtyChange(false);
+    setItems((current) => current.map((item) => (
+      item.id === result.resource.id
+        ? { ...item, excerpt: result.body.replace(/\s+/g, " ").slice(0, 160) }
+        : item
+    )));
+    void refreshMemorySummaries();
+  }, [handleDirtyChange, refreshMemorySummaries]);
+
+  return (
+    <div style={workspaceStyle}>
+      <aside style={listPanelStyle} aria-label="Agent memories">
+        <div style={listHeaderStyle}>
+          <div>
+            <div style={listTitleStyle}>AgentMemory</div>
+            <div style={listCountStyle}>{items.length} of {total} loaded</div>
+          </div>
+          <button
+            type="button"
+            aria-label="Refresh memories"
+            onClick={() => setReloadToken((value) => value + 1)}
+            disabled={loading || loadingMore || dirty}
+            title={dirty ? "Apply or cancel the current draft before refreshing" : "Refresh memories"}
+            style={{ ...iconButtonStyle, opacity: loading || loadingMore || dirty ? 0.55 : 1 }}
+          >
+            <RefreshCw size={14} aria-hidden="true" />
+          </button>
+        </div>
+        <div style={memoryListStyle}>
+          {items.map((item) => (
+            <button
+              type="button"
+              key={item.id}
+              onClick={() => void selectMemory(item.id)}
+              aria-current={item.id === selectedId ? "true" : undefined}
+              style={{
+                ...memoryButtonStyle,
+                ...(item.id === selectedId ? selectedMemoryButtonStyle : {}),
+              }}
+            >
+              <span style={memoryTypeStyle}>{item.type}</span>
+              <span style={memoryIdStyle}>{item.id}</span>
+              <span style={memoryExcerptStyle}>{item.excerpt || "Empty memory"}</span>
+            </button>
+          ))}
+          {nextOffset < total ? (
+            <button
+              type="button"
+              aria-label="Load more memories"
+              onClick={() => void loadMoreMemories()}
+              disabled={loading || loadingMore}
+              style={{ ...retryButtonStyle, opacity: loading || loadingMore ? 0.55 : 1 }}
+            >
+              {loadingMore ? "Loading…" : "Load more"}
+            </button>
+          ) : null}
+          {!loading && items.length === 0 ? (
+            <div style={emptyStyle}>
+              <Brain size={22} aria-hidden="true" />
+              <span>No AgentMemory items are available.</span>
+              <button type="button" onClick={() => setReloadToken((value) => value + 1)} style={retryButtonStyle}>
+                Refresh
+              </button>
+            </div>
+          ) : null}
+        </div>
+      </aside>
+
+      <main style={editorPanelStyle}>
+        {error ? <div role="alert" style={alertStyle}>{error}</div> : null}
+        {loading ? (
+          <div role="status" style={emptyStyle}>Loading memories…</div>
+        ) : selectedId ? (
+          <>
+            <div style={selectionHeaderStyle}>
+              <span style={selectionLabelStyle}>Selected memory</span>
+              <strong style={selectionIdStyle}>{selectedId}</strong>
+            </div>
+            <MarkdownContentViewer
+              content={selectedBody}
+              resource={{ kind: "agent-memory", id: selectedId }}
+              onApplied={applyMemory}
+              onDirtyChange={handleDirtyChange}
+            />
+          </>
+        ) : null}
+      </main>
+    </div>
+  );
+}
+
+const workspaceStyle: CSSProperties = {
+  display: "grid",
+  gridTemplateColumns: "minmax(220px, 300px) minmax(0, 1fr)",
+  height: "100%",
+  minHeight: 0,
+  background: GRAPH_THEME.ui.surface.stage,
+};
+
+const listPanelStyle: CSSProperties = {
+  display: "flex",
+  flexDirection: "column",
+  minHeight: 0,
+  borderRight: `1px solid ${GRAPH_THEME.ui.surface.panelBorder}`,
+  background: GRAPH_THEME.ui.surface.panel,
+};
+
+const listHeaderStyle: CSSProperties = {
+  display: "flex",
+  alignItems: "center",
+  justifyContent: "space-between",
+  gap: 12,
+  padding: 16,
+  borderBottom: `1px solid ${GRAPH_THEME.ui.surface.panelBorder}`,
+};
+
+const listTitleStyle: CSSProperties = {
+  color: GRAPH_THEME.ui.text.strong,
+  fontSize: 14,
+  fontWeight: 700,
+};
+
+const listCountStyle: CSSProperties = {
+  color: GRAPH_THEME.ui.text.muted,
+  fontSize: 11,
+  marginTop: 3,
+};
+
+const iconButtonStyle: CSSProperties = {
+  display: "inline-grid",
+  placeItems: "center",
+  width: 30,
+  height: 30,
+  borderRadius: 8,
+  border: `1px solid ${GRAPH_THEME.ui.surface.panelBorder}`,
+  background: "rgba(255, 255, 255, 0.04)",
+  color: GRAPH_THEME.ui.text.body,
+  cursor: "pointer",
+};
+
+const memoryListStyle: CSSProperties = {
+  display: "flex",
+  flexDirection: "column",
+  gap: 6,
+  minHeight: 0,
+  padding: 10,
+  overflowY: "auto",
+};
+
+const memoryButtonStyle: CSSProperties = {
+  display: "flex",
+  flexDirection: "column",
+  alignItems: "flex-start",
+  gap: 5,
+  padding: 10,
+  borderRadius: 9,
+  border: "1px solid transparent",
+  background: "transparent",
+  color: GRAPH_THEME.ui.text.body,
+  textAlign: "left",
+  cursor: "pointer",
+};
+
+const selectedMemoryButtonStyle: CSSProperties = {
+  border: `1px solid ${GRAPH_THEME.ui.control.activeBorder}`,
+  background: GRAPH_THEME.ui.timeline.playheadSoft,
+};
+
+const memoryTypeStyle: CSSProperties = {
+  color: GRAPH_THEME.ui.timeline.playhead,
+  fontSize: 10,
+  fontWeight: 700,
+  textTransform: "uppercase",
+};
+
+const memoryIdStyle: CSSProperties = {
+  maxWidth: "100%",
+  overflow: "hidden",
+  color: GRAPH_THEME.ui.text.strong,
+  fontFamily: "monospace",
+  fontSize: 12,
+  textOverflow: "ellipsis",
+  whiteSpace: "nowrap",
+};
+
+const memoryExcerptStyle: CSSProperties = {
+  display: "-webkit-box",
+  overflow: "hidden",
+  color: GRAPH_THEME.ui.text.muted,
+  fontSize: 11,
+  lineHeight: 1.45,
+  WebkitBoxOrient: "vertical",
+  WebkitLineClamp: 2,
+};
+
+const editorPanelStyle: CSSProperties = {
+  minWidth: 0,
+  minHeight: 0,
+  padding: 20,
+  overflowY: "auto",
+};
+
+const selectionHeaderStyle: CSSProperties = {
+  display: "flex",
+  flexDirection: "column",
+  gap: 4,
+  marginBottom: 12,
+};
+
+const selectionLabelStyle: CSSProperties = {
+  color: GRAPH_THEME.ui.text.muted,
+  fontSize: 11,
+};
+
+const selectionIdStyle: CSSProperties = {
+  color: GRAPH_THEME.ui.text.strong,
+  fontFamily: "monospace",
+  fontSize: 14,
+  wordBreak: "break-all",
+};
+
+const emptyStyle: CSSProperties = {
+  display: "flex",
+  flexDirection: "column",
+  alignItems: "center",
+  justifyContent: "center",
+  gap: 10,
+  minHeight: 160,
+  padding: 20,
+  color: GRAPH_THEME.ui.text.muted,
+  fontSize: 12,
+  textAlign: "center",
+};
+
+const retryButtonStyle: CSSProperties = {
+  padding: "6px 10px",
+  borderRadius: 7,
+  border: `1px solid ${GRAPH_THEME.ui.control.activeBorder}`,
+  background: GRAPH_THEME.ui.timeline.playheadSoft,
+  color: GRAPH_THEME.ui.timeline.playhead,
+  cursor: "pointer",
+};
+
+const alertStyle: CSSProperties = {
+  marginBottom: 12,
+  padding: "10px 12px",
+  borderRadius: 8,
+  border: "1px solid rgba(248, 81, 73, 0.28)",
+  background: "rgba(248, 81, 73, 0.1)",
+  color: "#ffb4ad",
+  fontSize: 12,
+};
@@ -42,6 +42,9 @@ async function startVite(): Promise<void> {
 }

 async function installApiFixture(page: Page): Promise<void> {
+  await page.route("**/api/info", async (route) => {
+    await route.fulfill({ json: { capabilities: { agent_memory: false } } });
+  });
  await page.route("**/api/graph/**", async (route) => {
    const pathname = new URL(route.request().url()).pathname;
    if (pathname === "/api/graph/stats") {
@@ -0,0 +1,74 @@
+import assert from "node:assert/strict";
+import test from "node:test";
+
+import { JSDOM } from "jsdom";
+import React from "react";
+
+import { fetchAgentMemoryAvailability } from "../src/explorerCapabilities.ts";
+
+(globalThis as typeof globalThis & { React: typeof React }).React = React;
+const dom = new JSDOM("<!doctype html><html><body></body></html>", {
+  url: "http://localhost",
+});
+Object.assign(globalThis, {
+  window: dom.window,
+  document: dom.window.document,
+  HTMLElement: dom.window.HTMLElement,
+  Node: dom.window.Node,
+});
+Object.defineProperty(globalThis, "navigator", {
+  configurable: true,
+  value: dom.window.navigator,
+});
+
+const { cleanup, render } = await import("@testing-library/react");
+const { ExploreWorkspaceTabs } = await import("../src/ExploreWorkspaceTabs.tsx");
+
+test.afterEach(cleanup);
+
+test("reports AgentMemory when the Explorer host provides it", async () => {
+  const available = await fetchAgentMemoryAvailability(async () => (
+    new Response(
+      JSON.stringify({ capabilities: { agent_memory: true } }),
+      { status: 200, headers: { "content-type": "application/json" } },
+    )
+  ));
+
+  assert.equal(available, true);
+});
+
+test("keeps AgentMemory hidden when the capability is absent or unavailable", async () => {
+  const absent = await fetchAgentMemoryAvailability(async () => (
+    new Response(JSON.stringify({ status: "active" }), { status: 200 })
+  ));
+  const unavailable = await fetchAgentMemoryAvailability(async () => {
+    throw new Error("network unavailable");
+  });
+
+  assert.equal(absent, false);
+  assert.equal(unavailable, false);
+});
+
+test("shows the Memories tab only when the host provides AgentMemory", () => {
+  const availableView = render(
+    <ExploreWorkspaceTabs
+      activeView="graph"
+      agentMemoryAvailable
+      onSelect={() => undefined}
+    />,
+  );
+  assert.ok(availableView.getByRole("button", { name: "Memories" }));
+  cleanup();
+
+  const unavailableView = render(
+    <ExploreWorkspaceTabs
+      activeView="graph"
+      agentMemoryAvailable={false}
+      onSelect={() => undefined}
+    />,
+  );
+  assert.equal(
+    unavailableView.queryByRole("button", { name: "Memories" }),
+    null,
+  );
+});
@@ -3,7 +3,7 @@ import assert from "node:assert/strict";
 import React from "react";
 import { renderToString } from "react-dom/server";

-(globalThis as any).React = React;
+(globalThis as typeof globalThis & { React: typeof React }).React = React;

 import { MarkdownContentViewer } from "../src/workspaces/GraphWorkspace/MarkdownContentViewer.tsx";
 import { isSafeUrl } from "../src/workspaces/GraphWorkspace/markdownUrlSafety.ts";
@@ -60,6 +60,29 @@ test("renders Preview mode with formatted Markdown elements and tabs", () => {
  assert.equal(html.includes("Item B"), true);
 });

