semantica/CHANGELOG.md

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

---

[Unreleased]

Added

• Global default persistent storage for ProvenanceManager, plus a working provenance CLI (#795, #802) by @Sameer6305 and @KaifAhmad1

  • Every ingestion/processing module (kg_provenance.py, pipeline_provenance.py, and 20+ other call sites) instantiated its own ProvenanceManager() with no storage_path, so all of them silently fell back to InMemoryStorage and the SQLite audit trail was never actually written. ProvenanceManager.set_default_storage_path(path) now sets a class-level default that every no-arg instantiation picks up, and Semantica.__init__ wires config.provenance.storage_path into it automatically during orchestrator init
  • Added the thread-safe default_storage_path(path) context manager (semantica.provenance.default_storage_path) for test isolation — it stacks nested overrides and guarantees restoration of the previous default on exit, even on exception, so tests can't leak global state into each other
  • Fixed ProvenanceManager.__init__ raising TypeError on the CLI's config= kwarg, and implemented the four methods the CLI already called but that didn't exist on the class: lineage(), audit_log(), export_prov() (W3C PROV-O turtle/ntriples/jsonld via rdflib), and check() — unblocking semantica provenance lineage|audit|export|check end-to-end
  • Follow-up review fixes: track_entity no longer aliases a caller-supplied used_entities list (it copied the reference and later mutated it in place via .append(), which could corrupt a list the caller still held); removed dead fallback branches in orchestrator.py/manager.py left over from not realizing Config.get() already resolves dotted paths; added a --dry-run option to provenance audit to match provenance export (previously only the global --dry-run flag worked, not a local one); and provenance check --strict no longer prints a green "✓" success line immediately before failing — a failing check now renders as a warning before the ClickException is raised

• Markdown round-trip export/import for AgentMemory (#765, #786) by @SaurabhScripts and @Sameer6305

  • AgentMemory.export(format="markdown") and import_data(format="markdown") add a human-editable, diff-friendly alternative to the existing JSON/dict serialization: one Markdown file per memory item, with id, created_at, updated_at, and type/kind in required YAML frontmatter and the memory content as the Markdown body
  • Exporting without a destination returns a single memory as a Markdown string; exporting a set requires a destination directory and writes one stable, content-hashed filename per memory ID, so re-exporting an unchanged set is byte-for-byte idempotent
  • Importing upserts by ID: unknown IDs create new memories, known IDs replace them atomically (local state and vector store are only mutated after the whole batch validates cleanly), and unchanged re-imports are a deterministic no-op
  • Malformed frontmatter, duplicate IDs within an import batch, and duplicate YAML keys are all rejected before any memory is mutated, with actionable error messages
  • Export refuses to overwrite symbolic links and replaces files atomically; import safely compares timezone-aware and timezone-naive timestamps so retention, recency sorting, and date filters stay correct across both
  • Entities and relationships round-trip as memory-local provenance only — Markdown import intentionally does not write into ContextGraph, matching the MVP scope agreed on in #765
  • Documented the file contract and workflow in docs/reference/context.md; 43 new tests in tests/context/test_agent_memory_markdown.py cover round-trip losslessness, idempotency, validation errors, rollback on failure, and vector-store sync ordering

Fixed

• Neptune cookbook CloudFormation stack exposed the database port to the entire internet and had no network audit trail (code scanning alert #28, #26, #27, AC_AWS_0276/AC_AWS_0369/AC_AWS_0148) by @KaifAhmad1

  • cookbook/introduction/neptune-setup.yaml's security group let anyone on 0.0.0.0/0 reach the Neptune Bolt/OpenCypher port (8182); it now requires a ClientCidr parameter (CIDR-validated, no default) so the stack can't be created without the deployer explicitly scoping access to their own IP or VPN/office range
  • Added AWS::EC2::FlowLog plus a dedicated CloudWatch Logs group and IAM role so all traffic in the stack's VPC is now logged
  • Left the account-wide IAM password policy check (AC_AWS_0148) unimplemented as a stack resource on purpose: AWS::IAM::AccountPasswordPolicy is an account singleton, and wiring it into a disposable per-learner tutorial stack would mean creating or deleting this stack also mutates or removes the account's real password policy — suppressed with a documented ts:skip=AC_AWS_0148 explaining why, rather than "fixed"
  • Updated 21_Amazon_Neptune_Store.ipynb's aws cloudformation create-stack instructions, prerequisites, and cost table to match the new required ClientCidr parameter and flow-log line item

• Follow-up to the knowledge-explorer Helm chart default-namespace/seccomp scanner findings reopening (code scanning alert #846, #847, #848, #68, #63, CKV_K8S_21/AC_K8S_0086/AC_K8S_0080) by @KaifAhmad1

  • The checkov.io/skip1 metadata annotation added previously (see the CKV_K8S_21 entry below) evidently isn't being honored by the Microsoft Defender for DevOps scan — the same finding reopened under new alert numbers on the current main. Added the more standard # checkov:skip=CKV_K8S_21 and # ts:skip=AC_K8S_0086 inline comments at the top of templates/deployment.yaml, templates/service.yaml, and templates/configmap.yaml as a second suppression path (matching the convention already used in deploy/gcp/cloudrun-service.yaml), plus # ts:skip=AC_K8S_0080 on templates/deployment.yaml for the seccomp finding, which trips for the same root cause: terrascan's static template scan never resolves {{ toYaml .Values.podSecurityContext }}, even though values.yaml sets seccompProfile.type: RuntimeDefault correctly
  • Confirmed the deploy/kubernetes/* (non-Helm) manifests already had TLS and seccomp configured correctly, so no code change was needed there for the corresponding alerts (#61 and the non-Helm seccomp finding) — expected to close on the next scan
  • Documented both suppression mechanisms and the reasoning in .checkov.yaml
  • Residual risk: this environment could not run checkov/terrascan locally to confirm the inline comments are actually honored during a Helm-rendered scan; if the alerts are still open after the next scan, the reliable fallback is splitting the CI checkov/terrascan invocation so deploy/helm/ is scanned with these specific checks excluded via --skip-check instead of relying on in-file suppression

• react-hooks/set-state-in-effect cascading renders across 12 Explorer workspace files (#769, #796) by @Sameer6305 and @KaifAhmad1

  • Replaced synchronous setState calls inside useEffect bodies with React's recommended "adjust state during render" pattern (if (x !== prevX) { setPrevX(x); ...setState... }) across OntologyWorkspace, ManageWorkspace, LineageWorkspace, and GraphWorkspace, and inlined async data-fetching effects with ignore flags to prevent race conditions and stale writes after unmount
  • Fixed a regression the inlining itself introduced: AlignmentsTab.tsx, KGOverviewTab.tsx, OntologyManager.tsx, and VersionsTab.tsx each duplicated their existing fetch callback (reload / fetchOverview / fetchRegistry / loadVersions+loadProposals) into a second, inline copy for the mount effect, and the copy silently dropped the setError/flashMsg calls the original had — re-introducing, on the very first page load, the exact error-swallowing behavior that #767/#790 had already fixed for these same files. The inline copies now mirror the original's error handling (including 207 partial-success messages) exactly
  • Fixed LineageDiagram.tsx only clearing the previously-rendered nodes/edges when the new activeId was falsy instead of on every id change, so switching directly between two lineage views briefly kept showing the previous view's stale diagram instead of clearing before the new fetch resolved
  • GraphWorkspace.tsx and GraphLoadingOverlay.tsx still have unrelated react-hooks/set-state-in-effect violations outside this PR's 12-file scope (confirmed via npx eslint .); left as follow-up work rather than expanding this PR further

• Checkov flagged the knowledge-explorer Helm chart for using the default Kubernetes namespace (code scanning alert #779, #778, #777, CKV_K8S_21) by @KaifAhmad1

  • templates/service.yaml, templates/deployment.yaml, and templates/configmap.yaml all already set metadata.namespace to {{ .Release.Namespace }}, which is only bound at helm install/helm template time; Checkov's helm framework renders the chart without a namespace override, so it always resolves to default and trips CKV_K8S_21 even though the chart is namespace-agnostic by design
  • Added a checkov.io/skip1: CKV_K8S_21 metadata annotation to each of the three files to suppress the scanner artifact false-positive properly in Helm templates, and documented the reasoning in .checkov.yaml

• No React error boundaries around lazy-loaded Explorer workspaces — a single render error crashed the whole app (#768, #794) by @Sameer6305

  • Added an ErrorBoundary class component (explorer/src/ErrorBoundary.tsx) and wrapped each lazy-loaded workspace's <Suspense> block in App.tsx with it, keyed on the active sub-view so navigating away from and back to a crashed tab remounts it cleanly
  • Failed retries are capped at 3 before the fallback UI switches from "Try Again" to a "Reload Application" dead-end, preventing infinite retry loops on deterministic crashes; raw error/stack details are logged via console.error only and never rendered into the fallback UI
  • Fixed the retry counter so it resets after a retry actually succeeds and stays error-free for a few seconds, instead of never resetting (which could permanently exhaust the retry budget on unrelated, individually-recoverable transient errors) or resetting on the very next commit (which could fire prematurely while Suspense was still showing its fallback)

• Explorer frontend workspaces silently swallowed network/server errors (#767, #790) by @Sameer6305