+test("stays read-only without a resource and exposes Edit for canonical resources", () => {
+  const readOnly = renderToString(React.createElement(MarkdownContentViewer, {
+    content: "Read-only body",
+  }));
+  const editable = renderToString(React.createElement(MarkdownContentViewer, {
+    content: "Editable body",
+    resource: { kind: "context-node", id: "node-1" },
+  }));
+
+  assert.equal(readOnly.includes(">Edit</button>"), false);
+  assert.equal(editable.includes(">Edit</button>"), true);
+});
+
+test("empty canonical resources still expose Edit", () => {
+  const html = renderToString(React.createElement(MarkdownContentViewer, {
+    content: "",
+    resource: { kind: "context-node", id: "empty-node" },
+  }));
+
+  assert.equal(html.includes("No content available for this node."), true);
+  assert.equal(html.includes(">Edit</button>"), true);
+});
+
 test("renders Source mode with exact unmodified text inside pre/code", () => {
  const markdown = `# Title 🚀\n\n  * Indented item\n\n\`\`\`python\ndef test():\n    return "α + β"\n\`\`\``;
  const html = renderToString(React.createElement(MarkdownContentViewer, { content: markdown, defaultMode: "source" }));
@@ -0,0 +1,639 @@
+import assert from "node:assert/strict";
+import test from "node:test";
+
+import { JSDOM } from "jsdom";
+import React from "react";
+
+(globalThis as typeof globalThis & { React: typeof React }).React = React;
+const dom = new JSDOM("<!doctype html><html><body></body></html>", {
+  url: "http://localhost",
+});
+Object.assign(globalThis, {
+  window: dom.window,
+  document: dom.window.document,
+  HTMLElement: dom.window.HTMLElement,
+  Node: dom.window.Node,
+});
+Object.defineProperty(globalThis, "navigator", {
+  configurable: true,
+  value: dom.window.navigator,
+});
+dom.window.confirm = () => true;
+
+// Testing Library and the components must load after the jsdom globals above.
+const { act, cleanup, fireEvent, render, waitFor } = await import(
+  "@testing-library/react"
+);
+const { MarkdownContentViewer } = await import(
+  "../src/workspaces/GraphWorkspace/MarkdownContentViewer.tsx"
+);
+const { MemoryWorkspace } = await import(
+  "../src/workspaces/MemoryWorkspace.tsx"
+);
+
+test.afterEach(() => {
+  cleanup();
+  dom.window.confirm = () => true;
+});
+
+const resource = { kind: "context-node" as const, id: "node-1" };
+const originalSource = "---\nid: node-1\ntype: Note\n---\n\nOriginal";
+
+function jsonResponse(body: unknown, status = 200): Response {
+  return new Response(JSON.stringify(body), {
+    status,
+    headers: { "Content-Type": "application/json" },
+  });
+}
+
+test("Edit and Apply send canonical Markdown and publish the applied result", async () => {
+  const requests: Array<{ url: string; init?: RequestInit }> = [];
+  let appliedBody = "";
+  globalThis.fetch = async (input, init) => {
+    requests.push({ url: String(input), init });
+    if (init?.method === "PUT") {
+      return jsonResponse({
+        resource,
+        source: originalSource.replace("Original", "Updated"),
+        body: "Updated",
+        revision: "sha256:updated",
+        editable: true,
+        changed: true,
+      });
+    }
+    return jsonResponse({
+      resource,
+      source: originalSource,
+      body: "Original",
+      revision: "sha256:original",
+      editable: true,
+    });
+  };
+
+  const view = render(
+    <MarkdownContentViewer
+      content="Original"
+      resource={resource}
+      onApplied={(result) => { appliedBody = result.body; }}
+    />,
+  );
+  fireEvent.click(view.getByRole("button", { name: "Edit" }));
+  const textarea = await view.findByRole("textbox", { name: "Markdown source" });
+  fireEvent.input(textarea, {
+    target: { value: originalSource.replace("Original", "Updated") },
+  });
+  fireEvent.click(view.getByRole("button", { name: "Apply" }));
+
+  await waitFor(() => assert.equal(appliedBody, "Updated"));
+  assert.deepEqual(requests.map(({ init }) => init?.method ?? "GET"), ["GET", "PUT"]);
+  assert.equal(
+    JSON.parse(String(requests[1].init?.body)).expected_revision,
+    "sha256:original",
+  );
+  assert.equal(
+    view.getByRole("tab", { name: "Preview" }).getAttribute("aria-selected"),
+    "true",
+  );
+});
+
+test("Cancel restores the previous view and never sends a PUT", async () => {
+  const methods: string[] = [];
+  globalThis.fetch = async (_input, init) => {
+    methods.push(init?.method ?? "GET");
+    return jsonResponse({
+      resource,
+      source: originalSource,
+      body: "Original",
+      revision: "sha256:original",
+      editable: true,
+    });
+  };
+
+  const view = render(<MarkdownContentViewer content="Original" resource={resource} />);
+  fireEvent.click(view.getByRole("button", { name: "Edit" }));
+  const textarea = await view.findByRole("textbox", { name: "Markdown source" });
+  fireEvent.input(textarea, {
+    target: { value: originalSource.replace("Original", "Draft") },
+  });
+  fireEvent.click(view.getByRole("button", { name: "Cancel" }));
+
+  assert.deepEqual(methods, ["GET"]);
+  assert.equal(view.queryByRole("textbox", { name: "Markdown source" }), null);
+  assert.equal(
+    view.getByRole("tab", { name: "Preview" }).getAttribute("aria-selected"),
+    "true",
+  );
+});
+
+test("validation failures keep the draft visible for correction", async () => {
+  globalThis.fetch = async (_input, init) => {
+    if (init?.method === "PUT") {
+      return jsonResponse({
+        detail: {
+          code: "invalid_markdown_frontmatter",
+          message: "Markdown frontmatter contains invalid YAML.",
+        },
+      }, 422);
+    }
+    return jsonResponse({
+      resource,
+      source: originalSource,
+      body: "Original",
+      revision: "sha256:original",
+      editable: true,
+    });
+  };
+
+  const view = render(<MarkdownContentViewer content="Original" resource={resource} />);
+  fireEvent.click(view.getByRole("button", { name: "Edit" }));
+  const textarea = await view.findByRole("textbox", { name: "Markdown source" });
+  const invalidDraft = "---\nid: [\n---\n\nDraft";
+  fireEvent.input(textarea, { target: { value: invalidDraft } });
+  fireEvent.click(view.getByRole("button", { name: "Apply" }));
+
+  const alert = await view.findByRole("alert");
+  assert.match(alert.textContent ?? "", /invalid YAML/);
+  assert.equal(
+    (view.getByRole("textbox", { name: "Markdown source" }) as HTMLTextAreaElement)
+      .value,
+    invalidDraft,
+  );
+});
+
+
+test("resource changes discard the previous editor session", async () => {
+  globalThis.fetch = async (input) => {
+    const id = String(input).endsWith("node-2") ? "node-2" : "node-1";
+    return jsonResponse({
+      resource: { kind: "context-node", id },
+      source: `---\nid: ${id}\ntype: Note\n---\n\n${id}`,
+      body: id,
+      revision: `sha256:${id}`,
+      editable: true,
+    });
+  };
+
+  const view = render(
+    <MarkdownContentViewer content="node-1" resource={resource} />,
+  );
+  fireEvent.click(view.getByRole("button", { name: "Edit" }));
+  const textarea = await view.findByRole("textbox", { name: "Markdown source" });
+  fireEvent.input(textarea, {
+    target: { value: `${(textarea as HTMLTextAreaElement).value}\nDraft` },
+  });
+
+  view.rerender(
+    <MarkdownContentViewer
+      content="node-2"
+      resource={{ kind: "context-node", id: "node-2" }}
+    />,
+  );
+  await waitFor(() => assert.equal(
+    view.queryByRole("textbox", { name: "Markdown source" }),
+    null,
+  ));
+
+  view.rerender(
+    <MarkdownContentViewer content="node-1" resource={resource} />,
+  );
+  await waitFor(() => assert.equal(
+    view.queryByRole("textbox", { name: "Markdown source" }),
+    null,
+  ));
+  assert.ok(view.getByRole("button", { name: "Edit" }));
+});
+
+
+test("unmounting a dirty editor clears the parent dirty guard", async () => {
+  const dirtyStates: boolean[] = [];
+  globalThis.fetch = async () => jsonResponse({
+    resource,
+    source: originalSource,
+    body: "Original",
+    revision: "sha256:original",
+    editable: true,
+  });
+
+  const view = render(
+    <MarkdownContentViewer
+      content="Original"
+      resource={resource}
+      onDirtyChange={(dirty) => dirtyStates.push(dirty)}
+    />,
+  );
+  fireEvent.click(view.getByRole("button", { name: "Edit" }));
+  const textarea = await view.findByRole("textbox", { name: "Markdown source" });
+  fireEvent.input(textarea, {
+    target: { value: originalSource.replace("Original", "Draft") },
+  });
+  await waitFor(() => assert.equal(dirtyStates.at(-1), true));
+
+  view.unmount();
+
+  assert.equal(dirtyStates.at(-1), false);
+});
+
+test("MemoryWorkspace protects a dirty memory draft when selection changes", async () => {
+  const requestedUrls: string[] = [];
+  globalThis.fetch = async (input) => {
+    const url = String(input);
+    requestedUrls.push(url);
+    if (url.startsWith("/api/memories")) {
+      return jsonResponse({
+        items: [
+          { id: "mem-1", type: "note", excerpt: "First", updated_at: null },
+          { id: "mem-2", type: "note", excerpt: "Second", updated_at: null },
+        ],
+        total: 2,
+        skip: 0,
+        limit: 100,
+      });
+    }
+    const id = url.endsWith("mem-2") ? "mem-2" : "mem-1";
+    return jsonResponse({
+      resource: { kind: "agent-memory", id },
+      source: `---\nid: ${id}\ntype: note\n---\n\n${id}`,
+      body: id,
+      revision: `sha256:${id}`,
+      editable: true,
+    });
+  };
+
+  const view = render(<MemoryWorkspace />);
+  await view.findByText("Selected memory");
+  fireEvent.click(view.getByRole("button", { name: "Edit" }));
+  const textarea = await view.findByRole("textbox", { name: "Markdown source" });
+  fireEvent.input(textarea, {
+    target: { value: `${(textarea as HTMLTextAreaElement).value}\nDraft` },
+  });
+
+  dom.window.confirm = () => false;
+  fireEvent.click(view.getByRole("button", { name: /mem-2/ }));
+
+  assert.equal(requestedUrls.some((url) => url.endsWith("mem-2")), false);
+  assert.equal(view.getByText("mem-1", { selector: "strong" }).textContent, "mem-1");
+});
+
+
+test("MemoryWorkspace loads memories beyond the first server page", async () => {
+  const requestedUrls: string[] = [];
+  const firstPage = Array.from({ length: 100 }, (_, index) => ({
+    id: `mem-${index + 1}`,
+    type: "note",
+    excerpt: `Memory ${index + 1}`,
+    updated_at: null,
+  }));
+  globalThis.fetch = async (input) => {
+    const url = String(input);
+    requestedUrls.push(url);
+    if (url === "/api/memories?skip=0&limit=100") {
+      return jsonResponse({
+        items: firstPage,
+        total: 101,
+        skip: 0,
+        limit: 100,
+      });
+    }
+    if (url === "/api/memories?skip=100&limit=100") {
+      return jsonResponse({
+        items: [{
+          id: "mem-101",
+          type: "note",
+          excerpt: "Memory 101",
+          updated_at: null,
+        }],
+        total: 101,
+        skip: 100,
+        limit: 100,
+      });
+    }
+    return jsonResponse({
+      resource: { kind: "agent-memory", id: "mem-1" },
+      source: "---\nid: mem-1\ntype: note\n---\n\nmem-1",
+      body: "mem-1",
+      revision: "sha256:mem-1",
+      editable: true,
+    });
+  };
+
+  const view = render(<MemoryWorkspace />);
+  await view.findByText("Selected memory");
+  fireEvent.click(view.getByRole("button", { name: "Load more memories" }));
+
+  await view.findByRole("button", { name: /mem-101/ });
+  assert.ok(requestedUrls.includes("/api/memories?skip=100&limit=100"));
+  assert.equal(view.getByText("101 of 101 loaded").textContent, "101 of 101 loaded");
+});
+
+
+test("MemoryWorkspace ignores stale selection responses", async () => {
+  let resolveMem2: ((response: Response) => void) | undefined;
+  let resolveMem3: ((response: Response) => void) | undefined;
+  const mem2Response = new Promise<Response>((resolve) => {
+    resolveMem2 = resolve;
+  });
+  const mem3Response = new Promise<Response>((resolve) => {
+    resolveMem3 = resolve;
+  });
+
+  globalThis.fetch = async (input) => {
+    const url = String(input);
+    if (url.startsWith("/api/memories")) {
+      return jsonResponse({
+        items: [
+          { id: "mem-1", type: "note", excerpt: "First", updated_at: null },
+          { id: "mem-2", type: "note", excerpt: "Second", updated_at: null },
+          { id: "mem-3", type: "note", excerpt: "Third", updated_at: null },
+        ],
+        total: 3,
+        skip: 0,
+        limit: 100,
+      });
+    }
+    if (url.endsWith("mem-2")) return mem2Response;
+    if (url.endsWith("mem-3")) return mem3Response;
+    return jsonResponse({
+      resource: { kind: "agent-memory", id: "mem-1" },
+      source: "---\nid: mem-1\ntype: note\n---\n\nmem-1",
+      body: "mem-1",
+      revision: "sha256:mem-1",
+      editable: true,
+    });
+  };
+
+  const view = render(<MemoryWorkspace />);
+  await view.findByText("Selected memory");
+  fireEvent.click(view.getByRole("button", { name: /mem-2/ }));
+  fireEvent.click(view.getByRole("button", { name: /mem-3/ }));
+
+  await act(async () => {
+    resolveMem3?.(jsonResponse({
+      resource: { kind: "agent-memory", id: "mem-3" },
+      source: "---\nid: mem-3\ntype: note\n---\n\nmem-3",
+      body: "mem-3",
+      revision: "sha256:mem-3",
+      editable: true,
+    }));
+    await mem3Response;
+  });
+  await waitFor(() => assert.equal(
+    view.getByText("mem-3", { selector: "strong" }).textContent,
+    "mem-3",
+  ));
+
+  await act(async () => {
+    resolveMem2?.(jsonResponse({
+      resource: { kind: "agent-memory", id: "mem-2" },
+      source: "---\nid: mem-2\ntype: note\n---\n\nmem-2",
+      body: "mem-2",
+      revision: "sha256:mem-2",
+      editable: true,
+    }));
+    await mem2Response;
+  });
+
+  assert.equal(
+    view.getByText("mem-3", { selector: "strong" }).textContent,
+    "mem-3",
+  );
+});
+
+
+test("MemoryWorkspace refreshes frontmatter summaries after apply", async () => {
+  let listRequests = 0;
+  globalThis.fetch = async (input, init) => {
+    const url = String(input);
+    if (url.startsWith("/api/memories")) {
+      listRequests += 1;
+      const saved = listRequests > 1;
+      return jsonResponse({
+        items: [{
+          id: "mem-1",
+          type: saved ? "decision" : "note",
+          excerpt: saved ? "Updated memory" : "Original memory",
+          updated_at: saved ? "2026-09-01T12:00:00+00:00" : null,
+        }],
+        total: 1,
+        skip: 0,
+        limit: 100,
+      });
+    }
+    if (init?.method === "PUT") {
+      return jsonResponse({
+        resource: { kind: "agent-memory", id: "mem-1" },
+        source: "---\nid: mem-1\ntype: decision\n---\n\nUpdated memory",
+        body: "Updated memory",
+        revision: "sha256:updated",
+        editable: true,
+        changed: true,
+      });
+    }
+    return jsonResponse({
+      resource: { kind: "agent-memory", id: "mem-1" },
+      source: "---\nid: mem-1\ntype: note\n---\n\nOriginal memory",
+      body: "Original memory",
+      revision: "sha256:original",
+      editable: true,
+    });
+  };
+
+  const view = render(<MemoryWorkspace />);
+  await view.findByText("Selected memory");
+  fireEvent.click(view.getByRole("button", { name: "Edit" }));
+  const textarea = await view.findByRole("textbox", { name: "Markdown source" });
+  fireEvent.input(textarea, {
+    target: {
+      value: (textarea as HTMLTextAreaElement).value
+        .replace("type: note", "type: decision")
+        .replace("Original memory", "Updated memory"),
+    },
+  });
+  fireEvent.click(view.getByRole("button", { name: "Apply" }));
+
+  await view.findByText("decision");
+  assert.equal(listRequests, 2);
+  assert.equal(view.getAllByText("Updated memory").length, 2);
+});
+
+test("HTTP 409 conflict preserves draft and shows conflict error with reload option", async () => {
+  // After a 409, the user's draft must be kept and a recovery path available.
+  const requests: Array<{ method: string; body?: unknown }> = [];
+  let fetchCount = 0;
+
+  globalThis.fetch = async (input, init) => {
+    fetchCount += 1;
+    const method = init?.method ?? "GET";
+    let parsedBody: unknown = undefined;
+    if (init?.body) {
+      try { parsedBody = JSON.parse(String(init.body)); } catch { /* ignore */ }
+    }
+    requests.push({ method, body: parsedBody });
+
+    if (method === "PUT") {
+      // First PUT returns 409 with current_revision
+      return jsonResponse({
+        detail: {
+          code: "markdown_revision_conflict",
+          message: "This item changed after editing began. Reload the latest version before applying.",
+          current_revision: "sha256:newer",
+        },
+      }, 409);
+    }
+    // All GETs return the canonical document
+    return jsonResponse({
+      resource,
+      source: originalSource,
+      body: "Original",
+      revision: "sha256:original",
+      editable: true,
+    });
+  };
+
+  const view = render(<MarkdownContentViewer content="Original" resource={resource} />);
+
+  // Enter edit mode
+  fireEvent.click(view.getByRole("button", { name: "Edit" }));
+  const textarea = await view.findByRole("textbox", { name: "Markdown source" });
+  const draftValue = originalSource.replace("Original", "My draft");
+  fireEvent.input(textarea, { target: { value: draftValue } });
+
+  // Apply → receives 409
+  fireEvent.click(view.getByRole("button", { name: "Apply" }));
+
+  // Conflict error must appear
+  const alert = await view.findByRole("alert");
+  assert.match(
+    alert.textContent ?? "",
+    /changed after editing|Reload/i,
+    "conflict error message must be shown",
+  );
+
+  // Draft must be preserved in the textarea
+  const textareaAfterConflict = view.getByRole("textbox", { name: "Markdown source" }) as HTMLTextAreaElement;
+  assert.equal(textareaAfterConflict.value, draftValue, "draft must be preserved after 409");
+
+  // A reload / recovery action must be available
+  const reloadButton = view.queryByRole("button", { name: /reload latest/i });
+  assert.ok(reloadButton !== null, "a 'Reload latest' recovery button must be shown");
+
+  // Click reload — should re-fetch the latest canonical document
+  await act(async () => {
+    fireEvent.click(reloadButton!);
+  });
+
+  // After reload the editor is re-initialized with the server's canonical source
+  await waitFor(() => {
+    const refreshedTextarea = view.queryByRole("textbox", { name: "Markdown source" });
+    assert.ok(refreshedTextarea !== null, "editor must still be open after reload");
+    assert.equal(
+      (refreshedTextarea as HTMLTextAreaElement).value,
+      originalSource,
+      "editor must show the server canonical source after reload",
+    );
+  });
+
+  // Reload must have triggered exactly one more GET
+  const getCount = requests.filter((r) => r.method === "GET").length;
+  assert.ok(getCount >= 2, "reload must issue a new GET to fetch the latest canonical document");
+});
+
+
+test("successful retry after 422 uses the original revision and persists changes", async () => {
+  // After a 422 (validation failure), the baseRevision must remain valid so that
+  // correcting the draft and re-applying succeeds without re-fetching the document.
+  let putCallCount = 0;
+
+  globalThis.fetch = async (_input, init) => {
+    const method = init?.method ?? "GET";
+    if (method === "PUT") {
+      putCallCount += 1;
+      if (putCallCount === 1) {
+        // First PUT: validation failure — resource is unchanged
+        return jsonResponse({
+          detail: {
+            code: "invalid_markdown_frontmatter",
+            message: "Markdown frontmatter contains invalid YAML.",
+          },
+        }, 422);
+      }
+      // Second PUT: success with the corrected Markdown
+      const body = JSON.parse(String(init?.body ?? "{}")) as { markdown: string };
+      const correctedBody = body.markdown.includes("Corrected") ? "Corrected body" : "body";
+      return jsonResponse({
+        resource,
+        source: originalSource.replace("Original", "Corrected"),
+        body: correctedBody,
+        revision: "sha256:after-retry",
+        editable: true,
+        changed: true,
+      });
+    }
+    return jsonResponse({
+      resource,
+      source: originalSource,
+      body: "Original",
+      revision: "sha256:original",
+      editable: true,
+    });
+  };
+
+  let appliedRevision = "";
+  const view = render(
+    <MarkdownContentViewer
+      content="Original"
+      resource={resource}
+      onApplied={(result) => { appliedRevision = result.revision; }}
+    />,
+  );
+
+  // Enter edit mode
+  fireEvent.click(view.getByRole("button", { name: "Edit" }));
+  const textarea = await view.findByRole("textbox", { name: "Markdown source" });
+
+  // First attempt: create an invalid draft
+  const invalidDraft = "---\nid: [\n---\n\nInvalid body";
+  fireEvent.input(textarea, { target: { value: invalidDraft } });
+  fireEvent.click(view.getByRole("button", { name: "Apply" }));
+
+  // 422 error appears, draft is preserved
+  const alert = await view.findByRole("alert");
+  assert.match(alert.textContent ?? "", /invalid YAML/i);
+  assert.equal(
+    (view.getByRole("textbox", { name: "Markdown source" }) as HTMLTextAreaElement).value,
+    invalidDraft,
+    "invalid draft must be preserved after 422",
+  );
+
+  // Correct the draft
+  const correctedDraft = originalSource.replace("Original", "Corrected");
+  fireEvent.input(view.getByRole("textbox", { name: "Markdown source" }), {
+    target: { value: correctedDraft },
+  });
+
+  // Apply is re-enabled (still dirty)
+  const applyButton = view.getByRole("button", { name: "Apply" });
+  assert.equal(
+    (applyButton as HTMLButtonElement).disabled,
+    false,
+    "Apply must be re-enabled after correcting the draft",
+  );
+
+  // Second attempt: apply corrected draft
+  fireEvent.click(applyButton);
+
+  // Must succeed — server returns new revision
+  await waitFor(() => assert.equal(appliedRevision, "sha256:after-retry"));
+
+  // Editor returns to preview mode after successful save
+  assert.equal(
+    view.getByRole("tab", { name: "Preview" }).getAttribute("aria-selected"),
+    "true",
+    "editor must return to preview after successful retry",
+  );
+
+  // Error is cleared
+  assert.equal(view.queryByRole("alert"), null, "error banner must be cleared after success");
+
+  // Both PUT attempts were made — retry used original revision (no extra GET between attempts)
+  assert.equal(putCallCount, 2, "exactly two PUT requests must be made (failed + successful retry)");
+});
@@ -0,0 +1,132 @@
+import assert from "node:assert/strict";
+import test from "node:test";
+
+import {
+  cancelEdit,
+  createEditSession,
+  createLoadingSession,
+  isDirty,
+  saveFailed,
+  saveStarted,
+  saveSucceeded,
+  shouldConfirmDiscard,
+  updateDraft,
+} from "../src/workspaces/GraphWorkspace/markdownEditorState.ts";
+
+const resource = { kind: "context-node" as const, id: "node-1" };
+
+
+test("enters loading and editing with the canonical source", () => {
+  const loading = createLoadingSession(resource);
+  const editing = createEditSession(resource, {
+    source: "---\nid: node-1\n---\n\nBody",
+    revision: "sha256:one",
+  });
+
+  assert.equal(loading.status, "loading-document");
+  assert.equal(editing.status, "editing");
+  assert.equal(editing.draft, editing.baseSource);
+  assert.equal(isDirty(editing), false);
+});
+
+
+test("draft changes derive dirty state and clear prior errors", () => {
+  const session = createEditSession(resource, {
+    source: "base",
+    revision: "sha256:one",
+  });
+  const failed = saveFailed(session, {
+    kind: "validation",
+    message: "Invalid",
+  });
+  const edited = updateDraft(failed, "draft");
+
+  assert.equal(edited.status, "editing");
+  assert.equal(edited.error, null);
+  assert.equal(isDirty(edited), true);
+  assert.equal(shouldConfirmDiscard(edited), true);
+});
+
+
+test("cancel discards the edit session without saving", () => {
+  const session = updateDraft(
+    createEditSession(resource, { source: "base", revision: "sha256:one" }),
+    "draft",
+  );
+
+  assert.equal(isDirty(session), true);
+  assert.equal(cancelEdit(), null);
+});
+
+
+test("no-op save never enters saving state", () => {
+  const session = createEditSession(resource, {
+    source: "base",
+    revision: "sha256:one",
+  });
+
+  assert.equal(saveStarted(session), session);
+  assert.equal(isDirty(session), false);
+});
+
+
+test("save success replaces the base source and revision", () => {
+  const session = saveStarted(
+    updateDraft(
+      createEditSession(resource, { source: "base", revision: "sha256:one" }),
+      "draft",
+    ),
+  );
+  const saved = saveSucceeded(session, {
+    source: "canonical saved",
+    revision: "sha256:two",
+  });
+
+  assert.equal(saved.status, "viewing");
+  assert.equal(saved.baseSource, "canonical saved");
+  assert.equal(saved.draft, "canonical saved");
+  assert.equal(saved.baseRevision, "sha256:two");
+  assert.equal(isDirty(saved), false);
+});
+
+
+test("validation, conflict, and save failures retain the draft for retry", () => {
+  const draft = updateDraft(
+    createEditSession(resource, { source: "base", revision: "sha256:one" }),
+    "draft",
+  );
+
+  const validation = saveFailed(draft, {
+    kind: "validation",
+    message: "Invalid",
+  });
+  const conflict = saveFailed(draft, {
+    kind: "conflict",
+    message: "Stale",
+    currentRevision: "sha256:two",
+  });
+  const network = saveFailed(draft, {
+    kind: "network",
+    message: "Offline",
+  });
+
+  assert.equal(validation.status, "validation-error");
+  assert.equal(conflict.status, "conflict");
+  assert.equal(network.status, "save-error");
+  assert.equal(validation.draft, "draft");
+  assert.equal(conflict.draft, "draft");
+  assert.equal(network.draft, "draft");
+  assert.equal(saveStarted(network).status, "saving");
+});
+
+
+test("saving sessions do not allow a competing discard confirmation", () => {
+  const saving = saveStarted(
+    updateDraft(
+      createEditSession(resource, { source: "base", revision: "sha256:one" }),
+      "draft",
+    ),
+  );
+
+  assert.equal(shouldConfirmDiscard(saving), false);
+});
@@ -0,0 +1,91 @@
+import assert from "node:assert/strict";
+import test from "node:test";
+
+import {
+  NodeMarkdownRefreshGuard,
+  buildNodeMarkdownAttributeUpdate,
+  readNodeMarkdownAttributeUpdate,
+} from "../src/workspaces/GraphWorkspace/nodeMarkdownSync.ts";
+
+
+test("saved Markdown updates graph content and its visible label", () => {
+  const update = buildNodeMarkdownAttributeUpdate(
+    "issue-1327",
+    "# Issue 1328888\n\nUpdated body",
+    { status: "implemented" },
+  );
+
+  assert.equal(update.content, "# Issue 1328888\n\nUpdated body");
+  assert.equal(update.label, "# Issue 1328888\n\nUpdated body");
+  assert.deepEqual(update.properties, {
+    status: "implemented",
+    content: "# Issue 1328888\n\nUpdated body",
+  });
+});
+
+
+test("empty Markdown falls back to the stable node id label", () => {
+  const update = buildNodeMarkdownAttributeUpdate("issue-1327", "", {});
+
+  assert.equal(update.label, "issue-1327");
+  assert.equal(update.content, "");
+});
+
+
+test("saved frontmatter is refreshed from the canonical graph node", async () => {
+  const update = await readNodeMarkdownAttributeUpdate(
+    "node/1",
+    async (input) => {
+      assert.equal(String(input), "/api/graph/node?node_id=node%2F1");
+      return new Response(JSON.stringify({
+        id: "node/1",
+        type: "Decision",
+        content: "Updated body",
+        properties: {
+          content: "Updated body",
+          status: "accepted",
+          valid_from: "2026-09-01T00:00:00Z",
+        },
+        valid_from: "2026-09-01T00:00:00Z",
+        valid_until: null,
+      }), {
+        status: 200,
+        headers: { "Content-Type": "application/json" },
+      });
+    },
+  );
+  assert.deepEqual(update, {
+    label: "Updated body",
+    content: "Updated body",
+    properties: {
+      content: "Updated body",
+      status: "accepted",
+      valid_from: "2026-09-01T00:00:00Z",
+    },
+    nodeType: "Decision",
+    valid_from: "2026-09-01T00:00:00Z",
+    valid_until: null,
+  });
+});
+
+
+test("realtime updates invalidate an older local-save refresh", () => {
+  const guard = new NodeMarkdownRefreshGuard();
+  const localSaveRefresh = guard.begin("node-1");
+
+  guard.invalidate("node-1");
+
+  assert.equal(guard.isCurrent("node-1", localSaveRefresh), false);
+});
+
+
+test("refresh invalidation is scoped to one node", () => {
+  const guard = new NodeMarkdownRefreshGuard();
+  const firstNodeRefresh = guard.begin("node-1");
+  const secondNodeRefresh = guard.begin("node-2");
+
+  guard.invalidate("node-1");
+
+  assert.equal(guard.isCurrent("node-1", firstNodeRefresh), false);
+  assert.equal(guard.isCurrent("node-2", secondNodeRefresh), true);
+});
@@ -65,9 +65,11 @@ import os
 import re
 import stat
 import tempfile
+import threading
 from collections import deque
 from dataclasses import dataclass, field
 from datetime import date, datetime, timedelta, timezone
+from functools import wraps
 from pathlib import Path
 from typing import Any, Dict, List, Optional, Tuple, Union

@@ -78,6 +80,12 @@ from ..utils.logging import get_logger
 from ..utils.progress_tracker import get_progress_tracker
 from ..utils.types import EntityDict, RelationshipDict
 from ._markdown_filesystem import find_filesystem_link
+from .markdown import (
+    MarkdownIdentityError,
+    MarkdownResourceNotFoundError,
+    MarkdownRevisionConflictError,
+    markdown_document_revision,
+)


 class _UniqueKeySafeLoader(yaml.SafeLoader):
@@ -158,6 +166,17 @@ class MemoryItem:
        )


+def _with_memory_lock(method):
+    """Serialize AgentMemory state mutations and Markdown revision checks."""
+
+    @wraps(method)
+    def locked(self, *args, **kwargs):
+        with self._memory_lock:
+            return method(self, *args, **kwargs)
+
+    return locked
+
+
 class AgentMemory:
    """
    Agent memory manager with RAG integration and Hierarchical Memory.
@@ -197,6 +216,7 @@ class AgentMemory:
        self.logger = get_logger("agent_memory")
        self.config = config or {}
        self.config.update(kwargs)
+        self._memory_lock = threading.RLock()

        self.vector_store = self.config.get("vector_store")
        self.knowledge_graph = self.config.get("knowledge_graph")
@@ -249,6 +269,7 @@ class AgentMemory:

        self.logger.info(f"Saved agent memory to {path}")

+    @_with_memory_lock
    def load(self, path: str) -> None:
        """
        Load memory state from disk.
@@ -298,6 +319,7 @@ class AgentMemory:

        self.logger.info(f"Loaded agent memory from {path}")

+    @_with_memory_lock
    def store(
        self,
        content: str,
@@ -574,6 +596,7 @@ class AgentMemory:
            "relationships": memory_item.relationships,
        }

+    @_with_memory_lock
    def delete_memory(self, memory_id: str, *, skip_vector: bool = False) -> bool:
        """
        Delete memory item.
@@ -625,6 +648,7 @@ class AgentMemory:
        self.logger.debug(f"Deleted memory item: {memory_id}")
        return True

+    @_with_memory_lock
    def vector_ids_for(self, memory_id: str) -> List[str]:
        """Return the vector-store ids owned by a memory item.