  • ShaclStudio.tsx, VersionsTab.tsx, SKOSVocabularyManager.tsx, EntityResolutionTab.tsx, LineageDiagram.tsx, DecisionWorkspace.tsx, KGOverviewTab.tsx, OntologyManager.tsx, OntologySearch.tsx, ReasoningWorkspace.tsx, and SparqlWorkspace.tsx now render a visible error banner instead of only console.error()-ing failed fetches
  • Added explicit response.status === 207 (Multi-Status) handling across these workspaces so partial backend failures surface a warning instead of reading as a full success (response.ok is true for all 2xx codes, including 207)
  • Added defensive JSON parsing so an unexpected non-JSON (e.g. HTML 500) response body no longer crashes the app with SyntaxError: Unexpected token < in JSON
  • Fixed KGOverviewTab.tsx dropping the /api/graph/nodes partial-success warning whenever /api/graph/stats also returned 207 — both warnings are now shown (appended) instead of one being silently discarded
  • Fixed HealthTab.tsx's registry load still using a bare .catch(() => {}) that swallowed errors identically to the pattern fixed elsewhere in this same folder; failures now populate the existing error banner
  • Fixed AlignmentsTab.tsx's reload() using Promise.allSettled but never handling the "rejected" branches for the registry/alignments fetches, so both failures previously vanished with no error surfaced and no logging

• tests/explorer/test_explorer_api.py failed with TypeError: Client.__init__() got an unexpected keyword argument 'app' on current httpx (#788, #789) by @Sameer6305

  • httpx>=0.28.0 removed the app= kwarg that Starlette's TestClient relies on to wrap a FastAPI app for testing; httpx wasn't pinned anywhere in pyproject.toml, so different environments could independently resolve an incompatible transitive version and hit the same break
  • Added an explicit httpx<0.28.0 constraint to the main [project.dependencies] array (not just a dev extra), so it applies globally across production, dev, and CI installs
  • Without the pin, the full test suite fails to even complete collection (fails immediately on tests/explorer/test_vocabulary.py with the same TestClient error); with it, tests/explorer/test_explorer_api.py goes from 7 failed/12 passed/58 errors to 77 passed, 0 errors

• Explorer backend routes returned HTTP 200 with error/empty bodies on failure, defeating frontend error handling (#770, #787) by @Sameer6305 and @KaifAhmad1

  • GET /api/temporal/patterns now raises HTTPException(500) on a genuine computation failure instead of silently returning an empty-but-valid TemporalPatternResponse; the ImportError fallback (optional kg extra not installed) is unchanged and still degrades gracefully to an empty list
  • POST /api/ontology/create now raises HTTPException(500) when ontology generation fails in either the sample_data or schema_text mode, instead of silently falling back to a partial/minimal ontology with a misleading nodes_added count
  • GET /api/analytics sets response.status_code = 207 (Multi-Status) when some, but not all, of the requested metrics fail, and raises HTTPException(500) when every requested metric fails — a plain 2xx (including 207) reads as success to callers that only check response.ok, so an all-failed request now surfaces as a hard error rather than a body full of {"error": ...}
  • Added regression tests covering all three failure paths (test_patterns_failure_returns_500, test_analytics_partial_failure_returns_207, test_analytics_total_failure_returns_500, and two TestOntologyCreateFailures cases)

[0.6.0] - 2026-07-21

Added

• Named-graph support for JenaStore via Dataset migration (#756, #757) by @Sameer6305 and @KaifAhmad1

  • JenaStore now backs onto rdflib.Dataset(default_union=False) instead of rdflib.Graph, closing #756 and fully closing out the #754/#756 cross-backend named-graph parity effort across Blazegraph, RDF4J, and Jena
  • default_union=False is explicitly set so existing execute_sparql()/get_triplets() calls that don't pass graph= keep seeing only the default graph, not a union across all named graphs
  • add_triplets() accepts a graph= option: when supplied, triples are written to that named graph (4-tuple add via Dataset.graph(uri)); when omitted, behavior is unchanged (3-tuple add routes to the default graph)
  • Fixed a pre-existing bug where the remote-endpoint path instantiated the read-only rdflib SPARQLStore instead of SPARQLUpdateStore, so every add_triplets() call against a remote Fuseki endpoint silently failed (TypeError swallowed, success=True/added=0 returned); also fixed a constructor bug where self.endpoint was always None regardless of how JenaStore was called, making the remote path unreachable in practice
  • serialize() now logs a warning instead of silently dropping named-graph content when the requested format (turtle, xml, n3, …) can only serialize the default graph; use format="trig" or format="nquads" to include all graphs
  • create_model()'s triplet_count now documented as counting across all graphs (default + named), not just the default graph, matching the Dataset-wide semantics
  • delete_triplet() remains scoped to the default graph only (named-graph parity for delete is an explicit follow-up, matching the maintainer's scoping of this migration to add_triplets); the removal is passed self.graph.default_graph explicitly as its context, since Dataset.remove() on a bare 3-tuple resolves to a wildcard context internally and would otherwise delete matching triples out of every named graph too — a follow-up fix to the initial PR #757 for a bug that had no test coverage
  • 9 new tests covering Dataset construction, default_union=False confirmation, named-graph write isolation, serialize() warning behavior, and delete_triplet()'s default-graph scoping

• SPARQL CONSTRUCT query templates (#752, #322, #755, #754) by @Sameer6305

  • Added parameterized, injection-safe CONSTRUCT templates (ConstructTemplate, ParameterDescriptor, ConstructTemplateRegistry)
  • Extended CONSTRUCT execution support from Blazegraph-only to the RDF4J and Jena backends (#755), closing #754
    • RDF4JStore.execute_sparql gains a CONSTRUCT-aware path (Accept: text/turtle, rdflib Turtle parsing, the same (s, p, o, metadata) 4-tuple contract) and named-graph writes via RDF4J's REST context parameter
    • JenaStore.execute_sparql gains the equivalent CONSTRUCT-aware path over its in-process rdflib.Graph
    • _CONSTRUCT_QUERY_RE moved to sparql_escaping.py as a shared, backend-agnostic constant used by all three backends
  • Added pipeline integration via the construct_template step type

• Databricks Connector (Unity Catalog + Delta Lake ingestion) (#747) by @KaifAhmad1

  • Added DatabricksIngestor (semantica/ingest/databricks_ingestor.py), mirroring SnowflakeIngestor's structure and public API shape: a DatabricksConnector connection handler, a DatabricksData dataclass, and an optional-import guard for databricks-sdk/databricks-sql-connector
  • Supports personal access token and OAuth M2M (service principal client_id/client_secret) authentication, configurable via constructor args or DATABRICKS_* environment variables
  • ingest_table()/ingest_query() run against a SQL warehouse or cluster via databricks-sql-connector, with where/order_by/limit/offset support and the same identifier-escaping and unsafe-ORDER BY rejection as SnowflakeIngestor; each call closes the SQL connection it opened unless one is already open (e.g. via the with DatabricksIngestor(...) context manager), which reuses and closes it exactly once instead of leaking a second connection per call
  • get_table_schema(), list_catalogs(), list_schemas(), and list_tables() introspect Unity Catalog via databricks-sdk's WorkspaceClient, validating both catalog and schema are resolved before calling the SDK; get_table_lineage() calls Unity Catalog's table-lineage REST API for upstream/downstream Table --DEPENDS_ON--> Table dependencies, plus an opt-in include_column_lineage=True that resolves per-column lineage via the column-lineage API
  • export_as_documents() converts ingested rows into Semantica document dicts for KG construction, matching SnowflakeIngestor.export_as_documents()'s shape
  • Registered as a lazy export in semantica.ingest (DatabricksIngestor, DatabricksData, DatabricksConnector) and as the db-databricks optional extra (pip install "semantica[db-databricks]") in pyproject.toml, included in db-all
  • New docs/integrations/databricks.md page modeled on docs/integrations/snowflake.md, plus a DatabricksIngestor section and table row in docs/reference/ingest.md and cross-links between the two integration pages
  • 35 unit tests in tests/test_databricks_ingestor.py covering both auth methods, table/query ingestion, connection lifecycle (including reuse under the context manager), pagination, unsafe ORDER BY rejection, catalog/schema validation, schema/catalog/table listing, table and column lineage, document export, and the missing-dependency error path, closing #747

• SQLite Vector Store Backend (sqlite-vec) (#726) by @Luffy2208 and @KaifAhmad1

  • Added SQLiteVecStore (semantica/vector_store/sqlite_vec_store.py), a disk-backed local vector store using the sqlite-vec extension's vec0 virtual tables, closing #240
  • Supports Cosine and L2 distance metrics, dynamic JSON metadata filtering, read-only mode, and an in-memory (:memory:) mode
  • Registered as the "sqlite" backend in VectorStore.SUPPORTED_BACKENDS, with db_path/sqlite_path config and a VECTOR_STORE_SQLITE_PATH environment variable
  • Batched add/delete/get and executemany-based update to avoid per-row round trips; optional use_wal=True enables journal_mode=WAL + synchronous=NORMAL for improved write concurrency
  • Lazy-imports sqlite-vec so the dependency stays fully optional (pip install semantica[vectorstore-sqlite]); table names and metadata filter keys are validated against a strict identifier pattern before SQL interpolation
  • Fixes VectorStore.update_vectors/delete_vectors to delegate to the active backend store instead of only mutating in-memory state, correcting existing behavior for all non-inmemory backends
  • 25 unit and integration tests in tests/vector_store/test_sqlite_vec_store.py covering init, add, search, get, update, delete, read-only mode, and stats

Fixed

• kg.ProvenanceTracker compatibility wrapper out of sync with ProvenanceManager, causing 9 pre-existing test failures (#744, #751) by @Sameer6305 and @KaifAhmad1

  • kg.ProvenanceTracker was a standalone in-memory implementation that never delegated to the unified ProvenanceManager backend; its own test suite asserted the existence of get_lineage, track_relationship, track_entities_batch, get_provenance, and _use_unified, none of which were ever implemented, plus a stale get_all_sources() assertion expecting "timestamp" instead of the actual "recorded_at" key
  • Rather than completing the abandoned compatibility layer, kg.ProvenanceTracker and its remaining supported methods (track_entity, get_all_sources, query_recorded_between, revision_history, export_audit_log) now emit DeprecationWarnings pointing callers to semantica.provenance.ProvenanceManager
  • Removed/rewrote the 9 tests that only exercised the never-implemented compatibility methods to instead verify the observable behavior of the still-supported API, and corrected the stale get_all_sources() assertion
  • Added the previously-missing docs/migration/kg-provenance-tracker.md migration guide referenced by every new deprecation warning, with a method-mapping table to ProvenanceManager and a before/after example, closing #744

• ProvenanceManager.track_entity silently overrides an explicit parent_entity_id/derived_from on re-track (#742) by @Sameer6305

  • track_entity() resolved parent_id via a documented precedence chain (parent_entity_id kwarg > metadata["derived_from"] > source-as-known-entity-id fallback), but the history-preservation block that runs afterward unconditionally overwrote that resolved value with an auto-generated f"{entity_id}:v:{existing.last_updated}" history pointer whenever the entity was being re-tracked, discarding whatever parent the caller had just explicitly supplied with no warning
  • track_entity() now records whether the precedence chain already resolved an explicit parent (parent_entity_id kwarg, metadata["derived_from"], or the source-as-known-entity-id fallback) before the history block runs, and only falls back to the auto-generated history pointer when the caller supplied no explicit parent on that call
  • The archived history entry for the previous version is still kept reachable in get_lineage() via used_entities (BFS-traversed by InMemoryStorage.trace_lineage()) even when an explicit parent is supplied, so re-tracking with a new parent no longer orphans the prior version from the lineage chain; when no explicit parent is supplied, used_entities is left alone since parent_entity_id already points at the same history id, avoiding a duplicate self-reference
  • Added test_retrack_with_explicit_parent_overrides_history_link, test_retrack_without_explicit_parent_still_uses_history_link, test_retrack_with_derived_from_overrides_history_link, and test_retrack_history_reachable_via_used_entities regression tests, closing #742

• ProvenanceManager.get_lineage does not link entities that share a source URL (#735) by @KaifAhmad1

  • track_entity()'s only auto-linking logic looked up source as if it were an existing entity's entity_id, so passing the same real URL/DOI as source for two conceptually linked entities (e.g. a document and a decision derived from it) never produced a parent link, leaving get_lineage() returning a chain of length 1
  • metadata["derived_from"] was preserved and echoed back in the output JSON but was never consulted by any linking or traversal code, so the caller's explicit relationship was silently inert
  • track_entity() now treats metadata["derived_from"] as an explicit parent link (unless parent_entity_id was already passed directly), so InMemoryStorage.trace_lineage()'s existing BFS over parent_entity_id picks it up for free
  • metadata["derived_from"] is now recognized on any collections.abc.Mapping, not just a concrete dict, so e.g. types.MappingProxyType metadata still creates the parent link
  • get_lineage()'s metadata aggregation now applies the queried entity's own metadata last so it wins over ancestor metadata on conflicting keys, matching the documented "most recent entry's metadata takes precedence" behavior — previously trace_lineage()'s BFS order caused ancestor metadata (now reachable via derived_from chains) to silently overwrite the queried entity's own values
  • Added 9 regression/edge-case tests in tests/provenance/test_manager.py covering the happy path, explicit parent_entity_id precedence over derived_from, precedence over the source-as-known-entity-id fallback, a derived_from pointing at a never-tracked entity, non-string/empty-string derived_from values being ignored, a self-referencing derived_from not hanging traversal, multi-hop derived_from chains, metadata precedence between a queried entity and its ancestors, and non-dict Mapping metadata, closing #735

• Reasoner.add_rule had no deduplication, doubling rules and silently emptying forward_chain() on rerun (#732) by @KaifAhmad1

  • add_rule() unconditionally appended to self.rules, so re-running the same setup code on an existing Reasoner instance (e.g. re-executing a Jupyter cell) duplicated every rule; since forward_chain() only records a conclusion if it isn't already in self.facts, the second run's duplicated rules matched but produced no new results, with no error or warning
  • add_rule() now compares an incoming rule's rule_type, conditions, and conclusion against existing rules and returns the existing Rule instead of appending a duplicate, keeping repeated add_rule() calls with the same definition idempotent
  • Added test_add_rule_deduplicates_identical_rule, test_add_rule_deduplication_is_idempotent_across_forward_chain, and test_add_rule_does_not_dedupe_distinct_rules regression tests

• InferenceResult.premises always empty from forward_chain/backward_chain (#739) by @Sameer6305

  • _match_rule() discarded matched facts and returned only instantiated conclusions, so ExplanationGenerator always produced empty premises lists regardless of which facts actually satisfied a rule, closing #733
  • _match_rule() now returns (conclusion, matched_facts) tuples; forward_chain() threads those facts into InferenceResult(premises=...), merging premises when the same conclusion is derived more than once within a pass
  • _prove_goal()'s base cases (goal already a known fact; goal matched via pattern unification) now return premises=[goal]/premises=[fact] instead of []
  • Facts are matched against a sorted() snapshot instead of the raw set so rule matching and premise selection are deterministic
  • Added test_forward_chaining_premises regression test mirroring the existing backward-chaining premises test

• Missing shacl optional-dependency extra (#736) by @Sameer6305

  • pip install semantica[shacl] referenced no matching extra in pyproject.toml, so pyshacl was never installed despite being documented as the fix in ontology_validator.py's ImportError message, the Explorer API, the healthcare cookbook notebook, and the changelog
  • Added shacl = ["pyshacl>=0.25.0"] to [project.optional-dependencies] and folded shacl into the all extra

• NodeEmbedder AttributeError masked in ContextGraph.analyze_graph_with_kg (#734) by @Sameer6305

  • analyze_graph_with_kg() called a non-existent NodeEmbedder.generate_embeddings(), and the surrounding broad except Exception swallowed the resulting AttributeError, silently returning {"error": "Graph analysis failed due to an internal error"} from get_causal_chain()'s supporting analytics and get_decision_insights()
  • Rewired the call site to the real NodeEmbedder.compute_embeddings(graph_store, node_labels, relationship_types) API, deriving node_labels/relationship_types from self.node_type_index/self.edge_type_index
  • Added a dedicated except AttributeError branch that logs distinctly and re-raises, so a broken internal method call surfaces as a diagnosable error instead of being indistinguishable from a legitimately empty analysis result

---

[0.5.1] - 2026-06-29

Added

• Apache Arrow & Feather File Ingestion (#705) by @Luffy2208

  • Added ArrowIngestor (semantica/ingest/arrow_ingestor.py) for reading .arrow, .feather, and .ipc files via PyArrow
  • Supports Arrow IPC File format (random-access), Arrow IPC Stream format, Feather v1 and v2
  • Selective column reads, optional row limits, and batch-aware iteration that stops early without scanning the full file
  • extract_schema() and extract_metadata() convenience methods for schema/metadata inspection without reading row data
  • _ArrowReaderWrapper provides a unified interface across all three reader types, preventing stream exhaustion during schema inspection
  • ingest_arrow() convenience function and ingest(..., source_type="arrow") unified dispatch
  • Automatic Arrow format detection in ingest() by file extension (.arrow, .feather, .ipc) and by Arrow IPC magic bytes (ARROW1\x00\x00) in FileTypeDetector
  • Registry integration under the arrow task namespace with file, schema, and metadata methods
  • Lazy-import exports of ArrowIngestor, ArrowData, and ingest_arrow from semantica.ingest
  • Optional dependency group: pip install semantica[ingest-arrow]; included in pip install semantica[all]
  • 34 tests covering schema extraction, metadata inspection, row limits, column selection, multi-batch reading, IPC stream format, Feather ingestion, empty datasets, null values, magic-byte detection, and failure modes

• Knowledge Explorer Deployment Templates (#684) by @ZohaibHassan16 and @KaifAhmad1

  • Added deploy/ directory with ready-to-use templates for 7 platforms, closing #681
  • Docker — fixed Dockerfile path (was broken on clean checkout), added non-root user, HEALTHCHECK, .dockerignore; fixed docker-compose.yml to start Explorer alongside FalkorDB on a shared network; added docker-compose.dev.yml with source volume-mounts for hot-reload (docker compose up brings up the full stack in one command)
  • Railway — deploy/railway/railway.toml with Dockerfile builder, healthcheck path, restart policy, and env vars wired from the Railway Redis plugin
  • Render — deploy/render/render.yaml Blueprint provisioning the web service and a Redis instance together with cross-linked env vars
  • Fly.io — deploy/fly/fly.toml with region, 512 MB VM, auto-stop, HTTP healthcheck, and a short README with four flyctl commands to deploy from zero
  • GCP Cloud Run — deploy/gcp/cloudbuild.yaml (build → push → deploy pipeline) and deploy/gcp/cloudrun-service.yaml (scale-to-zero, Secret Manager env vars, liveness probe)
  • Azure Container Apps — deploy/azure/azure.yaml, main.bicep (Container App + managed environment, HTTP ingress, HPA min 0 / max 10, liveness probe), and main.parameters.json; deployable with azd up
  • Kubernetes + Helm — raw manifests (namespace, configmap, secret.example, deployment with 2 replicas + rolling update, service, ingress with cert-manager TLS, kustomization); Helm chart with Chart.yaml, values.yaml, values.prod.yaml, HPA template, and helm lint-passing templates; all templates carry namespace: {{ .Release.Namespace }}
  • Added /api/health endpoint returning {"status": "ok"} used by all platform healthchecks
  • Wired ALLOWED_ORIGINS, FALKORDB_HOST, and FALKORDB_PORT from environment variables in semantica/explorer/app.py
  • Security hardened: non-root containers, readOnlyRootFilesystem, NetworkPolicy with explicit ingress/egress selectors, seccompProfile: RuntimeDefault, capabilities dropped; secrets via secret.yaml.example templates only — no committed credentials

Fixed

• Arrow ingestion double full-scan on every data read (#705) by @KaifAhmad1

  • ingest_file previously called _file_metadata (a full batch scan) before _read_batches, meaning every read scanned the entire file twice; for a limit=1 read on a large file the metadata pass visited every batch while the data pass read only one; replaced with a single-pass _read_batches_with_info that collects batch metadata as a side effect of the data read; _file_metadata is now only invoked for include_data=False