@@ -1136,6 +1160,7 @@ class AgentMemory:
        """
        return self.get_memory(memory_id)

+    @_with_memory_lock
    def update(
        self,
        memory_id: str,
@@ -1416,6 +1441,13 @@ class AgentMemory:

        return results

+    @_with_memory_lock
+    def list_snapshot(
+        self, *, limit: int = 100, offset: int = 0
+    ) -> Tuple[List[Dict[str, Any]], int]:
+        """Return one memory page and its total from the same locked state."""
+        return self.list(limit=limit, offset=offset), len(self.memory_items)
+
    def get_by_conversation(
        self, conversation_id: str, limit: int = 100
    ) -> List[Dict[str, Any]]:
@@ -1631,6 +1663,88 @@ class AgentMemory:
                updated += 1
        return updated

+    @_with_memory_lock
+    def export_item_markdown(self, memory_id: str) -> str:
+        """Return one existing memory item as canonical Markdown.
+
+        Args:
+            memory_id: Stable identifier of the memory item to export.
+
+        Returns:
+            Canonical Markdown containing the memory frontmatter and body.
+
+        Raises:
+            MarkdownResourceNotFoundError: If ``memory_id`` does not exist.
+        """
+        memory = self.get(memory_id)
+        if memory is None:
+            raise MarkdownResourceNotFoundError(
+                f"AgentMemory item {memory_id!r} was not found."
+            )
+        return self._memory_to_markdown(memory)
+
+    @_with_memory_lock
+    def apply_item_markdown(
+        self,
+        memory_id: str,
+        document: str,
+        *,
+        expected_revision: Optional[str] = None,
+    ) -> bool:
+        """Validate and atomically replace one existing memory item.
+
+        Args:
+            memory_id: Stable identifier of the memory item to update.
+            document: Canonical Markdown containing the replacement item.
+            expected_revision: Optional revision returned by
+                :meth:`export_item_markdown`. A mismatch rejects stale edits.
+
+        Returns:
+            ``True`` when the item changed, otherwise ``False``.
+
+        Raises:
+            ValueError: If the Markdown or frontmatter is invalid.
+            MarkdownIdentityError: If the frontmatter changes the memory ID.
+            MarkdownResourceNotFoundError: If ``memory_id`` does not exist.
+            MarkdownRevisionConflictError: If ``expected_revision`` is stale.
+            RuntimeError: If the validated item cannot be persisted.
+        """
+        memory = self._markdown_to_memory_dict(document, source=f"memory {memory_id!r}")
+        document_id = memory["memory_id"]
+        if document_id != memory_id:
+            raise MarkdownIdentityError(
+                f"Frontmatter id {document_id!r} does not match resource id "
+                f"{memory_id!r}."
+            )
+        if not self.exists(memory_id):
+            raise MarkdownResourceNotFoundError(
+                f"AgentMemory item {memory_id!r} was not found."
+            )
+        if expected_revision is not None:
+            # _memory_lock is an RLock; this re-entrant call into
+            # export_item_markdown (also @_with_memory_lock) is intentional
+            # and safe because RLock allows the same thread to re-acquire.
+            current_revision = markdown_document_revision(
+                self.export_item_markdown(memory_id)
+            )
+            if current_revision != expected_revision:
+                raise MarkdownRevisionConflictError(current_revision)
+        if self._markdown_record_matches(memory_id, memory):
+            return False
+
+        success = self._replace_memory_item(
+            memory_id,
+            memory["content"],
+            metadata=memory["metadata"],
+            entities=memory["entities"],
+            relationships=memory["relationships"],
+            timestamp=memory["timestamp"],
+            skip_graph=True,
+        )
+        if not success:
+            raise RuntimeError(f"AgentMemory item {memory_id!r} could not be replaced.")
+        return True
+
    # Export/Import
    def export(
        self,
@@ -1679,6 +1793,7 @@ class AgentMemory:
            return self._export_markdown(memories, destination=destination)
        return export_data

+    @_with_memory_lock
    def import_data(
        self, data: Union[str, Path, Dict[str, Any]], format: str = "json"
    ) -> int:
@@ -131,6 +131,12 @@ from ..utils.progress_tracker import get_progress_tracker
 from ..utils.skos import is_skos_hierarchy_edge, validate_skos_hierarchy
 from ._markdown_filesystem import find_filesystem_link
 from .entity_linker import EntityLinker
+from .markdown import (
+    MarkdownIdentityError,
+    MarkdownResourceNotFoundError,
+    MarkdownRevisionConflictError,
+    markdown_document_revision,
+)


 class _UniqueKeySafeLoader(yaml.SafeLoader):
@@ -1167,6 +1173,157 @@ class ContextGraph:
                )
            )

+    def export_node_markdown(self, node_id: str) -> str:
+        """Return one existing node as canonical Markdown.
+
+        Args:
+            node_id: Stable identifier of the node to export.
+
+        Returns:
+            Canonical Markdown containing the node frontmatter and body.
+
+        Raises:
+            MarkdownResourceNotFoundError: If ``node_id`` does not exist.
+        """
+        with self._lock:
+            node = self.nodes.get(node_id)
+            if node is None:
+                raise MarkdownResourceNotFoundError(
+                    f"ContextGraph node {node_id!r} was not found."
+                )
+            return self._node_markdown_source(node)
+
+    def _node_markdown_source(self, node: ContextNode) -> str:
+        frontmatter = {
+            "id": node.node_id,
+            "type": node.node_type,
+            "properties": copy.deepcopy(node.properties),
+            "metadata": copy.deepcopy(node.metadata),
+            "valid_from": node.valid_from,
+            "valid_until": node.valid_until,
+        }
+        return self._render_markdown_document(
+            frontmatter, node.content, f"node {node.node_id!r}"
+        )
+
+    def apply_node_markdown(
+        self,
+        node_id: str,
+        document: str,
+        *,
+        expected_revision: Optional[str] = None,
+    ) -> bool:
+        """Validate and atomically replace one existing node.
+
+        Args:
+            node_id: Stable identifier of the node to update.
+            document: Canonical Markdown containing the replacement node.
+            expected_revision: Optional revision returned by
+                :meth:`export_node_markdown`. A mismatch rejects stale edits.
+
+        Returns:
+            ``True`` when the node changed, otherwise ``False``.
+
+        Raises:
+            ValueError: If the Markdown or frontmatter is invalid.
+            MarkdownIdentityError: If the frontmatter changes the node ID.
+            MarkdownResourceNotFoundError: If ``node_id`` does not exist.
+            MarkdownRevisionConflictError: If ``expected_revision`` is stale.
+        """
+        source = f"node {node_id!r}"
+        frontmatter, body = self._parse_markdown_document(document, source)
+        candidate = self._parse_markdown_node(frontmatter, body, source)
+        if candidate.node_id != node_id:
+            raise MarkdownIdentityError(
+                f"Frontmatter id {candidate.node_id!r} does not match resource id "
+                f"{node_id!r}."
+            )
+
+        with self._lock:
+            existing = self.nodes.get(node_id)
+            if existing is None:
+                raise MarkdownResourceNotFoundError(
+                    f"ContextGraph node {node_id!r} was not found."
+                )
+            if expected_revision is not None:
+                current_revision = markdown_document_revision(
+                    self._node_markdown_source(existing)
+                )
+                if current_revision != expected_revision:
+                    raise MarkdownRevisionConflictError(current_revision)
+            if existing == candidate:
+                return False
+
+            # Decision index rebuilding can still reject YAML-valid property
+            # shapes, so retain every affected structure until commit succeeds.
+            decision_state_before = {}
+            if (
+                existing.node_type.lower() == "decision"
+                or candidate.node_type.lower() == "decision"
+            ):
+                for attribute in (
+                    "_decisions",
+                    "_decision_index",
+                    "_entity_index",
+                    "_temporal_index",
+                ):
+                    if not hasattr(self, attribute):
+                        decision_state_before[attribute] = None
+                    elif attribute in {"_decision_index", "_entity_index"}:
+                        decision_state_before[attribute] = {
+                            key: set(values)
+                            for key, values in getattr(self, attribute).items()
+                        }
+                    elif attribute == "_temporal_index":
+                        decision_state_before[attribute] = list(
+                            getattr(self, attribute)
+                        )
+                    else:
+                        decision_state_before[attribute] = dict(
+                            getattr(self, attribute)
+                        )
+
+            old_type = existing.node_type
+            try:
+                old_bucket = self.node_type_index.get(old_type)
+                if old_bucket is not None:
+                    old_bucket.discard(node_id)
+                    if not old_bucket:
+                        del self.node_type_index[old_type]
+
+                self.nodes[node_id] = candidate
+                self.node_type_index[candidate.node_type].add(node_id)
+                if (
+                    old_type.lower() == "decision"
+                    or candidate.node_type.lower() == "decision"
+                ):
+                    self._sync_decision_from_node(node_id)
+                payload = candidate.to_dict()
+                self._analytics_cache.clear()
+            except Exception:
+                self.nodes[node_id] = existing
+                candidate_bucket = self.node_type_index.get(candidate.node_type)
+                if candidate_bucket is not None:
+                    candidate_bucket.discard(node_id)
+                    if not candidate_bucket:
+                        del self.node_type_index[candidate.node_type]
+                self.node_type_index[old_type].add(node_id)
+                for attribute, state in decision_state_before.items():
+                    if state is None:
+                        if hasattr(self, attribute):
+                            delattr(self, attribute)
+                    else:
+                        restored = (
+                            defaultdict(set, state)
+                            if attribute in {"_decision_index", "_entity_index"}
+                            else state
+                        )
+                        setattr(self, attribute, restored)
+                raise
+
+        self._emit_mutation("UPDATE_NODE", node_id, payload)
+        return True
+
    def save_to_file(
        self, path: Union[str, Path], format: str = "json"
    ) -> None:
@@ -4947,32 +5104,42 @@ class ContextGraph:
    def _sync_decision_from_node(self, node_id: str) -> None:
        """Synchronise a single decision index entry from the node store.

-        Called after ``add_node_attribute`` mutates a decision node so that
-        ``_decisions`` and the derived indexes stay consistent without
-        requiring a full rebuild of all decisions.
+        Called after ``add_node_attribute`` mutates a decision node and after
+        ``apply_node_markdown`` replaces a node whose old or new type is
+        ``"decision"``, so that ``_decisions`` and the derived indexes stay
+        consistent without requiring a full rebuild of all decisions.
+
+        Temporal index cleanup (``_temporal_index``) runs unconditionally
+        before the node-type guard so that stale entries are removed even when
+        transitioning a decision node to a non-decision type.  Callers are
+        expected to ensure this is only invoked when at least one of the
+        current or previous node types is ``"decision"``; callers that bypass
+        that invariant will have ``node_id`` silently removed from
+        ``_temporal_index`` even if it was never a decision node.
        """
        node = self.nodes.get(node_id)
-        if node is None:
-            return
-        if (getattr(node, "node_type", None) or "").lower() != "decision":
-            return
-
        if not hasattr(self, "_decisions"):
            # Indexes don't exist yet — a full rebuild is safer.
            self._rebuild_decision_indexes()
            return

-        # Remove stale index entries for this decision ID.
-        old = self._decisions.get(node_id)
+        # Remove stale index entries before deciding whether the current node
+        # still belongs in the decision indexes.
+        old = self._decisions.pop(node_id, None)
        if old:
            old_cat = old.get("category", "")
-            if old_cat and node_id in self._decision_index.get(old_cat, set()):
+            if old_cat:
                self._decision_index[old_cat].discard(node_id)
            for ent in old.get("entities", []):
                self._entity_index[ent].discard(node_id)
-            self._temporal_index = [
-                (nid, ts) for nid, ts in self._temporal_index if nid != node_id
-            ]
+        self._temporal_index = [
+            (nid, ts) for nid, ts in self._temporal_index if nid != node_id
+        ]
+
+        if node is None:
+            return
+        if (getattr(node, "node_type", None) or "").lower() != "decision":
+            return

        # Rebuild the entry for this node and re-insert index entries.
        meta: Dict[str, Any] = {}
@@ -0,0 +1,25 @@
+"""Shared revision helpers and errors for single-resource Markdown operations."""
+
+import hashlib
+
+
+class MarkdownResourceNotFoundError(KeyError):
+    """Raised when a Markdown operation targets a missing resource."""
+
+
+class MarkdownIdentityError(ValueError):
+    """Raised when frontmatter changes a resource's stable identity."""
+
+
+class MarkdownRevisionConflictError(ValueError):
+    """Raised when a resource changed after an edit session began."""
+
+    def __init__(self, current_revision: str) -> None:
+        super().__init__("Markdown resource revision does not match.")
+        self.current_revision = current_revision
+
+
+def markdown_document_revision(source: str) -> str:
+    """Return the stable revision token for a canonical Markdown document."""
+    digest = hashlib.sha256(source.encode("utf-8")).hexdigest()
+    return f"sha256:{digest}"
@@ -8,16 +8,19 @@ from contextlib import asynccontextmanager
 from pathlib import Path
 from typing import Optional

-from fastapi import Depends, FastAPI, HTTPException, Request, WebSocket, WebSocketDisconnect
+from fastapi import Depends, FastAPI, HTTPException, Request
 from fastapi.middleware.cors import CORSMiddleware
 from fastapi.responses import FileResponse, HTMLResponse, JSONResponse
 from fastapi.staticfiles import StaticFiles

 from .. import __version__
+from ..context.agent_memory import AgentMemory
 from ..context.context_graph import ContextGraph
-from .dependencies import anonymous_access_allowed, get_expected_api_key, is_valid_api_key, require_auth
+from .dependencies import anonymous_access_allowed, get_expected_api_key, require_auth
+from .markdown_resources import MarkdownResourceRegistry
+from .runtime import explorer_capabilities, install_mutation_bridge
 from .session import GraphSession
-from .ws import ConnectionManager
+from .ws import ConnectionManager, install_graph_updates_websocket


 def _read_int_env(name: str, default: int) -> int:
@@ -53,37 +56,23 @@ def _read_explorer_settings() -> dict:
    }


-def _install_mutation_bridge(app: FastAPI, session: GraphSession) -> None:
-    if getattr(session.graph, "_mutation_bridge_installed", False):
-        return
-    session.graph._mutation_bridge_installed = True
-    previous_callback = getattr(session.graph, "mutation_callback", None)
-
-    def on_mutation(event_type: str, entity_id: str, payload: dict) -> None:
-        session.handle_graph_mutation(event_type, entity_id, payload)
-        if callable(previous_callback):
-            previous_callback(event_type, entity_id, payload)
-        loop = getattr(app.state, "event_loop", None)
-        manager = getattr(app.state, "ws_manager", None)
-        if loop is None or manager is None or loop.is_closed():
-            return
-        message = {
-            "event_type": event_type,
-            "entity_id": entity_id,
-            "payload": payload,
-        }
-        asyncio.run_coroutine_threadsafe(
-            manager.broadcast("graph_mutation", message),
-            loop,
-        )
-
-    session.graph.mutation_callback = on_mutation
-
-
 def create_app(
    session: Optional[GraphSession] = None,
    provenance_storage_path: Optional[str] = None,
+    agent_memory: Optional[AgentMemory] = None,
 ) -> FastAPI:
+    """Create an Explorer application over live graph and memory objects.
+
+    Args:
+        session: Graph session exposed by the Explorer. A new in-memory graph
+            session is created when omitted.
+        provenance_storage_path: Optional per-app provenance database path.
+        agent_memory: Existing AgentMemory instance to expose in the Memories
+            workspace. The workspace is unavailable when omitted.
+
+    Returns:
+        Configured FastAPI application.
+    """
    settings = _read_explorer_settings()
    prov_path = provenance_storage_path or settings.get("provenance_storage_path")
    if session is None:
@@ -95,6 +84,7 @@ def create_app(
        active_session = session
        if prov_path is not None:
            active_session.set_provenance_storage_path(prov_path)
+    markdown_resources = MarkdownResourceRegistry(active_session.graph, agent_memory)

    @asynccontextmanager
    async def lifespan(app: FastAPI):
@@ -117,7 +107,9 @@ def create_app(
        app.state.event_loop = asyncio.get_running_loop()
        app.state.ws_manager = ConnectionManager()
        app.state.session = active_session
-        _install_mutation_bridge(app, active_session)
+        app.state.agent_memory = agent_memory
+        app.state.markdown_resources = markdown_resources
+        install_mutation_bridge(app, active_session)
        yield

    app = FastAPI(
@@ -139,7 +131,7 @@ def create_app(
        CORSMiddleware,
        allow_origins=settings["allowed_origins"],
        allow_credentials=_allow_credentials,
-        allow_methods=["GET", "POST", "DELETE", "OPTIONS"],
+        allow_methods=["GET", "POST", "PUT", "DELETE", "OPTIONS"],
        allow_headers=["Content-Type", "Authorization", "X-API-Key"],
        max_age=600,
    )
@@ -170,6 +162,8 @@ def create_app(
    from .routes.enrich import router as enrich_router
    from .routes.export_import import router as export_import_router
    from .routes.graph import router as graph_router
+    from .routes.markdown import router as markdown_router
+    from .routes.memories import router as memories_router
    from .routes.ontology import router as ontology_router
    from .routes.provenance import router as provenance_router
    from .routes.sparql import router as sparql_router
@@ -183,54 +177,15 @@ def create_app(
    app.include_router(temporal_router, dependencies=_auth)
    app.include_router(enrich_router, dependencies=_auth)
    app.include_router(export_import_router, dependencies=_auth)
+    app.include_router(markdown_router, dependencies=_auth)
+    app.include_router(memories_router, dependencies=_auth)
    app.include_router(annotations_router, dependencies=_auth)
    app.include_router(sparql_router, dependencies=_auth)
    app.include_router(provenance_router, dependencies=_auth)
    app.include_router(vocabulary_router, dependencies=_auth)
    app.include_router(ontology_router, dependencies=_auth)