• Dead num_record_batches property on _ArrowReaderWrapper materialised all table batches (#705) by @KaifAhmad1

  • The property was never called by production code but its is_table branch called to_batches() purely to take len(), materialising the entire table in memory just for a count; property removed

• Arrow _open_file chained the wrong exception (#705) by @KaifAhmad1

  • The fallback cascade (IPC file → IPC stream → Feather) raised from feather_err, surfacing the least diagnostic error in the Python traceback chain; changed to from file_err so the IPC file open error — the most informative signal for unrecognised formats — appears as __cause__

• Neo4j Bulk CSV Export (#665) by @Luffy2208

  • Added Neo4jCSVExporter for generating Neo4j bulk-import CSV files compatible with neo4j-admin database import
  • Produces deterministic nodes.csv and relationships.csv with stable node IDs — reuses existing graph IDs or derives reproducible SHA-256 content-based IDs when none are present
  • Multi-label support via Neo4j :LABEL convention with configurable label_separator (default ;)
  • Alphabetically sorted property columns and deterministic row ordering for reproducible output across permuted inputs
  • Relationship endpoint resolution: aliases (name, text, label) automatically mapped to stable node IDs
  • Nested property serialisation to canonical JSON; flat scalar values written directly
  • dry_run() method for pre-flight CSV validation without writing files
  • validate_export() for post-write integrity checks (unique :id, consistent column widths, valid endpoint references)
  • export_nodes() and export_relationships() for partial exports
  • strict=True mode raises ValidationError on unresolved relationship endpoints
  • export_neo4j_csv() convenience function and format="neo4j_csv" / format="neo4j-csv" dispatch in export_knowledge_graph()
  • Registry integration under the neo4j_csv task namespace
  • Documentation added to semantica/export/export_usage.md with usage examples, mapping assumptions, and neo4j-admin import command
  • 13 tests covering headers, node/relationship CSV structure, multi-label, missing properties, deterministic output, CSV quoting/escaping, Unicode, empty graphs, dry-run, duplicate ID detection, ambiguous alias handling, nested property serialisation, and KnowledgeGraph integration

Fixed

• Neo4j CSV exporter _write_csv crashed with TypeError on dialect kwargs (#665) by @KaifAhmad1

  • Passing delimiter=, encoding=, or any caller kwarg to export_neo4j_csv caused csv.writer to receive unknown or duplicate keyword arguments; _write_csv now whitelists only valid csv.writer dialect params (quotechar, doublequote, skipinitialspace, escapechar, strict)

• export_neo4j_csv double-passed kwargs to both the constructor and export() (#665) by @KaifAhmad1

  • Constructor-level settings (node_file_name, relationship_file_name, encoding, delimiter, label_separator, strict) were merged into config for the constructor then re-forwarded as **kwargs to export_knowledge_graph, causing dialect params to collide; kwargs are now split into init_kwargs and call_kwargs before forwarding

• Dead node_id_lookup dict removed from _prepare_export (#665) by @KaifAhmad1

  • The {original_index → stable_id} mapping was built on every export but never consumed; removed to avoid misleading future readers

• Dropped ambiguous format="neo4j" alias from export_knowledge_graph dispatch (#665) by @KaifAhmad1

  • "neo4j" is used throughout the codebase to identify the live Bolt/Cypher graph store backend; routing it silently to the offline bulk-CSV exporter would have confused callers; only "neo4j_csv" and "neo4j-csv" are accepted

• export_usage.md documented non-existent constructor and function parameters (#665) by @KaifAhmad1

  • Examples showed node_label_sep (correct: label_separator), strict_validation (correct: strict), and nodes_path/rels_path kwargs that do not exist; all three examples corrected to match the actual API

• Public API Ingestion Support (#602) by @Luffy2208

  • Added PublicAPIIngestor class built on top of RESTIngestor for credential-free REST endpoints
  • Added PublicAPIExample and PublicAPIExamples catalog with 6 pre-configured no-auth examples:
    • jsonplaceholder_posts, jsonplaceholder_users, jsonplaceholder_todos — fake REST resources for testing
    • rest_countries_all — country reference data
    • data_gov_datasets — Data.gov CKAN catalog search
    • open_meteo_forecast — weather forecast (Berlin sample)
  • Added PublicAPIDetection dataclass for endpoint-level public/no-auth detection
  • Endpoint-level public API detection via detect_public_api() (informational, never raises)
  • No-auth validation: rejects Authorization, X-Api-Key, and all common auth headers before sending the request
  • Auth credential detection in URL query strings (api_key=, token=, access_token=, etc.)
  • Polite rate limiting with per-request and per-ingestor rate_limit_delay controls
  • Response parsing for JSON, CSV, and XML with response_format="auto" content-type detection
  • HTML response guard — text/html responses are never misclassified as XML
  • Nested record_path dot-notation extraction (e.g. "result.results" for Data.gov envelope)
  • _to_records normalization with automatic envelope unwrapping for items, data, results, records keys
  • batch_public_apis() for multi-endpoint ingestion with optional fail_fast
  • ingest_examples() for bulk example ingestion
  • sample_response() fixtures on PublicAPIExamples for mocked unit tests without live network calls
  • ingest_public_api() convenience function and ingest(..., source_type="public_api") unified dispatch
  • source_type="api" alias supported in ingest()
  • Registry integration: public_api and api task namespaces with endpoint, example, detect, batch, examples methods
  • Lazy-import exports of RESTIngestor, APIData, PublicAPIIngestor, PublicAPIExample, PublicAPIExamples, PublicAPIDetection from semantica.ingest
  • Documentation: updated docs/reference/ingest.md, docs/modules.md, and semantica/ingest/ingest_usage.md with full usage examples
  • 18 mocked tests covering JSON/CSV/XML parsing, nested record extraction, auth rejection, detection, string boolean config, batch dispatch, and unified ingest() routing
  • 3 optional-import tests covering defusedxml fallback path and import isolation without web-scraping backends

Fixed

• Public API XML parsing hardened against malicious payloads (#602) by @Luffy2208

  • Replaced stdlib xml.etree.ElementTree with defusedxml.ElementTree (XXE/entity-expansion safe); falls back to a hardened lxml parser (resolve_entities=False, no_network=True, load_dtd=False, huge_tree=False) when defusedxml is not installed
  • Added regression test asserting XXE entity payloads raise ProcessingError

• validate_no_auth config value not honoured when passed as a string (#602) by @Luffy2208

  • bool("false") evaluated to True, making validate_no_auth=False impossible via config files or environment variables; replaced with explicit _coerce_bool() that maps "false", "0", "no", "off" → False and rejects unrecognised strings with ValidationError

• Auth credential detection extended to URL query strings (#602) by @Sameer6305

  • detect_public_api() and ingest_public_api() now scan the endpoint URL itself for auth parameters (api_key, token, access_token, etc.) via urllib.parse.parse_qs, not only request headers and explicit params= dicts
  • Added regression tests for URL auth rejection (3 parametrized cases)

• ingest_examples and batch_public_apis mutable options mutation (#602)

  • Shared **options dict was passed by reference across loop iterations; mutable values such as params dicts were silently mutated after the first call, causing subsequent calls to receive a different (partially modified) options set; fixed by deep-copying options on each iteration

• rate_limit_delay forwarded twice in ingest_public_api method dispatcher (#602)

  • rate_limit_delay was consumed by the PublicAPIIngestor constructor via config but also leaked into request_kwargs forwarded to the ingestor method; added to the config_only_key strip list so it is consumed once at construction time only

• XML File Ingestion Support (#560) by @Luffy2208

  • Added XMLIngestor class with lxml backend for parsing local XML files
  • Nested element hierarchy and flat element list extraction
  • Namespace and prefix extraction with collision handling
  • Attribute and element metadata extraction
  • Optional XSD schema validation with detailed error reporting
  • Optional DTD validation (internal and external)
  • Secure-by-default parser (resolve_entities=False, no_network=True) blocking XXE attacks
  • ingest_xml() convenience function and ingest_file(..., method="xml") support
  • Unified .xml auto-detection via ingest("file.xml")
  • Directory ingestion with recursive scanning and fail_fast support
  • ingest_string() for in-memory XML bytes/str ingestion
  • Comprehensive test coverage (8/8 tests passing)

Fixed

• NERExtractor LLM method returning pattern-based output on custom gateways (#554, PR #556) by @KaifAhmad1

  NERExtractor(method="llm") silently fell back to regex/pattern extraction when used with OpenAI-compatible enterprise or self-hosted gateways (Qwen, LLaMA proxies, internal routing layers). Returned entities carried extraction_method='pattern' even though the LLM itself was producing correct tool-call output. Three root causes fixed:

  • Silent exception swallowing — exc_info=True was missing from the method-failure WARNING in NERExtractor.extract_entities. The full gateway-rejection traceback was invisible in logs even with DEBUG level enabled, making the failure impossible to diagnose without reading source code.

  • response_format=json_object sent to incompatible gateways — OpenAIProvider.generate_structured unconditionally included response_format={"type": "json_object"} in every API call. Custom/enterprise gateways frequently reject this parameter, causing both the instructor path and the manual repair loop to fail with the same error on every retry, eventually triggering _extract_fallback (pattern extraction).

  • No fallback in the generate_typed manual repair loop — when generate_structured itself raised (due to gateway rejection), the repair loop retried the identical failing call up to max_retries times before giving up. There was no path to recover via plain generate() + JSON parsing.

  Additional fixes applied during PR review:

  • Mode.JSON retry in generate_typed now strips response_format from create_kwargs before forwarding to the retry client, preventing incompatible kwargs from being sent to a client configured for a different instructor mode.
  • exc_info=True added to the generate_structured fallback warning in the manual repair loop for consistent observability across all failure paths.
  • Removed dead duplicate is_available definition in GroqProvider — Python silently kept only the second definition; the first was unreachable.
  • OpenAIProvider._init_client now validates base_url scheme at construction time. Non-HTTP(S) schemes (file://, ftp://, javascript:, etc.) raise ValueError immediately, preventing SSRF if base_url originates from configuration rather than hardcoded values.