-    _WS_MAX_MESSAGE_BYTES = 64 * 1024  # 64 KB — control messages only
-
-    @app.websocket("/ws/graph-updates")
-    async def websocket_endpoint(websocket: WebSocket):
-        # CORSMiddleware doesn't cover WebSocket handshakes (Starlette's
-        # CORS support only wraps HTTP), so under SEMANTICA_ALLOW_ANONYMOUS
-        # the key check below accepts any origin — loopback binding isn't a
-        # boundary against a browser, since any page the operator has open
-        # can still reach ws://localhost:.../ws/graph-updates directly.
-        # Reject a foreign Origin explicitly here, against the same
-        # allowlist CORSMiddleware already enforces for HTTP
-        # (GHSA-4643-wpgq-w329). Browsers always send Origin on a
-        # cross-origin WebSocket handshake; native/CLI clients omit it
-        # entirely, so a missing Origin is allowed through — the browser is
-        # the only threat this check is closing.
-        origin = websocket.headers.get("origin")
-        allowed_origins = app.state.explorer_settings["allowed_origins"]
-        if origin is not None and origin not in allowed_origins:
-            await websocket.close(code=4403)  # forbidden
-            return
-
-        # Browsers can't set custom headers on a WebSocket handshake, so
-        # accept the key via header (non-browser clients) or query param
-        # (browser clients), same SEMANTICA_API_KEY the REST routes check.
-        candidate = websocket.headers.get("x-api-key") or websocket.query_params.get("api_key")
-        if not is_valid_api_key(candidate):
-            await websocket.close(code=4401)  # unauthorized
-            return
-
-        manager: ConnectionManager = app.state.ws_manager
-        await manager.connect(websocket)
-        await manager.send_personal(websocket, "connection_ack", {"connected": True})
-        try:
-            while True:
-                message = await websocket.receive_text()
-                if len(message) > _WS_MAX_MESSAGE_BYTES:
-                    await websocket.close(code=1009)  # 1009 = message too big
-                    break
-                if message.strip().lower() == "ping":
-                    await manager.send_personal(websocket, "pong", {"ok": True})
-        except WebSocketDisconnect:
-            manager.disconnect(websocket)
+    install_graph_updates_websocket(app, settings["allowed_origins"])

    @app.get("/", include_in_schema=False)
    async def root():
@@ -269,6 +224,7 @@ def create_app(
            "name": "Semantica Knowledge Explorer",
            "version": __version__,
            "status": "active",
+            "capabilities": explorer_capabilities(agent_memory),
        }

    static_dir = Path(__file__).resolve().parent.parent / "static"
@@ -14,6 +14,8 @@ from typing import Optional
 from fastapi import Request, HTTPException, Security, status
 from fastapi.security.api_key import APIKeyHeader

+from ..context.agent_memory import AgentMemory
+from .markdown_resources import MarkdownResourceRegistry
 from .session import GraphSession

 _api_key_header = APIKeyHeader(name="X-API-Key", auto_error=False)
@@ -80,3 +82,25 @@ def get_session(request: Request) -> GraphSession:
            detail="GraphSession not initialized."
        )
    return request.app.state.session
+
+
+def get_markdown_resources(request: Request) -> MarkdownResourceRegistry:
+    """Retrieve the Markdown resource registry stored on ``app.state``."""
+    resources = getattr(request.app.state, "markdown_resources", None)
+    if resources is None:
+        raise HTTPException(
+            status_code=status.HTTP_503_SERVICE_UNAVAILABLE,
+            detail="Markdown resources are not initialized.",
+        )
+    return resources
+
+
+def get_agent_memory(request: Request) -> AgentMemory:
+    """Retrieve the optional AgentMemory configured for Explorer."""
+    memory = getattr(request.app.state, "agent_memory", None)
+    if memory is None:
+        raise HTTPException(
+            status_code=status.HTTP_503_SERVICE_UNAVAILABLE,
+            detail="AgentMemory is not configured for this Explorer instance.",
+        )
+    return memory
@@ -0,0 +1,283 @@
+"""Single-resource Markdown access for Explorer."""
+
+from __future__ import annotations
+
+from dataclasses import dataclass
+from enum import Enum
+from typing import Callable, Dict, Optional, Protocol, TypeVar
+
+import yaml
+
+from ..context.agent_memory import AgentMemory
+from ..context.context_graph import ContextGraph
+from ..context.markdown import (
+    MarkdownIdentityError,
+    MarkdownResourceNotFoundError,
+    MarkdownRevisionConflictError,
+    markdown_document_revision,
+)
+
+
+class MarkdownResourceKind(str, Enum):
+    CONTEXT_NODE = "context-node"
+    AGENT_MEMORY = "agent-memory"
+
+
+@dataclass(frozen=True)
+class MarkdownResourceRef:
+    kind: MarkdownResourceKind
+    id: str
+
+
+@dataclass(frozen=True)
+class MarkdownDocument:
+    resource: MarkdownResourceRef
+    source: str
+    body: str
+    revision: str
+
+
+@dataclass(frozen=True)
+class MarkdownApplyResult(MarkdownDocument):
+    changed: bool
+
+
+class MarkdownResourceError(Exception):
+    """Base class for safe, structured Explorer Markdown failures."""
+
+    code = "markdown_resource_error"
+
+    def __init__(
+        self,
+        message: str,
+        *,
+        field: Optional[str] = None,
+        current_revision: Optional[str] = None,
+    ) -> None:
+        super().__init__(message)
+        self.message = message
+        self.field = field
+        self.current_revision = current_revision
+
+
+class MarkdownResourceNotFound(MarkdownResourceError):
+    code = "markdown_resource_not_found"
+
+
+class InvalidMarkdownFrontmatter(MarkdownResourceError):
+    code = "invalid_markdown_frontmatter"
+
+
+class ResourceIdentityMismatch(MarkdownResourceError):
+    code = "resource_identity_mismatch"
+
+
+class MarkdownRevisionConflict(MarkdownResourceError):
+    code = "markdown_revision_conflict"
+
+
+class MarkdownSaveFailed(MarkdownResourceError):
+    code = "markdown_save_failed"
+
+
+class MarkdownAdapter(Protocol):
+    def export(self, resource_id: str) -> str:
+        ...
+
+    def apply(self, resource_id: str, source: str, expected_revision: str) -> bool:
+        ...
+
+
+_Result = TypeVar("_Result")
+
+
+def _translate_domain_errors(operation: Callable[[], _Result]) -> _Result:
+    try:
+        return operation()
+    except MarkdownResourceNotFoundError as exc:
+        raise MarkdownResourceNotFound(str(exc.args[0])) from exc
+    except MarkdownRevisionConflictError as exc:
+        raise MarkdownRevisionConflict(
+            "This item changed after editing began. Reload the latest "
+            "version before applying.",
+            current_revision=exc.current_revision,
+        ) from exc
+    except MarkdownIdentityError as exc:
+        raise ResourceIdentityMismatch(str(exc), field="id") from exc
+    except ValueError as exc:
+        raise InvalidMarkdownFrontmatter(_safe_validation_message(exc)) from exc
+
+
+class ContextGraphNodeMarkdownAdapter:
+    def __init__(self, graph: ContextGraph) -> None:
+        self._graph = graph
+
+    def export(self, resource_id: str) -> str:
+        return _translate_domain_errors(
+            lambda: self._graph.export_node_markdown(resource_id)
+        )
+
+    def apply(self, resource_id: str, source: str, expected_revision: str) -> bool:
+        return _translate_domain_errors(
+            lambda: self._graph.apply_node_markdown(
+                resource_id,
+                source,
+                expected_revision=expected_revision,
+            )
+        )
+
+
+class AgentMemoryItemMarkdownAdapter:
+    def __init__(self, memory: AgentMemory) -> None:
+        self._memory = memory
+
+    def export(self, resource_id: str) -> str:
+        return _translate_domain_errors(
+            lambda: self._memory.export_item_markdown(resource_id)
+        )
+
+    def apply(self, resource_id: str, source: str, expected_revision: str) -> bool:
+        return _translate_domain_errors(
+            lambda: self._memory.apply_item_markdown(
+                resource_id,
+                source,
+                expected_revision=expected_revision,
+            )
+        )
+
+
+def _safe_validation_message(exc: ValueError) -> str:
+    if isinstance(exc.__cause__, yaml.YAMLError):
+        return "Markdown frontmatter contains invalid YAML."
+    return str(exc)
+
+
+def document_revision(source: str) -> str:
+    """Return the stable revision token for canonical Markdown.
+
+    Args:
+        source: Canonical Markdown source.
+
+    Returns:
+        A SHA-256 revision token suitable for optimistic concurrency checks.
+    """
+    return markdown_document_revision(source)
+
+
+def markdown_body(source: str) -> str:
+    """Extract the body from canonical Markdown emitted by a domain model."""
+    lines = source.splitlines(keepends=True)
+    if not lines or lines[0].rstrip("\r\n") != "---":
+        raise MarkdownSaveFailed("The resource produced invalid canonical Markdown.")
+    closing_index = next(
+        (
+            index
+            for index, line in enumerate(lines[1:], start=1)
+            if line.rstrip("\r\n") == "---"
+        ),
+        None,
+    )
+    if closing_index is None:
+        raise MarkdownSaveFailed("The resource produced invalid canonical Markdown.")
+    body = "".join(lines[closing_index + 1 :])
+    if body.startswith("\r\n"):
+        return body[2:]
+    if body.startswith("\n"):
+        return body[1:]
+    return body
+
+
+class MarkdownResourceRegistry:
+    """Route Markdown operations to their owning domain models."""
+
+    def __init__(
+        self,
+        context_graph: ContextGraph,
+        agent_memory: Optional[AgentMemory] = None,
+    ) -> None:
+        self._adapters: Dict[MarkdownResourceKind, MarkdownAdapter] = {
+            MarkdownResourceKind.CONTEXT_NODE: ContextGraphNodeMarkdownAdapter(
+                context_graph
+            )
+        }
+        if agent_memory is not None:
+            self._adapters[
+                MarkdownResourceKind.AGENT_MEMORY
+            ] = AgentMemoryItemMarkdownAdapter(agent_memory)
+
+    def _adapter(self, kind: MarkdownResourceKind) -> MarkdownAdapter:
+        adapter = self._adapters.get(kind)
+        if adapter is None:
+            raise MarkdownResourceNotFound(
+                f"Markdown resource kind {kind.value!r} is not available."
+            )
+        return adapter
+
+    def read(self, ref: MarkdownResourceRef) -> MarkdownDocument:
+        """Read one resource as canonical Markdown.
+
+        Args:
+            ref: Explicit resource kind and stable identifier.
+
+        Returns:
+            The canonical source, body, and current revision.
+
+        Raises:
+            MarkdownResourceError: If the resource is unavailable or invalid.
+        """
+        try:
+            source = self._adapter(ref.kind).export(ref.id)
+        except MarkdownResourceError:
+            raise
+        except Exception as exc:
+            raise MarkdownSaveFailed(
+                "The Markdown resource could not be read."
+            ) from exc
+        return MarkdownDocument(
+            resource=ref,
+            source=source,
+            body=markdown_body(source),
+            revision=document_revision(source),
+        )
+
+    def apply(
+        self,
+        ref: MarkdownResourceRef,
+        markdown: str,
+        expected_revision: str,
+    ) -> MarkdownApplyResult:
+        """Apply validated Markdown to one existing resource.
+
+        Args:
+            ref: Explicit resource kind and stable identifier.
+            markdown: Replacement canonical Markdown.
+            expected_revision: Revision observed when editing began.
+
+        Returns:
+            The saved canonical document and whether it changed.
+
+        Raises:
+            MarkdownResourceError: If validation, identity, persistence, or
+                revision checks fail.
+        """
+        try:
+            changed = self._adapter(ref.kind).apply(
+                ref.id,
+                markdown,
+                expected_revision,
+            )
+        except MarkdownResourceError:
+            raise
+        except Exception as exc:
+            raise MarkdownSaveFailed(
+                "The edit could not be applied. The existing item was not changed."
+            ) from exc
+
+        saved = self.read(ref)
+        return MarkdownApplyResult(
+            resource=saved.resource,
+            source=saved.source,
+            body=saved.body,
+            revision=saved.revision,
+            changed=changed,
+        )
@@ -87,6 +87,13 @@ def _node_response(node: dict) -> NodeResponse:
    return NodeResponse(**node)


+async def _get_node_or_404(node_id: str, session: GraphSession) -> NodeResponse:
+    node = await asyncio.to_thread(session.get_node, node_id)
+    if node is None:
+        raise HTTPException(status_code=404, detail=f"Node '{node_id}' not found")
+    return _node_response(node)
+
+
 def _edge_response(edge: dict) -> EdgeResponse:
    return EdgeResponse(**edge)

@@ -121,15 +128,20 @@ async def list_nodes(
    )


+@router.get("/node", response_model=NodeResponse)
+async def get_node_by_query(
+    node_id: str = Query(..., description="Exact node ID"),
+    session: GraphSession = Depends(get_session),
+):
+    return await _get_node_or_404(node_id, session)
+
+
@router.get("/node/{node_id}", response_model=NodeResponse)
 async def get_node(
    node_id: str,
    session: GraphSession = Depends(get_session),
 ):
-    node = await asyncio.to_thread(session.get_node, node_id)
-    if node is None:
-        raise HTTPException(status_code=404, detail=f"Node '{node_id}' not found")
-    return _node_response(node)
+    return await _get_node_or_404(node_id, session)


@router.get("/node/{node_id}/neighbors", response_model=list[NeighborResponse])
@@ -441,7 +453,7 @@ async def distance_matrix(
                    embeddings = _get_cached_embeddings(session)
                    src_embedding = embeddings.get(src)
                    tgt_embedding = embeddings.get(tgt)
-                    
+
                    if src_embedding is None or tgt_embedding is None:
                        val = None
                    else:
@@ -451,7 +463,7 @@ async def distance_matrix(
                        tgt_vec = np.array(tgt_embedding)
                        sim = np.dot(src_vec, tgt_vec) / (np.linalg.norm(src_vec) * np.linalg.norm(tgt_vec))
                        val = 1.0 - float(sim) if isinstance(sim, (int, float)) else None
-                    
+
                    matrix[i][j] = val
                    matrix[j][i] = val
                elif path_finder is not None:
@@ -0,0 +1,115 @@
+"""Explorer routes for canonical Markdown resources."""
+
+from typing import NoReturn
+
+from fastapi import APIRouter, Depends, HTTPException, status
+
+from ..dependencies import get_markdown_resources
+from ..markdown_resources import (
+    InvalidMarkdownFrontmatter,
+    MarkdownApplyResult,
+    MarkdownDocument,
+    MarkdownResourceError,
+    MarkdownResourceKind,
+    MarkdownResourceNotFound,
+    MarkdownResourceRef,
+    MarkdownResourceRegistry,
+    MarkdownRevisionConflict,
+    ResourceIdentityMismatch,
+)
+from ..schemas import (
+    MarkdownApplyRequest,
+    MarkdownApplyResponse,
+    MarkdownDocumentResponse,
+)
+
+router = APIRouter(prefix="/api/markdown", tags=["markdown"])
+
+
+def _resource_ref(kind: str, resource_id: str) -> MarkdownResourceRef:
+    try:
+        resource_kind = MarkdownResourceKind(kind)
+    except ValueError:
+        _raise_http_error(
+            MarkdownResourceNotFound(
+                f"Markdown resource kind {kind!r} is not available."
+            )
+        )
+    return MarkdownResourceRef(kind=resource_kind, id=resource_id)
+
+
+def _error_detail(error: MarkdownResourceError) -> dict:
+    detail = {"code": error.code, "message": error.message}
+    if error.field is not None:
+        detail["field"] = error.field
+    if error.current_revision is not None:
+        detail["current_revision"] = error.current_revision
+    return detail
+
+
+def _raise_http_error(error: MarkdownResourceError) -> NoReturn:
+    if isinstance(error, MarkdownResourceNotFound):
+        status_code = status.HTTP_404_NOT_FOUND
+    elif isinstance(error, MarkdownRevisionConflict):
+        status_code = status.HTTP_409_CONFLICT
+    elif isinstance(error, (InvalidMarkdownFrontmatter, ResourceIdentityMismatch)):
+        status_code = status.HTTP_422_UNPROCESSABLE_ENTITY
+    else:
+        status_code = status.HTTP_500_INTERNAL_SERVER_ERROR
+    raise HTTPException(status_code=status_code, detail=_error_detail(error)) from error
+
+
+def _document_response(
+    document: MarkdownDocument,
+) -> MarkdownDocumentResponse:
+    return MarkdownDocumentResponse(
+        resource={
+            "kind": document.resource.kind.value,
+            "id": document.resource.id,
+        },
+        source=document.source,
+        body=document.body,
+        revision=document.revision,
+        editable=True,
+    )
+
+
+def _apply_response(result: MarkdownApplyResult) -> MarkdownApplyResponse:
+    document = _document_response(result)
+    return MarkdownApplyResponse(**document.model_dump(), changed=result.changed)
+
+
+@router.get(
+    "/{kind}/{resource_id:path}",
+    response_model=MarkdownDocumentResponse,
+)
+def read_markdown_resource(
+    kind: str,
+    resource_id: str,
+    resources: MarkdownResourceRegistry = Depends(get_markdown_resources),
+) -> MarkdownDocumentResponse:
+    try:
+        return _document_response(resources.read(_resource_ref(kind, resource_id)))
+    except MarkdownResourceError as error:
+        _raise_http_error(error)
+
+
+@router.put(
+    "/{kind}/{resource_id:path}",
+    response_model=MarkdownApplyResponse,
+)
+def apply_markdown_resource(
+    kind: str,
+    resource_id: str,
+    request: MarkdownApplyRequest,
+    resources: MarkdownResourceRegistry = Depends(get_markdown_resources),
+) -> MarkdownApplyResponse:
+    try:
+        result = resources.apply(
+            _resource_ref(kind, resource_id),
+            request.markdown,
+            request.expected_revision,
+        )
+        return _apply_response(result)
+    except MarkdownResourceError as error:
+        _raise_http_error(error)
@@ -0,0 +1,41 @@
+"""Minimal AgentMemory selection surface for Explorer."""
+
+from fastapi import APIRouter, Depends, Query
+
+from ...context.agent_memory import AgentMemory
+from ..dependencies import get_agent_memory
+from ..schemas import MemoryListResponse, MemorySummaryResponse
+
+router = APIRouter(prefix="/api/memories", tags=["memories"])
+
+
+@router.get("", response_model=MemoryListResponse)
+def list_memories(
+    skip: int = Query(default=0, ge=0),
+    limit: int = Query(default=100, ge=1, le=100),
+    memory: AgentMemory = Depends(get_agent_memory),
+) -> MemoryListResponse:
+    records, total = memory.list_snapshot(offset=skip, limit=limit)
+    items = []
+    for record in records:
+        metadata = record.get("metadata") or {}
+        content = record.get("content") or ""
+        excerpt = " ".join(str(content).split())[:160]
+        items.append(
+            MemorySummaryResponse(
+                id=record["memory_id"],
+                type=str(metadata.get("type") or "general"),
+                excerpt=excerpt,
+                updated_at=(
+                    str(metadata["updated_at"])
+                    if metadata.get("updated_at") is not None
+                    else None
+                ),
+            )
+        )
+    return MemoryListResponse(
+        items=items,
+        total=total,
+        skip=skip,
+        limit=limit,
+    )
@@ -0,0 +1,45 @@
+"""Shared runtime assembly for Explorer entry points."""
+
+import asyncio
+from typing import Dict, Optional
+
+from fastapi import FastAPI
+
+from ..context.agent_memory import AgentMemory
+from .session import GraphSession
+
+
+def explorer_capabilities(agent_memory: Optional[AgentMemory]) -> Dict[str, bool]:
+    """Describe optional Explorer features exposed by the current host."""
+    return {"agent_memory": agent_memory is not None}
+
+
+def install_mutation_bridge(app: FastAPI, session: GraphSession) -> None:
+    """Keep Explorer indexes and WebSocket clients in sync with graph writes."""
+    if getattr(app.state, "_semantica_mutation_bridge_session", None) is session:
+        return
+    previous_callback = getattr(session.graph, "mutation_callback", None)
+
+    def on_mutation(event_type: str, entity_id: str, payload: dict) -> None:
+        session.handle_graph_mutation(event_type, entity_id, payload)
+        if callable(previous_callback):
+            previous_callback(event_type, entity_id, payload)
+        loop = getattr(app.state, "event_loop", None)
+        manager = getattr(app.state, "ws_manager", None)
+        if loop is None or manager is None or loop.is_closed():
+            return
+        asyncio.run_coroutine_threadsafe(
+            manager.broadcast(
+                "graph_mutation",
+                {
+                    "event_type": event_type,
+                    "entity_id": entity_id,
+                    "payload": payload,
+                },
+            ),
+            loop,
+        )
+
+    session.graph.mutation_callback = on_mutation
+    app.state._semantica_mutation_bridge = on_mutation
+    app.state._semantica_mutation_bridge_session = session
@@ -428,3 +428,39 @@ class DistanceExportRequest(BaseModel):
    include: List[str] = Field(
        default_factory=lambda: ["source_id", "target_id", "hop_count", "distance_band"],
    )
+
+
+class MarkdownResourceRefResponse(BaseModel):
+    kind: Literal["context-node", "agent-memory"]
+    id: str
+
+
+class MarkdownDocumentResponse(BaseModel):
+    resource: MarkdownResourceRefResponse
+    source: str
+    body: str
+    revision: str
+    editable: bool = True
+
+
+class MarkdownApplyRequest(BaseModel):
+    markdown: str
+    expected_revision: str = Field(..., pattern=r"^sha256:[0-9a-f]{64}$")
+
+
+class MarkdownApplyResponse(MarkdownDocumentResponse):
+    changed: bool
+
+
+class MemorySummaryResponse(BaseModel):
+    id: str
+    type: str
+    excerpt: str
+    updated_at: Optional[str] = None
+
+
+class MemoryListResponse(BaseModel):
+    items: List[MemorySummaryResponse]
+    total: int
+    skip: int
+    limit: int
@@ -9,9 +9,53 @@ import asyncio
 import json
 import threading
 from datetime import datetime, timezone
-from typing import Any, Dict, Set
+from typing import Any, Dict, List, Sequence, Set