  17 regression tests added in tests/test_issue_554_fixes.py covering all bug paths, including harshalizode's exact gateway configuration.

Security

• GitHub Actions workflow permissions hardened — added explicit permissions: contents: read + security-events: write block to defender-for-devops.yml, resolving CodeQL alert actions/missing-workflow-permissions (CWE: principle of least privilege).

• DOMPurify upgraded to 3.4.0+ via npm overrides — monaco-editor pinned dompurify at 3.2.7; added overrides in explorer/package.json to force ^3.4.0 (resolved to 3.4.10). Fixes 6 Dependabot alerts:

  • Prototype pollution → XSS bypass via CUSTOM_ELEMENT_HANDLING fallback (CVE-2026-41238 / GHSA-v9jr-rg53-9pgp)
  • Mutation-XSS via re-contextualization into raw-text wrappers (GHSA-h8r8-wccr-v5f2)
  • SAFE_FOR_TEMPLATES bypass in RETURN_DOM mode (CVE-2026-41239 / GHSA-crv5-9vww-q3g8)
  • ADD_TAGS function-predicate bypasses FORBID_TAGS (GHSA-39q2-94rc-95cp / GHSA-h7mw-gpvr-xq4m)
  • ADD_ATTR predicate skips URI validation, allowing javascript: URLs (GHSA-cjmm-f4jc-qw8r)
  • USE_PROFILES prototype pollution allows event handlers (GHSA-cj63-jhhr-wcxv)

• uuid upgraded to 13.0.1+ via npm overrides — bumped from 13.0.0 to 13.0.2, fixing missing buffer bounds check in v3/v5/v6 APIs that allowed silent partial writes into caller-provided buffers (CVE-2026-41907 / GHSA-w5hq-g745-h8pq).

• Vite upgraded from 5.x to 6.4.3 — resolves path traversal in optimised-deps .map handling (CVE-2026-39365 / GHSA-4w7w-66w2-5vf9) and the esbuild dev-server CORS issue (GHSA-4w7w-66w2-5vf9). Bundled esbuild updated from 0.21.5 → 0.25.12.

• esbuild forced to 0.28.1+ via npm override — vite 6.4.3 bundles esbuild 0.25.12 which is vulnerable to missing binary integrity verification in the Deno distribution module (GHSA-gv7w-rqvm-qjhr); added "esbuild": "^0.28.1" to overrides in explorer/package.json. npm audit now reports 0 vulnerabilities (Dependabot #15).

• Leaked Groq API keys removed from cookbook notebooks — 6 hardcoded GROQ_API_KEY values (gsk_...) stripped from configuration cells in supply_chain/01, intelligence/01, cybersecurity/01, cybersecurity/02, finance/01, and blockchain/02; fallback replaced with empty string (secret scanning alerts #1–#6). Keys were already publicly exposed — rotate them in the Groq console.

• Leaked Groq API keys removed from additional cookbook notebooks — 6 distinct hardcoded GROQ_API_KEY values stripped from 4 additional notebooks: advanced_rag/01_GraphRAG_Complete, advanced_rag/02_RAG_vs_GraphRAG_Comparison, blockchain/01_DeFi_Protocol_Intelligence, and biomedical/01_Drug_Discovery_Pipeline (secret scanning alerts #1–#6). Affected keys: gsk_SLLE0..., gsk_S4dBVJ..., gsk_SLOv6..., gsk_lR6Qcj..., gsk_ToJis6..., gsk_LmbQBr...; all publicly exposed since Dec 2025 — revoke in the Groq console and close the GitHub secret scanning alerts as "Revoked" in the Security tab.

---

[0.5.0] - 2026-05-11

Added

• Distance Intelligence Embedding Cache Optimization by @KaifAhmad1
  • Implemented per-session graph revision-based embedding cache to avoid re-scanning all nodes on every request
  • Added get_cached_embeddings() method to GraphSession with thread-safe caching and automatic invalidation
  • Updated distance matrix and semantic neighborhood endpoints to use cached embeddings for significant performance improvement
  • Added graph revision tracking using hash-based identifiers for cache invalidation
  • Implemented force refresh capability and automatic cache invalidation on graph modifications (add_nodes/add_edges)
  • Resolved TODO in graph.py for embedding caching optimization
• Parquet File Ingestion Support (#548) by @Luffy2208
  • Added ParquetIngestor class with PyArrow backend
  • Single file and partitioned directory ingestion
  • Schema and metadata extraction capabilities
  • Selective column reading with memory efficiency
  • Hive-style partition discovery support
  • Unified dispatch integration
  • Optional dependency management (ingest-parquet extra)
  • Comprehensive test coverage (32/32 tests passing)

Ontology Hub (part of #517)

• Alignments tab (PR #524, @KaifAhmad1 @ZohaibHassan16) — cross-ontology alignment authoring UI:

  • Create/edit/delete alignments with source URI, target URI, relation selector (owl:equivalentClass, all five skos:*Match variants), confidence slider, provenance, and reviewer fields.
  • Pairwise alignment matrix: scrollable table for all loaded ontology pairs; clicking a badge pre-fills the form.
  • Alignment suggestions via POST /api/ontology/suggest-alignments — blended score (0.4×label + 0.6×TF-IDF char-ngram cosine); one-click accept.
  • Ephemeral-storage banner; all handlers wrapped in useCallback.

• Health Dashboard (PR #524) — per-ontology quality scoring across 5 dimensions:

  • Completeness, Consistency, SHACL (stub), Alignment, Documentation.
  • Total score computed as mean of scoreable dimensions only (SHACL excluded when unavailable).
  • Issue list with severity badges (error/warning/info), entity URI chip, "Fix in Editor" deep-link.
  • Downloadable JSON health report; GET /api/ontology/health with _MAX_ANALYSIS_NODES = 5 000 OOM cap.

• SHACL Studio (PR #524) — interactive SHACL shape authoring:

  • Shape generation via POST /api/ontology/shacl/generate (permissive/standard/strict tiers).
  • Shape library panel with per-shape Turtle extraction; "View all" restores full document.
  • Monaco editor with custom Monarch tokenizer for Turtle syntax.
  • Validation stub via POST /api/ontology/shacl/validate; rejects empty/invalid Turtle with HTTP 422.

• Visual Ontology Editor (PR #519, @KaifAhmad1) — @xyflow/react canvas for authoring classes/properties/individuals without hand-writing OWL/Turtle:

  • Context menus on nodes (rename, add super/subclass, restrictions, SKOS metadata, deprecation, delete with impact count) and edges (toggle functional/symmetric/transitive/inverse-functional, add inverse).
  • All edits debounced and staged as pending diffs via PATCH /api/ontology/draft; nothing commits until proposal publish.

• Versions & Proposals tab (PR #519) — version timeline, proposal review (approve/reject/publish), SHACL pre-validation, side-by-side diff via VersionManager.diff_ontologies().

• Ontology Registry (PR #518, @KaifAhmad1) — full CRUD with status/format badges, per-ontology stats, live search, filter pills (All/OWL/SKOS/Internal/External), action feedback auto-hide.

• Ontology Loader (PR #518) — three-mode modal: URL import (fetch preview + load), file upload (.ttl/.rdf/.owl/.nt/.jsonld/.n3), create new (scratch/from-data/from-text).

• Entity Search panel (PR #518) — debounced 320 ms search across all loaded ontologies; type filter pills; result detail panel with super/subclasses, domain/range, instance count.

• SKOS Vocabulary Manager (PR #518) — hierarchical concept browser with recursive ConceptTreeNode, client-side filterConcepts(), full SKOS annotation detail (definition, scopeNote, broader/narrower/related/exactMatch).

• 16 backend endpoints under /api/ontology — registry, preview, load, create, search, entity, skos/schemes, skos/concept, draft, proposals CRUD, versions, alignments, health, shacl/generate, shacl/shapes, shacl/validate (PRs #518, #519, #524).

• Explorer landing page redesign (PR #516, @ZohaibHassan16) — hero section, animated SVG graph preview, live /api/graph/stats metrics, workspace launcher; Space Grotesk / IBM Plex Sans fonts; prefers-reduced-motion support.

• Distance Intelligence (PR #502, @KaifAhmad1):

  • ContextGraph.get_neighbors(include_distance_metadata) — adds distance_band, confidence_decay, path_to_anchor per result.
  • AgentContext.retrieve() / find_precedents() blend graph proximity with semantic score (combined_score = (1−w)×semantic + w×proximity).
  • 5 new API endpoints: POST /api/graph/distance-matrix (N×N, upper-triangle mirrored), GET /api/graph/node/{id}/semantic-neighborhood, GET /api/decisions/causal-distance, GET /api/temporal/distance-history, POST /api/export/distance-enriched (CSV/JSONL, capped at 200 nodes).
  • Explorer UI: Ego Mode (BFS depth-of-field fading, depth slider 1–8), Structural overlay, Semantic overlay, Heatmap (green→red by hop); Path inspector with distance band chip, metric cards, bottleneck node highlight.
  • 57 new tests in tests/context/test_distance_intelligence.py.

• Graph Explorer visual refresh (PR #503, @ZohaibHassan16) — structured ui.* design-token namespace; per-shape biomolecule/condition/compound config; decomposed toolbar memos; typed sub-components (SearchCommandBar, ToolbarCluster, etc.); deterministic LOD edge classification via GraphFullEdgeClass.

• Graph Workspace declutter (PR #483, @ZohaibHassan16) — calmer default presentation for dense graphs, display-edge aggregation with raw-edge bundle retention, grouped community view, neighborhood collapse/expand.