-from fastapi import WebSocket
+from fastapi import FastAPI, WebSocket, WebSocketDisconnect
+
+from .dependencies import is_valid_api_key
+
+_WS_MAX_MESSAGE_BYTES = 64 * 1024
+
+
+def install_graph_updates_websocket(
+    app: FastAPI,
+    allowed_origins: Sequence[str],
+) -> None:
+    """Install the authenticated graph-mutation WebSocket endpoint."""
+    allowed_origin_set = frozenset(allowed_origins)
+
+    @app.websocket("/ws/graph-updates")
+    async def websocket_endpoint(websocket: WebSocket) -> None:
+        # CORSMiddleware does not cover WebSocket handshakes. Reject foreign
+        # browser origins against the same allowlist used for HTTP CORS.
+        origin = websocket.headers.get("origin")
+        if origin is not None and origin not in allowed_origin_set:
+            await websocket.close(code=4403)
+            return
+
+        # Browser clients pass the API key as a query parameter because the
+        # WebSocket API cannot set custom headers.
+        candidate = websocket.headers.get("x-api-key") or websocket.query_params.get(
+            "api_key"
+        )
+        if not is_valid_api_key(candidate):
+            await websocket.close(code=4401)
+            return
+
+        manager: ConnectionManager = app.state.ws_manager
+        await manager.connect(websocket)
+        await manager.send_personal(websocket, "connection_ack", {"connected": True})
+        try:
+            while True:
+                message = await websocket.receive_text()
+                if len(message) > _WS_MAX_MESSAGE_BYTES:
+                    await websocket.close(code=1009)
+                    break
+                if message.strip().lower() == "ping":
+                    await manager.send_personal(websocket, "pong", {"ok": True})
+        except WebSocketDisconnect:
+            manager.disconnect(websocket)


 class ConnectionManager:
@@ -63,8 +107,7 @@ class ConnectionManager:
        with self._lock:
            connections = set(self._active_connections)

-    
-        disconnected: list[WebSocket] = []
+        disconnected: List[WebSocket] = []
        for ws in connections:
            try:
                await ws.send_text(message)
@@ -301,7 +301,10 @@ class GraphValidator:
                    code="ORPHAN_NODES",
                    message=f"Found {len(isolates)} orphan nodes (no relationships).",
                    severity=ValidationSeverity.WARNING,
-                    details={"count": len(isolates), "ids": isolates[:10]} # Limit output
+                    details={
+                        "count": len(isolates),
+                        "ids": sorted(isolates, key=str)[:10],
+                    }  # Limit output deterministically
                ))

        except Exception as e:
@@ -35,7 +35,7 @@ Example Usage:
    >>> llm = LiteLLM(model="openai/gpt-4o", api_key="your-key")
    >>> response = llm.generate("Hello, world!")
    >>> # Or use other providers via LiteLLM
-    >>> llm = LiteLLM(model="anthropic/claude-sonnet-4-20250514")
+    >>> llm = LiteLLM(model="anthropic/claude-sonnet-5")
    >>> response = llm.generate("Hello, world!")
    >>>
    >>> # Anthropic provider
@@ -31,7 +31,7 @@ class LiteLLM:
    Provides unified interface to 100+ LLM providers through LiteLLM library.
    Supports providers like OpenAI, Anthropic, Groq, Azure, Bedrock, Vertex AI, etc.
    
-    Model format: "provider/model-name" (e.g., "openai/gpt-4o", "anthropic/claude-sonnet-4-20250514", "groq/llama-3.1-8b-instant")
+    Model format: "provider/model-name" (e.g., "openai/gpt-4o", "anthropic/claude-sonnet-5", "groq/llama-3.1-8b-instant")
    
    Example:
        >>> from semantica.llms import LiteLLM
@@ -39,7 +39,7 @@ class LiteLLM:
        >>> response = llm.generate("What is AI?")
        >>> 
        >>> # Use with different providers
-        >>> llm = LiteLLM(model="anthropic/claude-sonnet-4-20250514")
+        >>> llm = LiteLLM(model="anthropic/claude-sonnet-5")
        >>> response = llm.generate("Hello!")
    """

@@ -54,7 +54,7 @@ class LiteLLM:

        Args:
            model: Model identifier in format "provider/model-name"
-                   Examples: "openai/gpt-4o", "anthropic/claude-sonnet-4-20250514",
+                   Examples: "openai/gpt-4o", "anthropic/claude-sonnet-5",
                             "groq/llama-3.1-8b-instant", "azure/gpt-4", etc.
            api_key: API key (optional, can use environment variables)
            **kwargs: Additional LiteLLM options (temperature, max_tokens, etc.)
@@ -162,6 +162,13 @@ from .ontology_validator import (
    run_shacl_validation,
    validate_ontology,
 )
+from .quality_gate import (
+    OntologyQualityGate,
+    OntologyQualityReport,
+    QualityIssue,
+    QualitySeverity,
+    ontology_quality_check,
+)
 from .owl_generator import OWLGenerator
 from .property_generator import PropertyGenerator
 from .registry import MethodRegistry, method_registry
@@ -194,6 +201,11 @@ __all__ = [
    "SHACLValidationReport",
    "SHACLViolation",
    "run_shacl_validation",
+    "OntologyQualityGate",
+    "OntologyQualityReport",
+    "QualityIssue",
+    "QualitySeverity",
+    "ontology_quality_check",
    # OWL/RDF generation
    "OWLGenerator",
    # Requirements and competency questions
@@ -9,6 +9,7 @@ from .property_generator import PropertyGenerator
 from .owl_generator import OWLGenerator
 from .ontology_evaluator import OntologyEvaluator
 from .ontology_validator import OntologyValidator
+from .quality_gate import OntologyQualityGate, OntologyQualityReport
 from .llm_generator import LLMOntologyGenerator
 from ..semantic_extract.triplet_extractor import Triplet

@@ -25,6 +26,9 @@ class OntologyEngine:
        self.owl = OWLGenerator(**config)
        self.evaluator = OntologyEvaluator(**config)
        self.validator = OntologyValidator(**config)
+        self.quality_gate = OntologyQualityGate(
+            validator=self.validator, evaluator=self.evaluator
+        )
        self.llm = LLMOntologyGenerator(**config)
        self.store = config.get("store")

@@ -593,6 +597,15 @@ class OntologyEngine:
    def validate(self, ontology: Dict[str, Any], **options):
        return self.validator.validate(ontology, **options)

+    def quality_check(
+        self,
+        ontology: Dict[str, Any],
+        graph_data: Optional[Dict[str, Any]] = None,
+        **options,
+    ) -> OntologyQualityReport:
+        """Run deterministic ontology quality checks suitable for CI."""
+        return self.quality_gate.check(ontology, graph_data=graph_data, **options)
+
    def to_owl(self, ontology: Dict[str, Any], format: str = "turtle", **options):
        return self.owl.generate_owl(ontology, format=format, **options)

@@ -0,0 +1,855 @@
+"""Deterministic quality checks for ontology and KG pipelines."""
+
+import copy
+import math
+from dataclasses import dataclass, field
+from enum import Enum
+from typing import Any, Dict, Iterable, List, Mapping, Optional, Set, Tuple
+
+from ..kg.graph_validator import GraphValidator
+from .ontology_evaluator import OntologyEvaluator
+from .ontology_validator import OntologyValidator
+
+
+class QualitySeverity(str, Enum):
+    """Severity assigned to a quality finding."""
+
+    INFO = "info"
+    WARNING = "warning"
+    ERROR = "error"
+    CRITICAL = "critical"
+
+
+@dataclass
+class QualityIssue:
+    """A single machine-readable ontology quality finding."""
+
+    code: str
+    message: str
+    severity: QualitySeverity
+    element_id: Optional[str] = None
+    element_type: Optional[str] = None
+    details: Dict[str, Any] = field(default_factory=dict)
+
+    def to_dict(self) -> Dict[str, Any]:
+        """Return a JSON-friendly representation of the issue."""
+        return {
+            "code": self.code,
+            "message": self.message,
+            "severity": self.severity.value,
+            "element_id": self.element_id,
+            "element_type": self.element_type,
+            "details": self.details,
+        }
+
+
+@dataclass
+class OntologyQualityReport:
+    """Result returned by :class:`OntologyQualityGate`."""
+
+    passed: bool
+    issues: List[QualityIssue] = field(default_factory=list)
+    stats: Dict[str, int] = field(default_factory=dict)
+    metrics: Dict[str, float] = field(default_factory=dict)
+    thresholds: Dict[str, Optional[float]] = field(default_factory=dict)
+    threshold_failures: List[str] = field(default_factory=list)
+
+    @property
+    def error_count(self) -> int:
+        """Number of error and critical findings."""
+        return sum(
+            issue.severity in (QualitySeverity.ERROR, QualitySeverity.CRITICAL)
+            for issue in self.issues
+        )
+
+    @property
+    def warning_count(self) -> int:
+        """Number of warning findings."""
+        return sum(issue.severity == QualitySeverity.WARNING for issue in self.issues)
+
+    @property
+    def info_count(self) -> int:
+        """Number of informational findings."""
+        return sum(issue.severity == QualitySeverity.INFO for issue in self.issues)
+
+    def to_dict(self) -> Dict[str, Any]:
+        """Return a JSON-friendly representation of the report."""
+        return {
+            "passed": self.passed,
+            "issues": [issue.to_dict() for issue in self.issues],
+            "stats": {
+                **self.stats,
+                "issues": len(self.issues),
+                "errors": self.error_count,
+                "warnings": self.warning_count,
+                "infos": self.info_count,
+            },
+            "metrics": self.metrics,
+            "thresholds": self.thresholds,
+            "threshold_failures": self.threshold_failures,
+        }
+
+
+class OntologyQualityGate:
+    """Run deterministic, CI-friendly ontology quality checks."""
+
+    DEFAULT_THRESHOLDS: Dict[str, Optional[float]] = {
+        "min_coverage": 0.0,
+        "max_errors": 0.0,
+        "max_warnings": None,
+    }
+    _DATA_PROPERTY_TYPES = {"data", "datatype", "data_property", "literal"}
+    _OBJECT_PROPERTY_TYPES = {"object", "object_property", "relationship"}
+    _KNOWN_DATATYPES = {
+        "string",
+        "boolean",
+        "decimal",
+        "float",
+        "double",
+        "integer",
+        "int",
+        "long",
+        "short",
+        "byte",
+        "date",
+        "datetime",
+        "datetimestamp",
+        "time",
+        "duration",
+        "anyuri",
+    }
+    _BUILTIN_CLASSES = {
+        "owl:thing",
+        "rdfs:resource",
+        "http://www.w3.org/2002/07/owl#thing",
+        "http://www.w3.org/2000/01/rdf-schema#resource",
+    }
+
+    def __init__(
+        self,
+        validator: Optional[OntologyValidator] = None,
+        evaluator: Optional[OntologyEvaluator] = None,
+        thresholds: Optional[Mapping[str, Optional[float]]] = None,
+        fail_on_warnings: bool = False,
+    ) -> None:
+        self.validator = validator or OntologyValidator()
+        self.evaluator = evaluator or OntologyEvaluator()
+        self.thresholds = dict(self.DEFAULT_THRESHOLDS)
+        if thresholds:
+            self.thresholds.update(thresholds)
+        self.fail_on_warnings = fail_on_warnings
+
+    def check(
+        self,
+        ontology: Any,
+        graph_data: Optional[Dict[str, Any]] = None,
+        *,
+        thresholds: Optional[Mapping[str, Optional[float]]] = None,
+        fail_on_warnings: Optional[bool] = None,
+        competency_questions: Optional[List[str]] = None,
+    ) -> OntologyQualityReport:
+        """Check an ontology and optionally its instance graph.
+
+        ``graph_data`` is optional because an ontology can be checked before
+        instances are available. When omitted, embedded ``entities`` and
+        ``relationships`` are checked when present.
+        """
+        active_thresholds = dict(self.thresholds)
+        if thresholds:
+            active_thresholds.update(thresholds)
+        min_coverage, max_errors, max_warnings = self._validate_thresholds(
+            active_thresholds
+        )
+        should_fail_on_warnings = (
+            self.fail_on_warnings if fail_on_warnings is None else fail_on_warnings
+        )
+        issues: List[QualityIssue] = []
+
+        if not isinstance(ontology, dict):
+            self._add_issue(
+                issues,
+                "INVALID_ONTOLOGY",
+                "Ontology must be a dictionary.",
+                QualitySeverity.CRITICAL,
+                element_type="ontology",
+            )
+            return self._build_report(
+                issues,
+                classes=0,
+                properties=0,
+                entities=0,
+                relationships=0,
+                metrics={
+                    "coverage": 0.0,
+                    "class_coverage": 0.0,
+                    "property_coverage": 0.0,
+                },
+                thresholds=active_thresholds,
+                min_coverage=min_coverage,
+                max_errors=max_errors,
+                max_warnings=max_warnings,
+                fail_on_warnings=should_fail_on_warnings,
+            )
+
+        validation = self.validator.validate(ontology)
+        for message in getattr(validation, "errors", []) or []:
+            self._add_issue(
+                issues,
+                "VALIDATOR_ERROR",
+                str(message),
+                QualitySeverity.ERROR,
+                element_type="ontology",
+            )
+        for message in getattr(validation, "warnings", []) or []:
+            self._add_issue(
+                issues,
+                "VALIDATOR_WARNING",
+                str(message),
+                QualitySeverity.WARNING,
+                element_type="ontology",
+            )
+
+        classes = self._read_collection(ontology, "classes", issues)
+        properties = self._read_collection(ontology, "properties", issues)
+        class_aliases, class_ids = self._index_elements(
+            classes, "class", "MISSING_CLASS_ID", issues
+        )
+        referenced_classes: Set[str] = set()
+        self._mark_hierarchy(classes, class_aliases, referenced_classes)
+
+        property_with_endpoints = 0
+        for index, prop in enumerate(properties):
+            if not isinstance(prop, dict):
+                self._add_issue(
+                    issues,
+                    "INVALID_PROPERTY",
+                    f"Property at index {index} must be a dictionary.",
+                    QualitySeverity.ERROR,
+                    element_type="property",
+                    details={"index": index},
+                )
+                continue
+
+            prop_id = self._identifier(prop)
+            if prop_id is None:
+                prop_id = f"property[{index}]"
+                self._add_issue(
+                    issues,
+                    "MISSING_PROPERTY_ID",
+                    f"Property at index {index} has no name or URI.",
+                    QualitySeverity.ERROR,
+                    element_type="property",
+                    details={"index": index},
+                )
+
+            raw_prop_type = prop.get("type")
+            prop_type = (
+                str(raw_prop_type).strip().lower() if raw_prop_type is not None else ""
+            )
+            if not prop_type:
+                self._add_issue(
+                    issues,
+                    "MISSING_PROPERTY_TYPE",
+                    f"Property '{prop_id}' has no type.",
+                    QualitySeverity.ERROR,
+                    element_id=prop_id,
+                    element_type="property",
+                )
+            elif (
+                prop_type not in self._DATA_PROPERTY_TYPES | self._OBJECT_PROPERTY_TYPES
+            ):
+                self._add_issue(
+                    issues,
+                    "UNKNOWN_PROPERTY_TYPE",
+                    f"Property '{prop_id}' has unknown type '{prop_type}'.",
+                    QualitySeverity.WARNING,
+                    element_id=prop_id,
+                    element_type="property",
+                )
+
+            domains = self._values(prop, "domain")
+            ranges = self._values(prop, "range")
+            if domains or ranges:
+                property_with_endpoints += 1
+            else:
+                self._add_issue(
+                    issues,
+                    "ORPHAN_PROPERTY",
+                    f"Property '{prop_id}' has no domain or range.",
+                    QualitySeverity.WARNING,
+                    element_id=prop_id,
+                    element_type="property",
+                )
+            if not domains:
+                self._add_issue(
+                    issues,
+                    "MISSING_DOMAIN",
+                    f"Property '{prop_id}' has no domain.",
+                    QualitySeverity.WARNING,
+                    element_id=prop_id,
+                    element_type="property",
+                )
+            for domain in domains:
+                matched = self._match_class(domain, class_aliases)
+                if matched:
+                    referenced_classes.add(matched)
+                elif not self._is_builtin_class(domain):
+                    self._add_issue(
+                        issues,
+                        "UNKNOWN_DOMAIN",
+                        f"Property '{prop_id}' references unknown domain '{domain}'.",
+                        QualitySeverity.ERROR,
+                        element_id=prop_id,
+                        element_type="property",
+                        details={"domain": domain},
+                    )
+
+            if not ranges:
+                self._add_issue(
+                    issues,
+                    "MISSING_RANGE",
+                    f"Property '{prop_id}' has no range.",
+                    QualitySeverity.WARNING,
+                    element_id=prop_id,
+                    element_type="property",
+                )
+            for range_value in ranges:
+                matched = self._match_class(range_value, class_aliases)
+                if matched:
+                    referenced_classes.add(matched)
+                self._check_range(
+                    prop_id,
+                    prop_type,
+                    range_value,
+                    matched is not None,
+                    issues,
+                )
+
+        graph = graph_data
+        if graph is None and ("entities" in ontology or "relationships" in ontology):
+            graph = ontology
+        graph_entities, graph_relationships = self._read_graph(graph, issues)
+        if self._has_valid_graph_shape(graph):
+            self._check_graph(graph, issues)
+            self._mark_graph_types(graph_entities, class_aliases, referenced_classes)
+
+        for class_id in class_ids:
+            if class_id not in referenced_classes:
+                self._add_issue(
+                    issues,
+                    "ORPHAN_CLASS",
+                    f"Class '{class_id}' is not connected to a property, hierarchy, or graph entity type.",
+                    QualitySeverity.WARNING,
+                    element_id=class_id,
+                    element_type="class",
+                )
+
+        class_coverage = (
+            len(referenced_classes & set(class_ids)) / len(class_ids)
+            if class_ids
+            else 0.0
+        )
+        property_coverage = (
+            property_with_endpoints / len(properties) if properties else 1.0
+        )
+        metrics: Dict[str, float] = {
+            "coverage": (
+                (class_coverage + property_coverage) / 2
+                if classes or properties
+                else 0.0
+            ),
+            "class_coverage": class_coverage,
+            "property_coverage": property_coverage,
+            "validator_valid": 1.0 if getattr(validation, "valid", True) else 0.0,
+        }
+        if competency_questions is not None:
+            evaluation_ontology = self._prepare_for_evaluation(
+                ontology, classes, properties
+            )
+            evaluation = self._evaluate_competency_questions(
+                evaluation_ontology, competency_questions=competency_questions
+            )
+            metrics["competency_question_coverage"] = evaluation.coverage_score
+            metrics["completeness"] = evaluation.completeness_score
+
+        return self._build_report(
+            issues,
+            classes=len(classes),
+            properties=len(properties),
+            entities=len(graph_entities),
+            relationships=len(graph_relationships),
+            metrics=metrics,
+            thresholds=active_thresholds,
+            min_coverage=min_coverage,
+            max_errors=max_errors,
+            max_warnings=max_warnings,
+            fail_on_warnings=should_fail_on_warnings,
+        )
+
+    def _check_range(
+        self,
+        prop_id: str,
+        prop_type: str,
+        range_value: Any,
+        is_class: bool,
+        issues: List[QualityIssue],
+    ) -> None:
+        if prop_type in self._DATA_PROPERTY_TYPES:
+            if not self._is_known_datatype(range_value):
+                self._add_issue(
+                    issues,
+                    "INVALID_DATATYPE_RANGE",
+                    f"Data property '{prop_id}' has invalid range '{range_value}'.",
+                    QualitySeverity.ERROR,
+                    element_id=prop_id,
+                    element_type="property",
+                    details={"range": range_value},
+                )
+        elif prop_type in self._OBJECT_PROPERTY_TYPES:
+            if not is_class and not self._is_builtin_class(range_value):
+                self._add_issue(
+                    issues,
+                    "UNKNOWN_RANGE",
+                    f"Object property '{prop_id}' references unknown range '{range_value}'.",
+                    QualitySeverity.ERROR,
+                    element_id=prop_id,
+                    element_type="property",
+                    details={"range": range_value},
+                )
+        elif (
+            not is_class
+            and not self._is_builtin_class(range_value)
+            and not self._is_known_datatype(range_value)
+        ):
+            self._add_issue(
+                issues,
+                "UNKNOWN_RANGE",
+                f"Property '{prop_id}' references unknown range '{range_value}'.",
+                QualitySeverity.ERROR,
+                element_id=prop_id,
+                element_type="property",
+                details={"range": range_value},
+            )
+
+    def _build_report(
+        self,
+        issues: List[QualityIssue],
+        *,
+        classes: int,
+        properties: int,
+        entities: int,
+        relationships: int,
+        metrics: Dict[str, float],
+        thresholds: Dict[str, Optional[float]],
+        min_coverage: float,
+        max_errors: float,
+        max_warnings: Optional[float],
+        fail_on_warnings: bool,
+    ) -> OntologyQualityReport:
+        failures: List[str] = []
+        error_count = sum(
+            issue.severity in (QualitySeverity.ERROR, QualitySeverity.CRITICAL)
+            for issue in issues
+        )
+        warning_count = sum(
+            issue.severity == QualitySeverity.WARNING for issue in issues
+        )
+        if error_count > max_errors:
+            failures.append("max_errors")
+        if max_warnings is not None and warning_count > max_warnings:
+            failures.append("max_warnings")
+        if fail_on_warnings and warning_count:
+            failures.append("fail_on_warnings")
+        if metrics.get("coverage", 0.0) < min_coverage:
+            failures.append("min_coverage")
+        return OntologyQualityReport(
+            passed=not failures,
+            issues=issues,
+            stats={
+                "classes": classes,
+                "properties": properties,
+                "entities": entities,
+                "relationships": relationships,
+            },
+            metrics=metrics,
+            thresholds=thresholds,
+            threshold_failures=failures,
+        )
+
+    @staticmethod
+    def _validate_thresholds(
+        thresholds: Mapping[str, Optional[float]],
+    ) -> Tuple[float, float, Optional[float]]:
+        min_coverage = float(thresholds.get("min_coverage", 0.0) or 0.0)
+        max_errors = float(thresholds.get("max_errors", 0.0) or 0.0)
+        warning_value = thresholds.get("max_warnings")
+        max_warnings = None if warning_value is None else float(warning_value)
+        if not all(
+            math.isfinite(value)
+            for value in (min_coverage, max_errors)
+            if value is not None
+        ) or (max_warnings is not None and not math.isfinite(max_warnings)):
+            raise ValueError("quality thresholds must be finite numbers")
+        if not 0.0 <= min_coverage <= 1.0:
+            raise ValueError("min_coverage must be between 0.0 and 1.0")
+        if max_errors < 0 or (max_warnings is not None and max_warnings < 0):
+            raise ValueError("error and warning thresholds cannot be negative")
+        return min_coverage, max_errors, max_warnings
+
+    @classmethod
+    def _prepare_for_evaluation(
+        cls,
+        ontology: Dict[str, Any],
+        classes: Iterable[Any],
+        properties: Iterable[Any],
+    ) -> Dict[str, Any]:
+        """Make a shallow, evaluator-safe view without changing caller data."""
+        prepared = dict(ontology)
+        prepared["classes"] = [
+            cls._prepare_element(element)
+            for element in classes
+            if isinstance(element, dict)
+        ]
+        prepared["properties"] = [
+            cls._prepare_element(element)
+            for element in properties
+            if isinstance(element, dict)
+        ]
+        return prepared
+
+    @classmethod
+    def _prepare_element(cls, element: Dict[str, Any]) -> Dict[str, Any]:
+        prepared = dict(element)
+        identifier = cls._identifier(prepared)
+        if identifier is not None and not str(prepared.get("name", "")).strip():
+            prepared["name"] = identifier
+        return prepared
+
+    def _evaluate_competency_questions(
+        self, ontology: Dict[str, Any], competency_questions: List[str]
+    ) -> Any:
+        """Evaluate with an isolated question manager for repeatable checks."""
+        evaluator = copy.copy(self.evaluator)
+        manager = getattr(self.evaluator, "competency_questions_manager", None)
+        if manager is not None and hasattr(manager, "questions"):
+            isolated_manager = copy.copy(manager)
+            isolated_manager.questions = []
+            evaluator.competency_questions_manager = isolated_manager
+        return evaluator.evaluate_ontology(
+            ontology, competency_questions=competency_questions
+        )
+
+    @classmethod
+    def _read_collection(
+        cls, ontology: Dict[str, Any], key: str, issues: List[QualityIssue]
+    ) -> List[Any]:
+        if key not in ontology:
+            cls._add_issue(
+                issues,
+                f"MISSING_{key.upper()}",
+                f"Ontology has no {key} defined.",
+                QualitySeverity.WARNING,
+                element_type="ontology",
+            )
+            return []
+        value = ontology[key]
+        if not isinstance(value, list):
+            cls._add_issue(
+                issues,
+                f"INVALID_{key.upper()}",
+                f"Ontology '{key}' must be a list.",
+                QualitySeverity.ERROR,
+                element_type="ontology",
+            )
+            return []
+        return value
+
+    @classmethod
+    def _index_elements(
+        cls,
+        elements: Iterable[Any],
+        element_type: str,
+        missing_code: str,
+        issues: List[QualityIssue],
+    ) -> Tuple[Dict[str, str], List[str]]:
+        aliases: Dict[str, str] = {}
+        identifiers: List[str] = []
+        for index, element in enumerate(elements):
+            identifier = cls._identifier(element)
+            if identifier is None:
+                cls._add_issue(
+                    issues,
+                    missing_code,
+                    f"{element_type.title()} at index {index} has no name or URI.",
+                    QualitySeverity.ERROR,
+                    element_type=element_type,
+                    details={"index": index},
+                )
+                continue
+            identifiers.append(identifier)
+            aliases_for_element: Set[str] = set()
+            for candidate in cls._identifiers(element):
+                aliases_for_element.update(cls._term_aliases(candidate))
+            for alias in sorted(aliases_for_element):
+                aliases.setdefault(alias, identifier)
+        return aliases, identifiers
+
+    @classmethod
+    def _mark_hierarchy(
+        cls,
+        classes: Iterable[Any],
+        aliases: Mapping[str, str],
+        referenced: Set[str],
+    ) -> None:
+        for class_entry in classes:
+            if not isinstance(class_entry, dict):
+                continue
+            identifier = cls._identifier(class_entry)
+            for key in (
+                "subClassOf",
+                "subclassOf",
+                "parent",
+                "superclass",
+                "superclasses",
+            ):
+                parents = cls._values(class_entry, key)
+                if parents and identifier:
+                    referenced.add(identifier)
+                for parent in parents:
+                    matched = cls._match_class(parent, aliases)
+                    if matched:
+                        referenced.add(matched)
+
+    @classmethod
+    def _mark_graph_types(
+        cls,
+        entities: Iterable[Any],
+        aliases: Mapping[str, str],
+        referenced: Set[str],
+    ) -> None:
+        for entity in entities:
+            if not isinstance(entity, dict):
+                continue
+            entity_type = entity.get("type") or entity.get("entity_type")
+            matched = cls._match_class(entity_type, aliases)
+            if matched:
+                referenced.add(matched)
+
+    @classmethod
+    def _check_graph(cls, graph: Dict[str, Any], issues: List[QualityIssue]) -> None:
+        safe_graph = cls._prepare_graph_for_validation(graph, issues)
+        result = GraphValidator().validate(safe_graph)
+        code_map = {
+            "DANGLING_EDGE": "UNRESOLVED_RELATIONSHIP_ENDPOINT",
+            "ORPHAN_NODES": "ORPHAN_ENTITY",
+        }
+        severity_map = {
+            "info": QualitySeverity.INFO,
+            "warning": QualitySeverity.WARNING,
+            "error": QualitySeverity.ERROR,
+            "critical": QualitySeverity.CRITICAL,
+        }
+        for graph_issue in result.issues:
+            severity = severity_map.get(
+                graph_issue.severity.value, QualitySeverity.ERROR
+            )
+            details = dict(graph_issue.details or {})
+            if graph_issue.code == "ORPHAN_NODES" and "ids" in details:
+                details["ids"] = sorted(details["ids"], key=str)
+            cls._add_issue(
+                issues,
+                code_map.get(graph_issue.code, graph_issue.code),
+                graph_issue.message,
+                severity,
+                element_id=graph_issue.element_id,
+                element_type=graph_issue.element_type,
+                details=details,
+            )
+
+    @classmethod
+    def _prepare_graph_for_validation(
+        cls, graph: Dict[str, Any], issues: List[QualityIssue]
+    ) -> Dict[str, List[Dict[str, Any]]]:
+        """Normalize supported graph aliases and isolate malformed members."""
+        entities: List[Dict[str, Any]] = []
+        raw_entities = graph.get("entities", [])
+        raw_relationships = graph.get("relationships", [])
+        for index, entity in enumerate(
+            raw_entities if isinstance(raw_entities, list) else []
+        ):
+            if not isinstance(entity, dict):
+                cls._add_issue(
+                    issues,
+                    "INVALID_ENTITY",
+                    f"Entity at index {index} must be a dictionary.",
+                    QualitySeverity.ERROR,
+                    element_type="entity",
+                    details={"index": index},
+                )
+                continue
+            normalized = dict(entity)
+            if not normalized.get("name") and normalized.get("text") is not None:
+                normalized["name"] = normalized["text"]
+            entities.append(normalized)
+
+        relationships: List[Dict[str, Any]] = []
+        for index, relationship in enumerate(
+            raw_relationships if isinstance(raw_relationships, list) else []
+        ):
+            if not isinstance(relationship, dict):
+                cls._add_issue(
+                    issues,
+                    "INVALID_RELATIONSHIP",
+                    f"Relationship at index {index} must be a dictionary.",
+                    QualitySeverity.ERROR,
+                    element_type="relationship",
+                    details={"index": index},
+                )
+                continue
+            relationships.append(dict(relationship))
+
+        return {"entities": entities, "relationships": relationships}
+
+    @classmethod
+    def _read_graph(
+        cls, graph: Optional[Dict[str, Any]], issues: List[QualityIssue]
+    ) -> Tuple[List[Any], List[Any]]:
+        if graph is None:
+            return [], []
+        if not isinstance(graph, dict):
+            cls._add_issue(
+                issues,
+                "INVALID_GRAPH",
+                "Graph data must be a dictionary.",
+                QualitySeverity.ERROR,
+                element_type="graph",
+            )
+            return [], []
+        entities = graph.get("entities", [])
+        relationships = graph.get("relationships", [])
+        if not isinstance(entities, list) or not isinstance(relationships, list):
+            cls._add_issue(
+                issues,
+                "INVALID_GRAPH",
+                "Graph entities and relationships must be lists.",
+                QualitySeverity.ERROR,
+                element_type="graph",
+            )
+            return [], []
+        return entities, relationships
+
+    @staticmethod
+    def _has_valid_graph_shape(graph: Optional[Dict[str, Any]]) -> bool:
+        return bool(
+            isinstance(graph, dict)
+            and isinstance(graph.get("entities", []), list)
+            and isinstance(graph.get("relationships", []), list)
+        )
+
+    @staticmethod
+    def _identifier(element: Any) -> Optional[str]:
+        identifiers = OntologyQualityGate._identifiers(element)
+        return identifiers[0] if identifiers else None
+
+    @staticmethod
+    def _identifiers(element: Any) -> List[str]:
+        if isinstance(element, dict):
+            values = []
+            for key in ("name", "uri", "id", "@id"):
+                value = element.get(key)
+                if value is not None and str(value).strip():
+                    values.append(str(value).strip())
+            return values
+        if isinstance(element, str) and element.strip():
+            return [element.strip()]
+        return []
+
+    @staticmethod
+    def _values(element: Dict[str, Any], key: str) -> List[Any]:
+        value = element.get(key)
+        if value is None:
+            return []
+        if isinstance(value, (list, tuple, set)):
+            values = [item for item in value if item is not None and str(item).strip()]
+            return sorted(values, key=str) if isinstance(value, set) else values
+        return [value] if str(value).strip() else []
+
+    @classmethod
+    def _term_aliases(cls, value: Any) -> Set[str]:
+        if isinstance(value, dict):
+            value = cls._identifier(value)
+        if value is None:
+            return set()
+        text = str(value).strip().strip("<>")
+        if not text:
+            return set()
+        aliases = {text, text.lower()}
+        for separator in ("#", "/"):
+            if separator in text:
+                local = text.rstrip("/").rsplit(separator, 1)[-1]
+                aliases.update({local, local.lower()})
+        if ":" in text and not text.startswith(("http://", "https://")):
+            local = text.rsplit(":", 1)[-1]
+            aliases.update({local, local.lower()})
+        return aliases
+
+    @classmethod
+    def _match_class(cls, value: Any, aliases: Mapping[str, str]) -> Optional[str]:
+        for alias in sorted(cls._term_aliases(value)):
+            if alias in aliases:
+                return aliases[alias]
+        return None
+
+    @classmethod
+    def _is_builtin_class(cls, value: Any) -> bool:
+        return any(alias in cls._BUILTIN_CLASSES for alias in cls._term_aliases(value))
+
+    @classmethod
+    def _is_known_datatype(cls, value: Any) -> bool:
+        return any(alias in cls._KNOWN_DATATYPES for alias in cls._term_aliases(value))
+
+    @staticmethod
+    def _add_issue(
+        issues: List[QualityIssue],
+        code: str,
+        message: str,
+        severity: QualitySeverity,
+        *,
+        element_id: Optional[str] = None,
+        element_type: Optional[str] = None,
+        details: Optional[Dict[str, Any]] = None,
+    ) -> None:
+        issues.append(
+            QualityIssue(
+                code=code,
+                message=message,
+                severity=severity,
+                element_id=element_id,
+                element_type=element_type,
+                details=details or {},
+            )
+        )
+
+
+def ontology_quality_check(
+    ontology: Any,
+    graph_data: Optional[Dict[str, Any]] = None,
+    *,
+    thresholds: Optional[Mapping[str, Optional[float]]] = None,
+    fail_on_warnings: bool = False,
+    competency_questions: Optional[List[str]] = None,
+    validator: Optional[OntologyValidator] = None,
+    evaluator: Optional[OntologyEvaluator] = None,
+) -> OntologyQualityReport:
+    """Convenience wrapper around :class:`OntologyQualityGate`."""
+    gate = OntologyQualityGate(
+        validator=validator,
+        evaluator=evaluator,
+        thresholds=thresholds,
+        fail_on_warnings=fail_on_warnings,
+    )
+    return gate.check(
+        ontology,
+        graph_data=graph_data,
+        competency_questions=competency_questions,
+    )
@@ -5,6 +5,7 @@ This module provides the REST API server for the Semantica framework
 using FastAPI and uvicorn.
 """

+import asyncio
 import logging
 import os
 import uvicorn
@@ -24,8 +25,10 @@ from .utils.logging import setup_logging
 try:
    from .context.context_graph import ContextGraph
    from .explorer.session import GraphSession
-    from .explorer.ws import ConnectionManager
+    from .explorer.ws import ConnectionManager, install_graph_updates_websocket
    from .explorer.dependencies import anonymous_access_allowed, get_expected_api_key, require_auth
+    from .explorer.markdown_resources import MarkdownResourceRegistry
+    from .explorer.runtime import explorer_capabilities, install_mutation_bridge
    EXPLORER_AVAILABLE = True
 except ImportError:
    EXPLORER_AVAILABLE = False
@@ -57,16 +60,24 @@ async def lifespan(app: FastAPI):
            graph = ContextGraph()
            app.state.session = GraphSession(graph)
            app.state.ws_manager = ConnectionManager()
+            app.state.event_loop = asyncio.get_running_loop()
+            app.state.agent_memory = None
+            app.state.markdown_resources = MarkdownResourceRegistry(graph)
+            install_mutation_bridge(app, app.state.session)
            logging.info("Database Session and WebSockets attached to app state.")
        except Exception as e:
            logging.error(f"Failed to initialize GraphSession: {e}")
            app.state.session = None
            app.state.ws_manager = None
+            app.state.agent_memory = None
+            app.state.markdown_resources = None
    else:
        app.state.session = None
        app.state.ws_manager = None
+        app.state.agent_memory = None
+        app.state.markdown_resources = None

-    yield  
+    yield

    logging.info("Shutting down Semantica API...")
    if getattr(app.state, "session", None) and hasattr(app.state.session.graph, "close"):
@@ -91,11 +102,14 @@ app.add_middleware(
    CORSMiddleware,
    allow_origins=_cors_origins,
    allow_credentials=True,
-    allow_methods=["GET", "POST", "DELETE", "OPTIONS"],
+    allow_methods=["GET", "POST", "PUT", "DELETE", "OPTIONS"],
    allow_headers=["Content-Type", "Authorization", "X-API-Key"],
    max_age=600,
 )

+if EXPLORER_AVAILABLE:
+    install_graph_updates_websocket(app, _cors_origins)
+

 # --- Security response headers -------------------------------------
 class _SecurityHeadersMiddleware(BaseHTTPMiddleware):
@@ -138,7 +152,10 @@ async def root():
    return {
        "name": "Semantica API",
        "version": __version__,
-        "status": "active"
+        "status": "active",
+        "capabilities": explorer_capabilities(
+            getattr(app.state, "agent_memory", None)
+        ),
    }

@app.get("/health")
@@ -165,6 +182,8 @@ if EXPLORER_AVAILABLE:
            enrich,
            export_import,
            graph,
+            markdown,
+            memories,
            ontology,
            temporal,
            vocabulary,
@@ -179,6 +198,8 @@ if EXPLORER_AVAILABLE:
        app.include_router(enrich.router, dependencies=_auth)
        app.include_router(export_import.router, dependencies=_auth)
        app.include_router(graph.router, dependencies=_auth)
+        app.include_router(markdown.router, dependencies=_auth)
+        app.include_router(memories.router, dependencies=_auth)
        app.include_router(ontology.router, dependencies=_auth)
        app.include_router(temporal.router, dependencies=_auth)
        app.include_router(vocabulary.router, dependencies=_auth)
@@ -260,4 +281,4 @@ def main():
    uvicorn.run(app, host=host, port=8000)

 if __name__ == "__main__":
-    main()
+    main()
@@ -12,6 +12,10 @@ import pytest
 import yaml

 from semantica.context.agent_memory import AgentMemory
+from semantica.context.markdown import (
+    MarkdownRevisionConflictError,
+    markdown_document_revision,
+)