• Bidirectional path finding (closes #469, @KaifAhmad1) — directed=false query param on BFS and Dijkstra; undirected view built via graph.to_undirected() for traversal only; empty-path 404 guard; PathResponse.directed field.

• Node distance semantics in path responses (closes #472) — PathResponse gains hop_count and distance_band ("direct"/"near"/"mid-range"/"distant"); classify_path_distance() in semantica/utils/helpers.py; KGVisualizer.visualize_network(highlight_path) with band-scaled edge rendering.

• Native KnowledgeGraph type support in KGVisualizer (closes #471) — formal KnowledgeGraph dataclass (entities, relationships, metadata); _normalize_graph() duck-types input; raises clear ProcessingError on unknown types. 21 tests added.

• Indexed search for large graphs (PR #481, @ZohaibHassan16) — purpose-built inverted index with exact/token/prefix lookup tiers; LRU cache (128 slots); O(log n) mutation sync via bisect.insort; warm-query time 24 ms → 0.004 ms on 118 k-node graph.

• Provenance traversal multi-hop fix (PR #480, @Sameer6305) — undirected ego-graph expansion so upstream ancestors at depth ≥ 2 are no longer silently excluded; ProvenanceEdge.direction field (upstream/downstream/lateral); grouped markdown report under ## Upstream/Downstream/Lateral sections.

• TripletStore ontology namespace (PR #447, @KaifAhmad1) — _resolve_iri() applies base_uri before urn: fallback; W3C prefix expansion table (owl/xsd/rdf/rdfs/skos) expands to canonical IRIs regardless of base_uri.

• Blazegraph literal serialization (PR #448, @KaifAhmad1) — _format_object_for_sparql() selects IRI/typed-literal/language-tagged-literal/plain-literal token; _resolve_datatype_iri() with prefix expansion; RFC 5646 language-tag validation; _escape_literal() for string escaping.

• DeepSeek provider via OpenAI SDK (PR #482, @liling) — _init_client rewritten using openai.OpenAI(base_url=self.base_url) instead of defunct deepseek package; verbose_mode assignment fix; pyproject.toml updated to openai>=1.0.0.

• DuplicateDetector result limiting and ranking (issue #534, by @KaifAhmad1):

  • max_results — hard global cap on returned candidates; applied after sorting. None means no limit.
  • top_k_per_entity — keep at most k candidates per entity (by the sort field) so no single entity floods the output. None means no per-entity limit.
  • min_similarity — extra similarity floor on top of similarity_threshold; candidates below it are dropped before ranking. None means no extra floor.
  • sort_by — ranking field before limits are applied; accepts "confidence" (default) or "similarity_score". Invalid values raise ValueError at construction time.
  • All four options are applied by the new _apply_result_limits helper and are respected by both detect_duplicates() and incremental_detect().
  • 15 new tests in TestResultLimiting covering each option in isolation and in combination.
  • Follow-up Qodo review fixes (by @KaifAhmad1):
    • top_k_per_entity now uses OR semantics — a candidate is kept if either entity is still under quota, preventing high-quality pairs from being silently dropped when a popular counterpart saturates its limit.
    • max_results and top_k_per_entity now validated at construction time; negative or non-integer values raise ValueError.
    • min_similarity now validated in [0.0, 1.0] at construction; out-of-range values raise ValueError.
    • Added _normalize_entity_id helper (always returns str) used consistently in both _apply_result_limits and _build_duplicate_groups, eliminating int vs str ID key mismatches.
    • Updated detect_duplicates and incremental_detect docstrings to reflect the configurable sort_by field.

Fixed

• Fix: ConflictDetector.detect_conflicts() raises AttributeError when called with method= or property_name= kwargs (issue #533, PR conflicts, by @KaifAhmad1):

  • detect_conflicts was defined twice in conflict_detector.py; Python silently overwrote the first (dispatcher) definition with the second (comprehensive), which accepted no method or property_name parameters — causing AttributeError or TypeError for any caller using those kwargs.
  • Removed the first (dead) definition and merged its dispatcher logic into the surviving method. New signature: detect_conflicts(entities, method="all", property_name=None, entity_type=None, **kwargs).
  • Supported method values: "all" (default, comprehensive), "value", "property", "type", "relationship", "temporal", "logical", "entity". Unknown values raise ValueError.
  • Fixed method="relationship" silently defaulting relationships to the entities list, which caused entity dicts to be iterated as relationship dicts producing silent wrong results (None_None_None keys). Now defaults to [] with dict normalization.
  • Removed unreachable dead code (for field_name in fields_to_check loop after try/except raise) in detect_entity_conflicts.
  • Follow-up Qodo review fix — hardened method="relationship" normalization: when relationships kwarg is a dict whose "relationships" value is itself a non-list (or the key is absent), the value is now always wrapped in a list before being passed to detect_relationship_conflicts, guaranteeing List[Dict] input in all cases.

• Fix: semantica[all] installation fails on Windows due to faiss-gpu dependency (issue #532, PR #utlis, by @KaifAhmad1):

  • [all] bundled the [gpu] extra (faiss-gpu>=1.7.0, cupy>=10.0.0), which has no Windows builds, causing pip install "semantica[all]" to fail with No matching distribution found for faiss-gpu>=1.7.0.
  • Removed gpu from both [all] lines in pyproject.toml — [all] now installs only cross-platform dependencies. Users on Linux who need GPU acceleration can install semantica[gpu] explicitly.

• Fix: Progress tracker crashes with UnicodeEncodeError on Windows cp1252 consoles (issue #531, PR #utlis, by @KaifAhmad1):

  • ConsoleProgressDisplay.update() had 5 direct sys.stdout.write() calls that bypassed the existing _safe_write() guard, causing UnicodeEncodeError when emoji characters (🧠, 📊) were written to cp1252-encoded consoles during any progress-tracked operation.
  • All 5 calls replaced with self._safe_write(), which catches UnicodeEncodeError and re-encodes output with errors="replace" so progress output never crashes the process.
  • Added TestProgressTrackerEncoding regression class (3 tests) covering _safe_write safety, pipeline header write, and auto emoji-disable on cp1252 stdout.

• Fix: Break circular import in semantic_extract; address Qodo review bug (issue #528, PR #536, by @ZohaibHassan16, review fixes by @KaifAhmad1):

  • Root cause — ner_extractor.py imported get_entity_method from methods.py, while methods.py imported Entity from ner_extractor.py, creating a circular import that raised ImportError: cannot import name 'Entity' from partially initialized module on any import of semantica.semantic_extract.
  • semantica/semantic_extract/types.py (new) — shared Entity, Relation, and Triplet dataclasses extracted into a dedicated module that neither side of the old cycle imports, so both ner_extractor, relation_extractor, triplet_extractor, and methods can import from it freely.
  • semantica/semantic_extract/__init__.py — lazy-loads package-level exports so core extractor imports do not pull in optional modules (e.g. the YAML-backed semantic network extractor); added TripleExtractor as a compatibility alias for TripletExtractor; legacy re-exports from the individual extractor modules preserved for backward compatibility.
  • semantica/semantic_extract/methods.py — updated to import shared types from types.py; extractor-specific imports moved to function scope where needed to prevent re-introducing the cycle.
  • Added regression tests (tests/semantic_extract/test_imports.py) covering import order independence (methods-before-extractors and extractors-before-methods), legacy type import compatibility, TripleExtractor alias, and that core imports do not require yaml.
  • Review fix (Qodo — Py3.8 test import crash): test_imports.py annotated _run_python as -> subprocess.CompletedProcess[str], which is not subscriptable at runtime on Python 3.8 (generic subscript on built-in types requires 3.9+). Added from __future__ import annotations (PEP 563) so all annotations are lazy strings never evaluated at import time, restoring compatibility with the declared requires-python = ">=3.8" without any behaviour change on 3.9+.

• Fix: Lazy-load optional ingest backends; address Qodo review bugs (issue #527, PR #535, by @ZohaibHassan16, review fixes by @KaifAhmad1):

  • semantica/ingest/__init__.py — core exports (FileIngestor, ingest_file, config, registry) remain eagerly imported; all optional backends (WebIngestor, FeedIngestor, RepoIngestor, EmailIngestor, StreamIngestor, DBIngestor, MCPIngestor, OntologyIngestor, SnowflakeIngestor) are now deferred behind a module-level __getattr__, so from semantica.ingest import FileIngestor no longer fails when GitPython or BeautifulSoup4 are absent.
  • semantica/ingest/methods.py — backend imports relocated into their respective ingestion functions (ingest_web, ingest_feed, ingest_repository, ingest_email) with helper _missing_optional_dependency() / _is_missing_dependency() for consistent, actionable error messages.
  • Review fix (Bug 1 — overbroad missing-dep detection): replaced except ImportError with except ModuleNotFoundError in all four function-level import guards and in __getattr__. ImportError catches failures thrown by code inside a successfully found module, masking real bugs with a misleading "package not installed" message; ModuleNotFoundError (its subclass) is specific to absent modules. Simplified _is_missing_dependency to rely solely on exc.name now that ModuleNotFoundError always sets it.
  • Review fix (Bug 2 — expected errors logged as failures): added except ConfigurationError: raise before the blanket except Exception handlers in ingest_web, ingest_feed, ingest_repository, and ingest_email. Missing optional dependencies are expected user-configuration issues and must not produce error-level log entries.
  • Review fix (Bug 3 — test blocker not setting exc.name): OptionalDependencyBlocker.find_spec now sets err.name = root_name on the manually constructed ModuleNotFoundError, matching what Python's import machinery does, so _is_missing_dependency correctly identifies the missing package in tests.
  • Added regression tests (tests/ingest/test_optional_imports.py) that block the git and bs4 modules via a custom meta path finder and assert core imports succeed and backends raise ConfigurationError with an actionable message.