 _ERROR_PRIVILEGE_NOT_HELD = 1314

@@ -1008,3 +1012,144 @@ def test_markdown_import_rejects_non_regular_file(tmp_path):
    with patch("semantica.context.agent_memory.os.fstat", return_value=fake_stat):
        with pytest.raises(ValueError, match="not a regular file"):
            memory._read_markdown_file_content(real_file)
+
+
+def _editable_memory(vector_store=None):
+    memory = AgentMemory(vector_store=vector_store)
+    memory.store(
+        "Original body",
+        memory_id="mem-edit",
+        timestamp=datetime.fromisoformat("2026-07-22T09:00:00+00:00"),
+        metadata={
+            "type": "note",
+            "updated_at": "2026-07-22T10:00:00+00:00",
+            "owner": "before",
+        },
+    )
+    return memory
+
+
+def test_single_item_markdown_export_and_apply_preserve_identity():
+    memory = _editable_memory()
+    source = memory.export_item_markdown("mem-edit")
+    edited = source.replace("owner: before", "owner: after").replace(
+        "Original body", "Updated body"
+    )
+
+    assert memory.apply_item_markdown("mem-edit", edited) is True
+    item = memory.get("mem-edit")
+    assert item["memory_id"] == "mem-edit"
+    assert item["content"] == "Updated body"
+    assert item["metadata"]["owner"] == "after"
+    assert memory.export_item_markdown("mem-edit") == edited
+
+
+def test_single_item_markdown_rejects_identity_change_without_mutation():
+    memory = _editable_memory()
+    before = deepcopy(memory.get("mem-edit"))
+    document = memory.export_item_markdown("mem-edit").replace(
+        "id: mem-edit", "id: mem-other"
+    )
+
+    with pytest.raises(ValueError, match="does not match resource id"):
+        memory.apply_item_markdown("mem-edit", document)
+
+    assert memory.get("mem-edit") == before
+    assert memory.get("mem-other") is None
+
+
+def test_single_item_markdown_invalid_document_and_missing_item_are_safe():
+    memory = _editable_memory()
+    before = deepcopy(memory.get("mem-edit"))
+
+    with pytest.raises(ValueError, match="Invalid Markdown frontmatter"):
+        memory.apply_item_markdown("mem-edit", "---\nid: [\n---\n\nBroken")
+    with pytest.raises(KeyError, match="missing"):
+        memory.export_item_markdown("missing")
+
+    assert memory.get("mem-edit") == before
+
+
+def test_single_item_markdown_noop_skips_vector_sync():
+    vector_store = TrackingVectorStore()
+    memory = _editable_memory(vector_store=vector_store)
+    vector_store.events.clear()
+    source = memory.export_item_markdown("mem-edit")
+
+    assert memory.apply_item_markdown("mem-edit", source) is False
+    assert vector_store.events == []
+
+
+def test_single_item_markdown_revision_check_is_atomic_with_apply():
+    memory = _editable_memory()
+    source = memory.export_item_markdown("mem-edit")
+    expected_revision = markdown_document_revision(source)
+    memory.update("mem-edit", content="Concurrent update")
+
+    with pytest.raises(MarkdownRevisionConflictError):
+        memory.apply_item_markdown(
+            "mem-edit",
+            source.replace("Original body", "Stale edit"),
+            expected_revision=expected_revision,
+        )
+
+    assert memory.get("mem-edit")["content"] == "Concurrent update"
+
+
+def test_apply_item_markdown_rolls_back_all_state_on_store_failure():
+    """apply_item_markdown fully restores AgentMemory state if _replace_memory_item fails.
+
+    The test injects a failure inside the store() call that executes AFTER
+    delete_memory() completes, so that at least one real mutation has already
+    happened before the exception is raised.  This verifies the rollback path
+    inside _replace_memory_item is exercised through apply_item_markdown, not
+    just through import_data.
+    """
+    memory = _editable_memory()
+    original_item = deepcopy(memory.get("mem-edit"))
+    original_stats = deepcopy(memory.stats)
+    original_index = list(memory.memory_index)
+    original_stm = [item.memory_id for item in memory.short_term_memory]
+    original_vector_ids = deepcopy(memory._vector_ids)
+
+    source = memory.export_item_markdown("mem-edit")
+    edited = source.replace("Original body", "Replaced body")
+
+    # Patch store() so it succeeds (internally called by _replace_memory_item),
+    # but raises AFTER the new item is inserted into memory_items.
+    real_store = memory.store
+    call_count = 0
+
+    def fail_after_store(*args, **kwargs):
+        nonlocal call_count
+        call_count += 1
+        result = real_store(*args, **kwargs)
+        # Raise on the store() call made by _replace_memory_item (not any earlier call)
+        if call_count >= 1:
+            raise RuntimeError("simulated store failure after insert")
+        return result
+
+    with patch.object(memory, "store", side_effect=fail_after_store):
+        with pytest.raises(RuntimeError, match="simulated store failure"):
+            memory.apply_item_markdown("mem-edit", edited)
+
+    # All state must be identical to what it was before apply_item_markdown
+    assert memory.get("mem-edit") == original_item, (
+        "memory_items must be fully restored after rollback"
+    )
+    assert memory.stats == original_stats, (
+        "stats must be fully restored after rollback"
+    )
+    assert list(memory.memory_index) == original_index, (
+        "memory_index must be fully restored after rollback"
+    )
+    assert [item.memory_id for item in memory.short_term_memory] == original_stm, (
+        "short_term_memory must be fully restored after rollback"
+    )
+    assert memory._vector_ids == original_vector_ids, (
+        "vector_ids must be fully restored after rollback"
+    )
+    # The memory must still be readable and unchanged
+    assert memory.exists("mem-edit")
+    assert memory.get("mem-edit")["content"] == "Original body"
+    assert memory.count() == 1
@@ -9,6 +9,10 @@ import yaml
 import semantica.context.context_graph as context_graph_module
 from semantica.change_management.managers import TemporalVersionManager
 from semantica.context.context_graph import ContextEdge, ContextGraph, ContextNode
+from semantica.context.markdown import (
+    MarkdownRevisionConflictError,
+    markdown_document_revision,
+)


 def _read_markdown(path: Path):
@@ -711,3 +715,135 @@ def test_json_remains_default_and_unknown_format_is_rejected(tmp_path):

    with pytest.raises(ValueError, match="Unsupported context graph"):
        graph.save_to_file(tmp_path / "graph", format="html")
+
+
+def _edited_node_document(graph, node_id, **frontmatter_updates):
+    source = graph.export_node_markdown(node_id)
+    frontmatter, body = graph._parse_markdown_document(source, f"node {node_id!r}")
+    frontmatter.update(frontmatter_updates)
+    return graph._render_markdown_document(
+        frontmatter,
+        body.replace("Keep evidence.", "Keep verified evidence."),
+        f"node {node_id!r}",
+    )
+
+
+def test_single_node_markdown_export_and_apply_update_complete_node_state():
+    graph, _, _ = _sample_graph()
+    events = []
+    graph.mutation_callback = lambda *event: events.append(event)
+    before_edges = list(graph.edges)
+    document = _edited_node_document(
+        graph,
+        "policy/\u6771\u4eac",
+        type="Decision",
+        properties={"category": "retention", "reviewed": True},
+        metadata={"source": "editor"},
+        valid_from="2026-03-01T00:00:00+00:00",
+        valid_until="2028-01-01T00:00:00+00:00",
+    )
+
+    assert graph.apply_node_markdown("policy/\u6771\u4eac", document) is True
+    node = graph.nodes["policy/\u6771\u4eac"]
+    assert node.node_id == "policy/\u6771\u4eac"
+    assert node.node_type == "Decision"
+    assert node.content == "# Retention\n\nKeep verified evidence.\n---\n"
+    assert node.properties == {"category": "retention", "reviewed": True}
+    assert node.metadata == {"source": "editor"}
+    assert node.valid_from == "2026-03-01T00:00:00+00:00"
+    assert node.valid_until == "2028-01-01T00:00:00+00:00"
+    assert "policy/\u6771\u4eac" not in graph.node_type_index.get("Policy", set())
+    assert "policy/\u6771\u4eac" in graph.node_type_index["Decision"]
+    assert "policy/\u6771\u4eac" in graph._decisions
+    assert graph.edges == before_edges
+    assert [event[0] for event in events] == ["UPDATE_NODE"]
+    assert graph.export_node_markdown("policy/\u6771\u4eac") == document
+
+
+def test_single_node_markdown_transition_out_of_decision_clears_indexes():
+    graph = ContextGraph(advanced_analytics=False)
+    graph.add_node("decision-1", "Decision", "Choose")
+    graph._rebuild_decision_indexes()
+    source = graph.export_node_markdown("decision-1")
+    document = source.replace("type: Decision", "type: Policy")
+
+    assert graph.apply_node_markdown("decision-1", document) is True
+    assert "decision-1" not in graph._decisions
+
+
+def test_single_node_markdown_invalid_identity_and_yaml_do_not_mutate():
+    graph, _, _ = _sample_graph()
+    events = []
+    graph.mutation_callback = lambda *event: events.append(event)
+    before = _normalized_state(graph)
+    source = graph.export_node_markdown("policy/\u6771\u4eac")
+
+    with pytest.raises(ValueError, match="does not match resource id"):
+        graph.apply_node_markdown(
+            "policy/\u6771\u4eac",
+            source.replace("id: policy/", "id: changed/"),
+        )
+    with pytest.raises(ValueError, match="Invalid Markdown frontmatter"):
+        graph.apply_node_markdown("policy/\u6771\u4eac", "---\nid: [\n---\n\nBroken")
+
+    assert _normalized_state(graph) == before
+    assert events == []
+
+
+def test_single_node_markdown_index_failure_rolls_back_complete_graph_state():
+    graph, _, _ = _sample_graph()
+    graph._analytics_cache["cached"] = {"score": 1.0}
+    events = []
+    graph.mutation_callback = lambda *event: events.append(event)
+    before = _normalized_state(graph)
+    before_type_index = {
+        node_type: set(node_ids)
+        for node_type, node_ids in graph.node_type_index.items()
+    }
+    before_analytics_cache = dict(graph._analytics_cache)
+    document = _edited_node_document(
+        graph,
+        "policy/\u6771\u4eac",
+        type="Decision",
+        properties={"confidence": {"invalid": True}},
+    )
+
+    with pytest.raises(TypeError):
+        graph.apply_node_markdown("policy/\u6771\u4eac", document)
+
+    assert _normalized_state(graph) == before
+    assert {
+        node_type: set(node_ids)
+        for node_type, node_ids in graph.node_type_index.items()
+    } == before_type_index
+    assert graph._analytics_cache == before_analytics_cache
+    assert not hasattr(graph, "_decisions")
+    assert events == []
+
+
+def test_single_node_markdown_noop_and_missing_node_are_safe():
+    graph, _, _ = _sample_graph()
+    events = []
+    graph.mutation_callback = lambda *event: events.append(event)
+    source = graph.export_node_markdown("policy/\u6771\u4eac")
+
+    assert graph.apply_node_markdown("policy/\u6771\u4eac", source) is False
+    with pytest.raises(KeyError, match="missing"):
+        graph.export_node_markdown("missing")
+    assert events == []
+
+
+def test_single_node_markdown_revision_check_is_atomic_with_apply():
+    graph, _, _ = _sample_graph()
+    source = graph.export_node_markdown("policy/\u6771\u4eac")
+    expected_revision = markdown_document_revision(source)
+    graph.add_node_attribute("policy/\u6771\u4eac", {"concurrent": True})
+
+    with pytest.raises(MarkdownRevisionConflictError):
+        graph.apply_node_markdown(
+            "policy/\u6771\u4eac",
+            source.replace("Keep evidence.", "Stale edit."),
+            expected_revision=expected_revision,
+        )
+
+    assert graph.nodes["policy/\u6771\u4eac"].properties["concurrent"] is True
@@ -247,6 +247,18 @@ class TestGraphNodes:
        assert payload["id"] == "python"
        assert payload["properties"]["content"] == "Python programming language"

+    def test_get_node_by_query_preserves_path_ids(self):
+        graph = ContextGraph(advanced_analytics=False)
+        graph.add_node("folder/node", node_type="note", content="Nested ID")
+        with TestClient(create_app(session=GraphSession(graph))) as test_client:
+            response = test_client.get(
+                "/api/graph/node",
+                params={"node_id": "folder/node"},
+            )
+
+        assert response.status_code == 200
+        assert response.json()["id"] == "folder/node"
+
    def test_get_neighbors(self, client):
        response = client.get("/api/graph/node/python/neighbors?depth=2")
        assert response.status_code == 200
@@ -60,6 +60,22 @@ def test_write_route_also_refuses_when_auth_not_configured(client, monkeypatch):
    assert resp.status_code == 503