• Fix: Ontology Hub post-review bug fixes and security hardening (follow-up to #518, closes security advisory #23, by @KaifAhmad1):

  • Broken registry filters — fetchRegistry was sending toolbar filter values (owl, skos, internal, external) to the backend as the status query param, which only accepts published|draft|external, causing those filters to return empty lists. Removed the spurious status param; all format/kind filtering is now applied client-side via filteredEntries, which already had the correct logic.
  • Toggle/refresh URI corruption — toggle_ontology and refresh_ontology applied .removesuffix("/toggle") / .removesuffix("/refresh") to the captured path parameter, which would silently corrupt any ontology URI that legitimately ends with those strings. Starlette's route regex (/{uri:path}/toggle) already strips the literal suffix via backtracking, so the removesuffix calls were removed and the raw ontology_uri parameter is used directly.
  • SSRF in URL fetch — _fetch_url_sync() accepted arbitrary user-supplied URLs and called requests.get() with no validation, enabling server-side request forgery against internal services. Added _validate_fetch_url() which rejects non-http/https schemes and resolves the hostname via socket.getaddrinfo, blocking loopback, private, link-local, reserved, and multicast addresses.
  • File upload format misdetected — the file picker accepted .xml and .json but fmtMap had no entries for those extensions, causing them to default to turtle. Added xml: "xml" and json: "json-ld" mappings. Changed the unknown-extension fallback from || "turtle" to ?? "" (empty string), and omit the format key from the request body when empty so the backend _detect_format() runs instead of receiving a forced incorrect value. Also added .n3 to the accepted extension list and dropzone hint.
  • Inconsistent XML hardening — _parse_rdf_sync() called rdflib.Graph().parse() directly, bypassing the defusedxml-based XXE protection already present in semantica/explorer/utils/rdf_parser.py. Now routes through _safe_parse_rdf() from that module, applying consistent protection for all RDF/XML parse paths.
  • Search scans whole graph (GET /api/ontology/search) — the endpoint fetched up to 999 999 nodes and performed a linear Python substring scan on every request. Replaced with session.search(q, limit * 6) which uses the GraphSearchIndex; results are then post-filtered by _SEARCHABLE_TYPES and entity_type before being returned up to the requested limit.
  • ReDoS in format detector (security advisory #23, CodeQL py/polynomial-redos, CWE-1333/730/400) — _detect_format() used re.match(r"_:\w+|<[^>]+>\s+<[^>]+>", ...) to detect N-Triples content. The <[^>]+>\s+<[^>]+> alternative was flagged as a polynomial regular expression on uncontrolled data. The URI-subject branch was already unreachable (strings starting with < return "xml" two lines above), so the entire regex was replaced with two O(1) string operations: stripped.startswith("_:") and " <" in stripped. import re removed as now unused.

• OWLExporter Turtle syntax (closes #478) — invalid multi-block output fixed via _ttl_block(); data properties no longer silently dropped; _escape_ttl_str() applied to all label/comment/version sites. 43 tests added.

• OWLGenerator schema compatibility (Issue #446) — label-first IRI fallback, list-typed datatype ranges, per-call namespace consistency, subClassOf/subclassOf parity.

• TripletStore IRI regressions (PR #447 follow-up) — non-string IDs coerced to str(); W3C prefix expansion now correct regardless of base_uri.

• KGVisualizer accepts KnowledgeGraph objects (closes #458) — _normalize_graph() duck-types input; raises clear ProcessingError on unknown types. 21 tests added.

• Semantic Distance UI slash-safe routes (PR #515, @ZohaibHassan16) — query-param routes /api/graph/semantic-neighborhood?node_id= and /api/graph/path?source=&target= bypass FastAPI's %2F pre-decode; legacy path-segment routes kept as deprecated aliases.

• Explorer Distance Intelligence rendering (PR #513, @ZohaibHassan16) — distance state flows through Sigma reducer/theme pipeline instead of mutating raw graph attributes; restoreNodeColors() race eliminated by merging ego/heatmap useEffect hooks.

• Distance Intelligence code review regressions (PR #502 follow-up, @KaifAhmad1) — top_k param name fix; include_distance_metadata gated behind False default; weakest_link key standardized; temporal sampling uses timedelta not timetuple; O(E×L) decay replaced with O(E) index; AgentContext._apply_proximity_metadata stores graph_node_id separately; sweep animation sweepGeneration counter fix; HTTP 413 for >200 node subsets; upper-triangle distance matrix.

• Knowledge Explorer blockers (PR #420, @ZohaibHassan16):

  • Dockerfile: renamed DockerFile → Dockerfile; fixed CMD module path; added app = create_app() at module level.
  • CORS: default origins narrowed from "*" to localhost:5173 only.
  • get_ws_manager() now raises HTTP 503 instead of unhandled AttributeError.
  • SPARQL: read-only enforcement — INSERT/DELETE/UPDATE/LOAD/DROP rejected.
  • Vocabulary: 10 MB upload cap; JSON-LD format auto-detection for .jsonld/.json-ld/.json.
  • Annotation O(1) lookup via GraphSession.get_annotation(id).
  • Self-loop guard in batchMergeEdges prevents Graphology crash.
  • Static build artifacts removed from git; semantica/static/ added to .gitignore.

• Ontology Hub post-review hardening (PR #518 follow-up, @KaifAhmad1):

  • Registry filter: status param removed from fetchRegistry; filtering applied client-side.
  • Toggle/refresh URI: removed .removesuffix() calls that corrupted URIs ending with those strings.
  • Format detector: _detect_format() ReDoS eliminated — re.match replaced with two O(1) string ops.
  • Broken fmtMap entries: added xml/json mappings; unknown-extension fallback changed from || "turtle" to ?? "".
  • XML hardening: _parse_rdf_sync() now routes through _safe_parse_rdf() for consistent defusedxml XXE protection.
  • Search: replaced O(999 999) linear scan with GraphSearchIndex-backed session.search().

Security

• 12 vulnerability fixes (PR security-enhancement, @KaifAhmad1):
  • [CRITICAL — CWE-95] Eval injection in media_parser.py: replaced eval(ffprobe_output) with fractions.Fraction.
  • [CRITICAL — CWE-502] Pickle deserialization in agent_memory.py: replaced with JSON; legacy .pkl files detected and refused with migration message.
  • [HIGH — CWE-89] SQL injection in snowflake_ingestor.py: LIMIT/OFFSET parameterized; ORDER BY regex-validated; WHERE clauses containing semicolons rejected.
  • [HIGH — CWE-611] XXE in rdf_parser.py: defusedxml.defuse_stdlib() before all RDF/XML parsing.
  • [HIGH — CWE-346/200] Missing security headers in server.py: CORSMiddleware, X-Content-Type-Options, X-Frame-Options, HSTS, generic 500 handler.
  • [HIGH — CWE-346/400] Overpermissive CORS in explorer/app.py: methods/headers narrowed; 64 KB WebSocket frame cap.
  • [MEDIUM — CWE-20] Algorithm param unconstrained in graph.py: enum-validated bfs|dijkstra only.
  • [MEDIUM — CWE-434] RDF upload without extension check in vocabulary.py: .ttl/.rdf/.owl/.xml/.jsonld allowlist enforced.
  • [MEDIUM — CWE-1336] Prompt injection in llm_extraction.py: user-supplied content wrapped in json.dumps().
  • [MEDIUM — CWE-95] Dynamic __import__() in pipeline_validator.py: replaced with proper module-level import.
  • [MEDIUM — CWE-1333] ReDoS in enrich.py: whitespace-normalize then split on literal " AND ".
  • [LOW — CWE-22] Path traversal in server.py SPA route: Path.resolve().relative_to() guard; 400 on escape.
  • [LOW — CWE-400] Unbounded SPARQL in sparql.py: 5 000-row cap, 30 s asyncio.wait_for timeout, Semaphore(4) concurrency cap; SparqlResponse.truncated field added.
  • [LOW — CWE-434] Import upload in export_import.py: 50 MB cap; {.json,.csv} allowlist.
  • CodeQL paths-ignore for cookbook/**/*.html to suppress false-positive JS alerts #15–18.
• SSRF in Ontology Hub (PR #518 follow-up): _validate_fetch_url() rejects non-http/https schemes and resolves hostname via socket.getaddrinfo, blocking loopback/private/link-local/multicast addresses.

---

[0.4.0] - 2026-04-08

Added

Temporal Intelligence (@KaifAhmad1, PRs #396–#402)

• Core Temporal Data Model (PR #396) — semantica.kg.temporal_model with shared parsing/normalization/serialization helpers; TemporalBound and BiTemporalFact exported from semantica.kg; valid-time and transaction-time filtering; TemporalValidationError on invalid inputs; history-preserving revisions in TemporalVersionManager.apply_revision() with supersession semantics.
• Point-in-Time Query Engine (PR #397) — TemporalGraphQuery.reconstruct_at_time(graph, at_time) builds consistent point-in-time subgraphs without mutating source; TemporalConsistencyReport detects inverted intervals, relationships outside entity lifetimes, overlapping same-type relationships, and temporal gaps; sequence/cycle pattern detection; calendar-aligned evolution bucketing via temporal_granularity; causal ordering controls on find_temporal_paths() (strict/overlap/loose).
• Deterministic Temporal Reasoning Engine (PR #398) — semantica.kg.temporal_reasoning; full Allen interval algebra via IntervalRelation (all 13 relations); TemporalReasoningEngine with interval merging, gap analysis, coverage calculation, timelines, retroactive coverage; zero LLM calls; circular import risk between semantica.reasoning and semantica.kg eliminated.
• Temporal Awareness in ContextGraph (PR #399) — Decision dataclass gains valid_from/valid_until; superseded decisions remain in graph (immutable history); find_precedents_by_scenario(include_superseded, as_of); ContextGraph.state_at(timestamp) serializable snapshot; CausalChainAnalyzer.trace_at_time(event_id, at_time); AgentContext.checkpoint(label), diff_checkpoints(), flush_checkpoint().
• Temporal Metadata Extraction from Text (PR #400):
  • extract_relations_llm(extract_temporal_bounds=True) — each Relation gains valid_from, valid_until, temporal_confidence (0.0–1.0), temporal_source_text; default False is 100% backward-compatible.
  • Calibrated confidence anchors: 1.00 = full ISO date → 0.00 = no temporal signal.
  • TemporalNormalizer (zero LLM calls, pure regex + dateutil): normalize(value) → UTC datetime tuple or None; normalize_phrase(phrase) → metadata dict or None; 13-domain default phrase map; TemporalAmbiguityWarning for ambiguous DD/MM/YYYY inputs (never silently guesses locale).
• Temporal Provenance & OWL-Time Export (PR #401):
  • ProvenanceTracker.track_entity() auto-stamps recorded_at on every new record.
  • query_recorded_between(start, end), revision_history(fact_id), export_audit_log(fact_ids, format) (JSON/CSV).
  • RDFExporter.export_to_rdf(include_temporal=True, time_axis="valid|transaction|both") — emits OWL-Time triples for all temporally-annotated relationships.
  • create_snapshot() stamps "format_version": "1.0"; validate_snapshot() and migrate_snapshot() for stable snapshot lifecycle.
• Temporal GraphRAG Integration (PR #402) — TemporalGraphRetriever filters retrieved context to a point in time; ContextRetriever.query_with_reasoning(at_time, header_template) prepends structured temporal header; TemporalQueryRewriter extracts temporal intent (before/after/at/during/between) from natural language; regex-only by default, optional LLM-assisted mode.