+def test_markdown_write_route_requires_api_key(client, monkeypatch):
+    monkeypatch.delenv("SEMANTICA_ALLOW_ANONYMOUS", raising=False)
+    monkeypatch.setenv("SEMANTICA_API_KEY", "correct-key")
+
+    response = client.put(
+        "/api/markdown/context-node/python",
+        headers={"X-API-Key": "wrong-key"},
+        json={
+            "markdown": "---\nid: python\ntype: language\n---\n\nChanged",
+            "expected_revision": "sha256:" + ("0" * 64),
+        },
+    )
+
+    assert response.status_code == 401
+
+
 # ---------------------------------------------------------------------------
 # Configured key: wrong/missing key rejected, correct key accepted.
 # ---------------------------------------------------------------------------
@@ -0,0 +1,135 @@
+import pytest
+
+pytest.importorskip("fastapi")
+
+from fastapi import FastAPI  # noqa: E402
+from starlette.testclient import TestClient  # noqa: E402
+
+from semantica.context.context_graph import ContextGraph  # noqa: E402
+from semantica.explorer.runtime import install_mutation_bridge  # noqa: E402
+from semantica.explorer.session import GraphSession  # noqa: E402
+
+
+def test_mutation_bridge_supports_multiple_apps_for_one_graph(monkeypatch):
+    graph = ContextGraph(advanced_analytics=False)
+    first_session = GraphSession(graph)
+    second_session = GraphSession(graph)
+    received = []
+    graph.mutation_callback = lambda *event: received.append(("original", event))
+
+    monkeypatch.setattr(
+        first_session,
+        "handle_graph_mutation",
+        lambda *event: received.append(("first", event)),
+    )
+    monkeypatch.setattr(
+        second_session,
+        "handle_graph_mutation",
+        lambda *event: received.append(("second", event)),
+    )
+
+    first_app = FastAPI()
+    first_app.state.event_loop = None
+    first_app.state.ws_manager = None
+    second_app = FastAPI()
+    second_app.state.event_loop = None
+    second_app.state.ws_manager = None
+
+    install_mutation_bridge(first_app, first_session)
+    install_mutation_bridge(second_app, second_session)
+    graph.mutation_callback("UPDATE_NODE", "node-1", {"content": "Updated"})
+
+    assert [receiver for receiver, _ in received] == ["second", "first", "original"]
+
+
+def test_mutation_bridge_is_idempotent_for_one_app(monkeypatch):
+    graph = ContextGraph(advanced_analytics=False)
+    session = GraphSession(graph)
+    received = []
+    monkeypatch.setattr(
+        session,
+        "handle_graph_mutation",
+        lambda *event: received.append(event),
+    )
+    app = FastAPI()
+    app.state.event_loop = None
+    app.state.ws_manager = None
+
+    install_mutation_bridge(app, session)
+    install_mutation_bridge(app, session)
+    graph.mutation_callback("UPDATE_NODE", "node-1", {"content": "Updated"})
+
+    assert len(received) == 1
+
+
+def test_mutation_bridge_reinstalls_for_new_session_on_same_app(monkeypatch):
+    first_session = GraphSession(ContextGraph(advanced_analytics=False))
+    second_session = GraphSession(ContextGraph(advanced_analytics=False))
+    received = []
+    monkeypatch.setattr(
+        first_session,
+        "handle_graph_mutation",
+        lambda *event: received.append(("first", event)),
+    )
+    monkeypatch.setattr(
+        second_session,
+        "handle_graph_mutation",
+        lambda *event: received.append(("second", event)),
+    )
+    app = FastAPI()
+    app.state.event_loop = None
+    app.state.ws_manager = None
+
+    install_mutation_bridge(app, first_session)
+    install_mutation_bridge(app, second_session)
+    second_session.graph.mutation_callback(
+        "UPDATE_NODE",
+        "node-2",
+        {"content": "Updated"},
+    )
+
+    assert [receiver for receiver, _ in received] == ["second"]
+
+
+def test_legacy_server_mounts_editable_markdown_routes(monkeypatch):
+    monkeypatch.setenv("SEMANTICA_ALLOW_ANONYMOUS", "true")
+
+    from semantica import server
+
+    paths = {route.path for route in server.app.routes}
+    assert "/api/markdown/{kind}/{resource_id:path}" in paths
+    assert "/api/memories" in paths
+    assert "/ws/graph-updates" in paths
+
+    with TestClient(server.app) as client:
+        with client.websocket_connect("/ws/graph-updates") as websocket:
+            acknowledgement = websocket.receive_json()
+            assert acknowledgement["event"] == "connection_ack"
+            assert acknowledgement["data"] == {"connected": True}
+            assert acknowledgement["timestamp"]
+        info = client.get("/api/info")
+        memories = client.get("/api/memories")
+        server.app.state.session.graph.add_node(
+            "server-node",
+            "Note",
+            "Original server content",
+        )
+        current = client.get("/api/markdown/context-node/server-node").json()
+        saved = client.put(
+            "/api/markdown/context-node/server-node",
+            json={
+                "markdown": current["source"].replace(
+                    "Original server content",
+                    "Updated server content",
+                ),
+                "expected_revision": current["revision"],
+            },
+        )
+
+    assert info.json()["capabilities"]["agent_memory"] is False
+    assert memories.status_code == 503
+    assert memories.json()["detail"] == (
+        "AgentMemory is not configured for this Explorer instance."
+    )
+    assert saved.status_code == 200
+    assert saved.json()["body"] == "Updated server content"
@@ -0,0 +1,319 @@
+"""Integration tests for Explorer Markdown and memory selection routes."""
+
+import threading
+from datetime import datetime
+
+import pytest
+
+from semantica.context.agent_memory import AgentMemory
+from semantica.context.context_graph import ContextGraph
+
+pytest.importorskip("fastapi")
+
+from starlette.testclient import TestClient  # noqa: E402
+
+from semantica.explorer.app import create_app  # noqa: E402
+from semantica.explorer.session import GraphSession  # noqa: E402
+
+
+@pytest.fixture
+def markdown_client():
+    graph = ContextGraph(advanced_analytics=False)
+    graph.add_node("editable-node", "Note", "Original node")
+    events = []
+    graph.mutation_callback = lambda *event: events.append(event)
+    memory = AgentMemory()
+    memory.store(
+        "Original memory",
+        memory_id="editable-memory",
+        metadata={
+            "type": "note",
+            "updated_at": "2026-07-22T10:00:00+00:00",
+        },
+        timestamp=datetime.fromisoformat("2026-07-22T09:00:00+00:00"),
+    )
+    app = create_app(session=GraphSession(graph), agent_memory=memory)
+    with TestClient(app) as test_client:
+        yield test_client, graph, memory, events
+
+
+def test_gets_context_node_and_agent_memory(markdown_client):
+    client, _, _, _ = markdown_client
+
+    info = client.get("/api/info")
+    node = client.get("/api/markdown/context-node/editable-node")
+    memory = client.get("/api/markdown/agent-memory/editable-memory")
+
+    assert info.json()["capabilities"]["agent_memory"] is True
+    assert node.status_code == 200
+    assert node.json()["body"] == "Original node"
+    assert node.json()["resource"] == {
+        "kind": "context-node",
+        "id": "editable-node",
+    }
+    assert node.json()["source"].startswith("---\n")
+    assert node.json()["revision"].startswith("sha256:")
+    assert node.json()["editable"] is True
+    assert memory.status_code == 200
+    assert memory.json()["body"] == "Original memory"
+
+
+@pytest.mark.parametrize(
+    ("kind", "resource_id", "original", "updated"),
+    [
+        ("context-node", "editable-node", "Original node", "Updated node"),
+        ("agent-memory", "editable-memory", "Original memory", "Updated memory"),
+    ],
+)
+def test_put_updates_exact_resource(
+    markdown_client, kind, resource_id, original, updated
+):
+    client, graph, memory, _ = markdown_client
+    current = client.get(f"/api/markdown/{kind}/{resource_id}").json()
+
+    response = client.put(
+        f"/api/markdown/{kind}/{resource_id}",
+        json={
+            "markdown": current["source"].replace(original, updated),
+            "expected_revision": current["revision"],
+        },
+    )
+
+    assert response.status_code == 200
+    assert response.json()["changed"] is True
+    assert response.json()["body"] == updated
+    assert client.get(f"/api/markdown/{kind}/{resource_id}").json()["body"] == updated
+    if kind == "context-node":
+        assert graph.nodes[resource_id].content == updated
+        assert memory.get("editable-memory")["content"] == "Original memory"
+    else:
+        assert memory.get(resource_id)["content"] == updated
+        assert graph.nodes["editable-node"].content == "Original node"
+
+
+def test_validation_and_identity_errors_preserve_resource(markdown_client):
+    client, graph, _, events = markdown_client
+    current = client.get("/api/markdown/context-node/editable-node").json()
+
+    invalid = client.put(
+        "/api/markdown/context-node/editable-node",
+        json={
+            "markdown": "---\nid: [\n---\n\nBroken",
+            "expected_revision": current["revision"],
+        },
+    )
+    mismatch = client.put(
+        "/api/markdown/context-node/editable-node",
+        json={
+            "markdown": current["source"].replace(
+                "id: editable-node", "id: different-node"
+            ),
+            "expected_revision": current["revision"],
+        },
+    )
+
+    assert invalid.status_code == 422
+    assert invalid.json()["detail"]["code"] == "invalid_markdown_frontmatter"
+    assert "invalid YAML" in invalid.json()["detail"]["message"]
+    assert mismatch.status_code == 422
+    assert mismatch.json()["detail"] == {
+        "code": "resource_identity_mismatch",
+        "message": (
+            "Frontmatter id 'different-node' does not match resource id "
+            "'editable-node'."
+        ),
+        "field": "id",
+    }
+    assert graph.nodes["editable-node"].content == "Original node"
+    assert events == []
+
+
+def test_stale_revision_returns_conflict(markdown_client):
+    client, _, _, _ = markdown_client
+    current = client.get("/api/markdown/context-node/editable-node").json()
+    first = client.put(
+        "/api/markdown/context-node/editable-node",
+        json={
+            "markdown": current["source"].replace("Original node", "First update"),
+            "expected_revision": current["revision"],
+        },
+    )
+    stale = client.put(
+        "/api/markdown/context-node/editable-node",
+        json={
+            "markdown": current["source"].replace("Original node", "Stale update"),
+            "expected_revision": current["revision"],
+        },
+    )
+
+    assert first.status_code == 200
+    assert stale.status_code == 409
+    assert stale.json()["detail"]["code"] == "markdown_revision_conflict"
+    assert stale.json()["detail"]["current_revision"] == first.json()["revision"]
+
+
+def test_missing_and_failed_apply_are_structured(markdown_client, monkeypatch):
+    client, graph, _, _ = markdown_client
+    missing = client.get("/api/markdown/context-node/missing")
+    current = client.get("/api/markdown/context-node/editable-node").json()
+
+    def fail(_resource_id, _document):
+        raise RuntimeError("private storage failure")
+
+    monkeypatch.setattr(graph, "apply_node_markdown", fail)
+    failed = client.put(
+        "/api/markdown/context-node/editable-node",
+        json={
+            "markdown": current["source"].replace("Original node", "Will fail"),
+            "expected_revision": current["revision"],
+        },
+    )
+
+    assert missing.status_code == 404
+    assert missing.json()["detail"]["code"] == "markdown_resource_not_found"
+    assert failed.status_code == 500
+    assert failed.json()["detail"] == {
+        "code": "markdown_save_failed",
+        "message": "The edit could not be applied. The existing item was not changed.",
+    }
+    assert "private storage failure" not in failed.text
+
+
+def test_lists_memories_for_selection(markdown_client):
+    client, _, _, _ = markdown_client
+
+    response = client.get("/api/memories?skip=0&limit=100")
+
+    assert response.status_code == 200
+    assert response.json() == {
+        "items": [
+            {
+                "id": "editable-memory",
+                "type": "note",
+                "excerpt": "Original memory",
+                "updated_at": "2026-07-22T10:00:00+00:00",
+            }
+        ],
+        "total": 1,
+        "skip": 0,
+        "limit": 100,
+    }
+
+
+def test_memory_listing_is_atomic_with_concurrent_mutation(
+    markdown_client, monkeypatch
+):
+    client, _, memory, _ = markdown_client
+    snapshot_ready = threading.Event()
+    mutation_attempted = threading.Event()
+    mutation_finished = threading.Event()
+
+    def pause_after_key_snapshot(*, offset=0, limit=100, **_filters):
+        memory_ids = list(memory.memory_items.keys())[offset : offset + limit]
+        snapshot_ready.set()
+        assert mutation_attempted.wait(timeout=1)
+        mutation_finished.wait(timeout=0.25)
+
+        records = []
+        for memory_id in memory_ids:
+            # Match AgentMemory.list()'s second dictionary lookup after its
+            # key snapshot; this raises if a concurrent writer is not excluded.
+            memory.memory_items[memory_id]
+            record = memory.get_memory(memory_id)
+            if record is not None:
+                records.append(record)
+        return records
+
+    monkeypatch.setattr(memory, "list", pause_after_key_snapshot)
+
+    def delete_memory() -> None:
+        assert snapshot_ready.wait(timeout=1)
+        mutation_attempted.set()
+        memory.delete_memory("editable-memory")
+        mutation_finished.set()
+
+    writer = threading.Thread(target=delete_memory)
+    writer.start()
+    try:
+        response = client.get("/api/memories?skip=0&limit=100")
+    finally:
+        writer.join(timeout=2)
+
+    assert not writer.is_alive()
+    assert response.status_code == 200
+    assert [item["id"] for item in response.json()["items"]] == ["editable-memory"]
+    assert response.json()["total"] == 1
+
+
+def test_default_app_does_not_expose_agent_memory():
+    graph = ContextGraph(advanced_analytics=False)
+    with TestClient(create_app(session=GraphSession(graph))) as client:
+        info = client.get("/api/info")
+        memories = client.get("/api/memories")
+
+    assert info.json()["capabilities"]["agent_memory"] is False
+    assert memories.status_code == 503
+    assert memories.json()["detail"] == (
+        "AgentMemory is not configured for this Explorer instance."
+    )
+
+
+def test_slash_and_unicode_node_id_round_trip():
+    """GET and PUT work for node IDs containing '/' and Unicode characters.
+
+    Tests that:
+    - encodeURIComponent-style %2F encoding is transparent to the route
+    - the returned resource.id is the original unencoded string
+    - a successful PUT persists to the correct node
+    - the node ID is never rewritten by the apply path
+    """
+    node_id = "policy/\u6771\u4eac"  # "policy/東京"
+    graph = ContextGraph(advanced_analytics=False)
+    graph.add_node(node_id, "Policy", "Original slash body")
+    app = create_app(session=GraphSession(graph))
+
+    with TestClient(app) as client:
+        # GET via percent-encoded path (what encodeURIComponent produces)
+        import urllib.parse
+        encoded_id = urllib.parse.quote(node_id, safe="")
+        get_response = client.get(f"/api/markdown/context-node/{encoded_id}")
+
+        assert get_response.status_code == 200, (
+            f"GET returned {get_response.status_code}: {get_response.text}"
+        )
+        doc = get_response.json()
+        assert doc["resource"]["id"] == node_id, (
+            f"resource.id should be {node_id!r}, got {doc['resource']['id']!r}"
+        )
+        assert doc["resource"]["kind"] == "context-node"
+        assert doc["body"] == "Original slash body"
+        assert doc["revision"].startswith("sha256:")
+
+        # PUT an edit back via the same percent-encoded URL
+        updated_markdown = doc["source"].replace("Original slash body", "Updated slash body")
+        put_response = client.put(
+            f"/api/markdown/context-node/{encoded_id}",
+            json={
+                "markdown": updated_markdown,
+                "expected_revision": doc["revision"],
+            },
+        )
+
+        assert put_response.status_code == 200, (
+            f"PUT returned {put_response.status_code}: {put_response.text}"
+        )
+        result = put_response.json()
+        assert result["changed"] is True
+        assert result["body"] == "Updated slash body"
+        assert result["resource"]["id"] == node_id, (
+            f"PUT response resource.id should be {node_id!r}, got {result['resource']['id']!r}"
+        )
+
+        # Verify the live graph node was updated
+        assert graph.nodes[node_id].content == "Updated slash body"
+
+        # Also verify GET via literal slash URL works identically
+        get_literal = client.get(f"/api/markdown/context-node/{node_id}")
+        assert get_literal.status_code == 200
+        assert get_literal.json()["resource"]["id"] == node_id
+        assert get_literal.json()["body"] == "Updated slash body"
@@ -0,0 +1,115 @@
+from datetime import datetime
+
+import pytest
+
+from semantica.context.agent_memory import AgentMemory
+from semantica.context.context_graph import ContextGraph
+from semantica.explorer.markdown_resources import (
+    MarkdownResourceKind,
+    MarkdownResourceNotFound,
+    MarkdownResourceRef,
+    MarkdownResourceRegistry,
+    MarkdownRevisionConflict,
+    MarkdownSaveFailed,
+    document_revision,
+)
+
+
+def _resources():
+    graph = ContextGraph(advanced_analytics=False)
+    graph.add_node("node-1", "Note", "Original node")
+    graph.add_node("node-2", "Note", "Second node")
+    memory = AgentMemory()
+    memory.store(
+        "Original memory",
+        memory_id="mem-1",
+        timestamp=datetime.fromisoformat("2026-07-22T09:00:00+00:00"),
+        metadata={
+            "type": "note",
+            "updated_at": "2026-07-22T10:00:00+00:00",
+        },
+    )
+    return graph, memory, MarkdownResourceRegistry(graph, memory)
+
+
+def test_registry_routes_context_nodes_and_returns_canonical_revision():
+    graph, _, resources = _resources()
+    ref = MarkdownResourceRef(MarkdownResourceKind.CONTEXT_NODE, "node-1")
+
+    document = resources.read(ref)
+
+    assert document.body == "Original node"
+    assert document.revision == document_revision(document.source)
+    edited = document.source.replace("Original node", "Updated node")
+    result = resources.apply(ref, edited, document.revision)
+    assert result.changed is True
+    assert result.body == "Updated node"
+    assert graph.nodes["node-1"].content == "Updated node"
+
+
+def test_registry_routes_agent_memory_without_touching_context_graph():
+    graph, memory, resources = _resources()
+    ref = MarkdownResourceRef(MarkdownResourceKind.AGENT_MEMORY, "mem-1")
+    before_nodes = dict(graph.nodes)
+    document = resources.read(ref)
+
+    result = resources.apply(
+        ref,
+        document.source.replace("Original memory", "Updated memory"),
+        document.revision,
+    )
+
+    assert result.body == "Updated memory"
+    assert memory.get("mem-1")["content"] == "Updated memory"
+    assert graph.nodes == before_nodes
+
+
+def test_registry_rejects_stale_revision_and_preserves_current_state():
+    graph, _, resources = _resources()
+    ref = MarkdownResourceRef(MarkdownResourceKind.CONTEXT_NODE, "node-1")
+    first = resources.read(ref)
+    resources.apply(
+        ref,
+        first.source.replace("Original node", "First update"),
+        first.revision,
+    )
+
+    with pytest.raises(MarkdownRevisionConflict) as error:
+        resources.apply(
+            ref,
+            first.source.replace("Original node", "Stale update"),
+            first.revision,
+        )
+
+    assert error.value.current_revision == resources.read(ref).revision
+    assert graph.nodes["node-1"].content == "First update"
+
+
+def test_registry_missing_adapter_and_resources_are_structured():
+    graph, _, _ = _resources()
+    resources = MarkdownResourceRegistry(graph)
+
+    with pytest.raises(MarkdownResourceNotFound, match="not available"):
+        resources.read(MarkdownResourceRef(MarkdownResourceKind.AGENT_MEMORY, "mem-1"))
+    with pytest.raises(MarkdownResourceNotFound, match="missing"):
+        resources.read(
+            MarkdownResourceRef(MarkdownResourceKind.CONTEXT_NODE, "missing")
+        )
+
+
+def test_registry_does_not_expose_adapter_failure_details(monkeypatch):
+    graph, _, resources = _resources()
+    ref = MarkdownResourceRef(MarkdownResourceKind.CONTEXT_NODE, "node-1")
+    document = resources.read(ref)
+
+    def fail(_resource_id, _source):
+        raise RuntimeError("private backend failure")
+
+    monkeypatch.setattr(graph, "apply_node_markdown", fail)
+    with pytest.raises(MarkdownSaveFailed) as error:
+        resources.apply(ref, document.source + "changed", document.revision)
+
+    assert error.value.message == (
+        "The edit could not be applied. The existing item was not changed."
+    )
+    assert "private backend failure" not in error.value.message
@@ -0,0 +1,308 @@
+from types import SimpleNamespace
+
+import pytest
+
+from semantica.ontology import (
+    OntologyEngine,
+    OntologyQualityGate,
+    QualitySeverity,
+    ontology_quality_check,
+)
+
+
+def _ontology():
+    return {
+        "classes": [
+            {"name": "Person", "uri": "https://example.org/Person"},
+            {"name": "Company", "uri": "https://example.org/Company"},
+        ],
+        "properties": [
+            {
+                "name": "worksFor",
+                "type": "object",
+                "domain": ["Person"],
+                "range": ["Company"],
+            },
+            {
+                "name": "name",
+                "type": "data",
+                "domain": ["Person"],
+                "range": "string",
+            },
+        ],
+    }
+
+
+def test_quality_gate_reports_a_healthy_ontology():
+    report = ontology_quality_check(_ontology())
+
+    assert report.passed
+    assert report.metrics["coverage"] == 1.0
+    assert report.error_count == 0
+    assert report.to_dict()["stats"]["properties"] == 2
+
+
+def test_quality_gate_finds_schema_and_endpoint_problems():
+    ontology = {
+        "classes": [{"name": "Person"}, {"name": "Unused"}],
+        "properties": [
+            {
+                "name": "worksFor",
+                "type": "object",
+                "domain": ["Person"],
+                "range": ["MissingCompany"],
+            },
+            {"name": "unattached", "type": "data"},
+        ],
+    }
+    graph = {
+        "entities": [{"id": "p1", "type": "Person"}],
+        "relationships": [
+            {"source_id": "p1", "target_id": "missing", "type": "worksFor"}
+        ],
+    }
+
+    report = OntologyQualityGate().check(ontology, graph_data=graph)
+    codes = {issue.code for issue in report.issues}
+
+    assert not report.passed
+    assert "UNKNOWN_RANGE" in codes
+    assert "UNRESOLVED_RELATIONSHIP_ENDPOINT" in codes
+    assert "ORPHAN_CLASS" in codes
+    assert "MISSING_RANGE" in codes
+    assert any(issue.severity == QualitySeverity.WARNING for issue in report.issues)
+
+
+def test_quality_gate_supports_thresholds_and_legacy_endpoint_keys():
+    ontology = {
+        "classes": [{"name": "Person"}],
+        "properties": [
+            {
+                "name": "name",
+                "type": "data",
+                "domain": "Person",
+                "range": "xsd:string",
+            }
+        ],
+    }
+    graph = {
+        "entities": [{"entity_id": "p1", "type": "Person", "name": "Alice"}],
+        "relationships": [{"source": "p1", "target": "p1", "type": "knows"}],
+    }
+
+    report = ontology_quality_check(
+        ontology,
+        graph_data=graph,
+        thresholds={"min_coverage": 1.0},
+    )
+
+    assert report.passed
+    assert report.threshold_failures == []
+    assert report.stats["relationships"] == 1
+
+
+def test_quality_gate_can_fail_on_warnings():
+    report = ontology_quality_check(
+        {"classes": [{"name": "Person"}], "properties": []},
+        fail_on_warnings=True,
+    )
+
+    assert not report.passed
+    assert "fail_on_warnings" in report.threshold_failures
+
+
+def test_engine_exposes_quality_check():
+    report = OntologyEngine().quality_check(_ontology())
+
+    assert report.passed
+
+
+def test_quality_gate_accepts_canonical_context_graph_entities_without_mutation():
+    graph = {
+        "entities": [
+            {"id": "p1", "text": "Alice", "type": "Person"},
+            {"id": "c1", "text": "Acme", "type": "Company"},
+        ],
+        "relationships": [{"source_id": "p1", "target_id": "c1", "type": "worksFor"}],
+    }
+
+    report = OntologyQualityGate().check(_ontology(), graph_data=graph)
+
+    assert report.passed
+    assert "name" not in graph["entities"][0]
+
+
+def test_quality_gate_returns_structured_findings_for_malformed_graph_members():
+    report = OntologyQualityGate().check(
+        _ontology(), graph_data={"entities": [None], "relationships": [None]}
+    )
+    codes = {issue.code for issue in report.issues}
+
+    assert not report.passed
+    assert {"INVALID_ENTITY", "INVALID_RELATIONSHIP"} <= codes
+
+
+def test_quality_gate_returns_structured_finding_for_invalid_graph_containers():
+    report = OntologyQualityGate().check(
+        _ontology(), graph_data={"entities": None, "relationships": []}
+    )
+
+    assert not report.passed
+    assert any(issue.code == "INVALID_GRAPH" for issue in report.issues)
+
+
+def test_quality_gate_does_not_undercount_identical_dangling_edges():
+    graph = {
+        "entities": [{"id": "p1", "name": "Alice", "type": "Person"}],
+        "relationships": [
+            {"source": "p1", "target": "missing", "type": "knows"},
+            {"source": "p1", "target": "missing", "type": "knows"},
+        ],
+    }
+
+    report = OntologyQualityGate().check(
+        _ontology(), graph_data=graph, thresholds={"max_errors": 1}
+    )
+
+    endpoint_errors = [
+        issue
+        for issue in report.issues
+        if issue.code == "UNRESOLVED_RELATIONSHIP_ENDPOINT"
+    ]
+    assert len(endpoint_errors) == 2
+    assert not report.passed
+    assert "max_errors" in report.threshold_failures
+
+
+def test_quality_gate_indexes_all_class_identifier_aliases():
+    ontology = {
+        "classes": [{"name": "Person", "uri": "https://example.org/PersonType"}],
+        "properties": [
+            {
+                "name": "name",
+                "type": "data",
+                "domain": "https://example.org/PersonType",
+                "range": "string",
+            }
+        ],
+    }
+
+    report = OntologyQualityGate().check(ontology)
+    codes = {issue.code for issue in report.issues}
+
+    assert "UNKNOWN_DOMAIN" not in codes
+
+
+def test_quality_gate_recognizes_subclass_of_hierarchies():
+    ontology = {
+        "classes": [
+            {"name": "Person"},
+            {"name": "Employee", "subclassOf": "Person"},
+        ],
+        "properties": [
+            {
+                "name": "name",
+                "type": "data",
+                "domain": "Employee",
+                "range": "string",
+            }
+        ],
+    }
+
+    report = OntologyQualityGate().check(ontology)
+
+    assert not any(issue.code == "ORPHAN_CLASS" for issue in report.issues)
+
+
+def test_quality_gate_keeps_competency_evaluation_safe_for_malformed_members():
+    ontology = {
+        "classes": [None, {"uri": "https://example.org/Person"}],
+        "properties": [
+            None,
+            {
+                "uri": "https://example.org/name",
+                "type": "data",
+                "domain": "Person",
+                "range": "string",
+            },
+        ],
+    }
+
+    report = OntologyQualityGate().check(
+        ontology, competency_questions=["What is a person's name?"]
+    )
+
+    assert isinstance(report.metrics["competency_question_coverage"], float)
+
+
+def test_quality_gate_reports_validator_warnings_and_can_fail_on_them():
+    class WarningValidator:
+        def validate(self, ontology):
+            return SimpleNamespace(valid=True, errors=[], warnings=["review me"])
+
+    report = OntologyQualityGate(
+        validator=WarningValidator(), fail_on_warnings=True
+    ).check(_ontology())
+
+    assert report.warning_count == 1
+    assert "VALIDATOR_WARNING" in {issue.code for issue in report.issues}
+    assert "fail_on_warnings" in report.threshold_failures
+
+
+def test_quality_gate_isolates_competency_questions_between_checks():
+    engine = OntologyEngine()
+
+    first = engine.quality_check(_ontology(), competency_questions=["Who is a Person?"])
+    second = engine.quality_check(
+        _ontology(), competency_questions=["What is a location?"]
+    )
+
+    assert first.metrics["competency_question_coverage"] == 1.0
+    assert second.metrics["competency_question_coverage"] == 0.0
+
+
+def test_quality_gate_treats_null_property_type_as_missing():
+    ontology = {
+        "classes": [{"name": "Person"}],
+        "properties": [
+            {
+                "name": "value",
+                "type": None,
+                "domain": "Person",
+                "range": "string",
+            }
+        ],
+    }
+
+    report = OntologyQualityGate().check(ontology)
+
+    assert "MISSING_PROPERTY_TYPE" in {issue.code for issue in report.issues}
+    assert report.error_count >= 1
+
+
+def test_quality_gate_normalizes_orphan_ids_for_deterministic_reports():
+    graph = {
+        "entities": [
+            {"id": "p2", "name": "Bob", "type": "Person"},
+            {"id": "p1", "name": "Alice", "type": "Person"},
+        ],
+        "relationships": [],
+    }
+
+    report = OntologyQualityGate().check(_ontology(), graph_data=graph)
+    orphan = next(issue for issue in report.issues if issue.code == "ORPHAN_ENTITY")
+
+    assert orphan.details["ids"] == ["p1", "p2"]
+
+
+@pytest.mark.parametrize(
+    "thresholds",
+    [
+        {"min_coverage": float("nan")},
+        {"max_errors": float("inf")},
+        {"max_warnings": float("-inf")},
+    ],
+)
+def test_quality_gate_rejects_non_finite_thresholds(thresholds):
+    with pytest.raises(ValueError, match="finite"):
+        OntologyQualityGate().check(_ontology(), thresholds=thresholds)
Author	SHA1	Message	Date
Yuang PengandSameer Kadam	b92fed2d0a	feat(explorer): add Markdown editor write path (#1349 ) * feat(explorer): add Markdown editing write path * fix(explorer): address markdown editor review findings * feat(explorer): add Markdown editor write path --------- Co-authored-by: Sameer Kadam <sskadam6305@gmail.com>	2026-09-04 01:38:17 +05:30
Mohd Kaif	8910d2c949	docs(ontology): document quality gate threshold semantics (#1450 ) * docs(ontology): document quality gate threshold semantics The Ontology Quality Gate section (#1397) showed a thresholds={...} example but never explained what min_coverage, max_errors, max_warnings, or fail_on_warnings actually mean, or that fail_on_warnings is a separate parameter rather than a thresholds key. Add a concise defaults/semantics table, verified against OntologyQualityGate.DEFAULT_THRESHOLDS and __init__ in quality_gate.py. * docs(ontology): explain thresholds as prose instead of a table A four-row table was heavier than this needed; each threshold's meaning reads faster as two connected sentences.	2026-09-03 23:40:00 +05:30
Mohd Kaif	4c6eab632f	Merge pull request #1397 from T1mn/codex/ontology-quality-gate feat(ontology): add a CI-friendly quality gate	2026-09-03 23:20:39 +05:30
Mohd Kaif	103ab04970	Merge branch 'main' into codex/ontology-quality-gate	2026-09-03 23:05:32 +05:30
Zohaib Hassnain	88a57b39f9	docs: replace retired claude-sonnet-4-20250514 model id (#1449 ) * docs: replace retired claude-sonnet-4-20250514 model id * docs(llms): sweep retired model as qodo found	2026-09-03 22:29:50 +05:00
Guofang.Tang	75bcb64681	Merge branch 'main' into codex/ontology-quality-gate	2026-09-04 00:34:54 +08:00
T1mn	8778e6a837	fix(ontology): address follow-up quality gate findings	2026-09-03 19:57:04 +08:00
Mohd Kaif	01352d5fd5	Merge branch 'main' into codex/ontology-quality-gate	2026-09-03 15:18:43 +05:30
T1mn	6a55b8c3ee	fix(ontology): address quality gate review findings	2026-09-03 10:04:46 +08:00
T1mn	471cbe7711	feat(ontology): add deterministic quality gate	2026-09-03 02:00:57 +08:00