Ontology (@KaifAhmad1 @ZohaibHassan16)

• SHACL Shape Generation & Validation (PR #318) — SHACLGenerator derives SHACL node/property shapes from any ontology dict; three quality tiers (basic/standard/strict); Turtle/JSON-LD/N-Triples output; iterative multi-level inheritance propagation, cycle-safe; OntologyEngine.to_shacl(), export_shacl(), validate_graph(explain=True); SHACLValidationReport with plain-English explanations for all 7 constraint types. pip install semantica[shacl].
• SKOS Vocabulary Module (PR #319) — TripletStore.add_skos_concept() / get_skos_concepts(scheme_uri); OntologyEngine.list_vocabularies(), list_concepts(), search_concepts(); NamespaceManager.get_skos_uri() / build_concept_scheme_uri(); SPARQL injection hardened.
• Ontology Alignment API (PR #361) — OntologyEngine.create_alignment(), get_alignments(), list_alignments(); OWL/SKOS standard predicates (owl:equivalentClass, all five skos:*Match); ReuseManager.suggest_alignments(); QueryEngine.expand_entity_uri(use_alignments=True) with SPARQL VALUES clause injection; SPARQL injection hardened.
• Ontology Diff & Migration (PR #367) — VersionManager.diff_ontologies() covering classes/properties/individuals/axioms; ChangeLogAnalyzer.analyze() classifying CRITICAL/HIGH/MEDIUM/INFO impact; ImpactReport, generate_change_report(); OntologyEngine.compare_versions() end-to-end orchestrator with optional validation and graph-instance checks.

Knowledge Explorer API (@ZohaibHassan16 @KaifAhmad1)

• Full FastAPI backend (PR #384) — semantica.explorer package with graph, analytics, decisions, temporal, enrichment, export/import, annotations routes; 12 export formats; WebSocket progress for import; 99 integration tests. pip install semantica[explorer]; CLI: semantica-explorer --graph my_graph.json.
• Thread safety (PR #385) — ContextGraph and GraphSession protected with threading.RLock; 8 analytics components lazily initialized under lock.
• In-memory fallbacks (PR #386) — All 7 DecisionQuery and 4 DecisionRecorder methods have ContextGraph fallback paths for in-memory usage without a graph DB.
• Snapshot schema compatibility (PR #393) — accepts both nodes/edges and entities/relationships snapshot schemas transparently; metadata counts always accurate.
• Audit trail & rollback protection (PR #394) — mutation-level audit tracking, named version tags, restore_snapshot() requires explicit confirmation, get_node_history(), diff() Git-like alias.
• SKOS Vocabulary REST API (PR #426) — GET /api/vocabulary/schemes, GET /api/vocabulary/hierarchy?scheme=<uri> with cycle detection, POST /api/vocabulary/import (.ttl/.rdf/.owl; HTTP 422 on invalid).
• O(N) → O(limit) Pagination (PR #431) — find_nodes/find_edges use itertools.islice on generators; ghost-node fix (accepts source_id/target_id and source/target key names); deterministic page boundaries via sorted(); stats() applies same validity filters as pagination.
• Named graph support (PR #432, @Sameer6305) — enable_named_graphs flag forwarded correctly through TripletStore.execute_query(); duplicate FROM/FROM NAMED clauses prevented; graph URIs percent-encoded in DROP statements.

Integrations

• Agno Agentic Framework (Issue #249, @KaifAhmad1) — 5 components, all degrading gracefully when agno is not installed:
  • AgnoContextStore — graph-backed agent memory implementing agno.memory.db.base.MemoryDb.
  • AgnoKnowledgeGraph — multi-hop GraphRAG knowledge base implementing agno.knowledge.base.AgentKnowledge.
  • AgnoDecisionKit — 6 decision-intelligence tools (record_decision, find_precedents, trace_causal_chain, analyze_impact, check_policy, get_decision_summary).
  • AgnoKGToolkit — 7 KG pipeline tools (extract_entities, extract_relations, add_to_graph, query_graph, find_related, infer_facts, export_subgraph).
  • AgnoSharedContext — team coordinator with single shared ContextGraph; bind_agent(role) returns role-scoped view; thread-safe via RLock.
  • 110 integration tests; 3 cookbook notebooks. pip install semantica[agno].
• Novita AI Provider (PR #374, @Alex-wuhu) — OpenAI-compatible; default model deepseek/deepseek-v3.2; NOVITA_API_KEY; create_provider("novita").

Reasoning

• Native Datalog Reasoning Engine (PR #371, @ZohaibHassan16) — pure-Python bottom-up semi-naive fixpoint with guaranteed termination; recursive Horn clause rules (e.g. ancestor(X,Y) :- parent(X,Z), ancestor(Z,Y).); O(1) delta-index lookup; load_from_graph(ContextGraph); query("pred(?X, ?Y)") with optional bindings=; DatalogReasoner, DatalogFact, DatalogRule exported from semantica.reasoning.

Fixed

• Pattern Matcher restored (PR #387, @ZohaibHassan16) — dead code silently overwrote _match_pattern regex (pre-bound variable embedding, repeated-variable backreferences) with re.escape, breaking transitivity/symmetry/self-join rules; removed. re.error now surfaced instead of swallowed.
• OllamaProvider base_url ignored (PR #408, @AlexeyMyslin) — ollama.Client(host=self.base_url) instead of raw module assignment; remote Ollama servers now reachable.
• spaCy runtime fallback — NERExtractor now catches runtime initialization failures, not just missing-model errors.
• CentralityCalculator crash — _build_adjacency() handles both ContextGraph dataclass edges (source_id/target_id) and plain dicts.
• find_path always used BFS (PR #384) — algorithm query param now correctly dispatched to dijkstra_shortest_path or bfs_shortest_path.
• Event loop blocked in /api/enrich/links (PR #385) — score_link scoring loop wrapped in asyncio.to_thread.
• Temp file leak in export_graph (PR #384) — try/finally cleanup for all error paths.
• ChangeCategory enum typo (PR #367) — "potenitally_breaking" → "potentially_breaking".
• DecisionQuery/DecisionRecorder fallbacks (PR #386) — type() guard instead of isinstance() for Mock safety; flat property storage in _store_decision_node; spurious properties={} kwarg removed; tz-aware/naive datetime mismatch resolved; find_edges() hoisted out of BFS loop (O(nodes×edges) → O(1) per call).
• Snapshot schema (PR #393) — silent restore failures when nodes/edges schema didn't match legacy entities/relationships expectations.
• Context explainability (@KaifAhmad1) — decision nodes now store full scenario/reasoning text; causal/precedent reconstruction returns enriched Decision objects; PolicyEngine.get_affected_decisions() consistent across Cypher and fallback branches.

Security

• CWE-312/359/532 — Removed api_key debug print blocks from relation_extractor.py and triplet_extractor.py.
• CWE-20 — URL sanitization: "url" in urls replaced with any(url == "url" for url in urls), eliminating substring match.
• CI overpermissions — permissions: contents: read added to benchmark.yml and security.yml.
• SHACL path traversal (PR #318) — replaced len < 500 and "\n" not in s heuristic with os.path.exists().
• SHACL inheritance mutation (PR #318) — _propagate_inheritance uses dataclasses.replace() instead of appending parent PropertyShape objects by reference.
• SPARQL injection (PR #361) — search_concepts, list_alignments, build_values_clause fully hardened.

---

[0.3.0] - 2026-03-10

Added

• Context Graph Feature Completeness (@KaifAhmad1):
  • ContextNode / ContextEdge gain valid_from / valid_until with is_active(at_time) -> bool.
  • ContextGraph.find_active_nodes(node_type, at_time) — temporal node filtering.
  • get_neighbors(min_weight) — confidence-filtered BFS (default 0.0 passes all edges).
  • link_graph() / navigate_to() / resolve_links(registry) — cross-graph navigation with full save/load round-trip.
  • graph_id UUID field persisted to JSON.

Fixed

• is_active() tz-aware/naive datetime normalization.
• valid_from/valid_until serialization in add_nodes(), add_edges(), to_dict(), from_dict().
• Cross-graph link phantom-node prevention in link_graph().
• pipeline_builder.add_step() return type annotation.
• test_hybrid_search_performance timing computation; threshold raised to < 5.0 s.
• ProvenanceTracker added to semantica/kg/__init__.py exports.
• Duplicate relation creation in _parse_relation_result — orphaned legacy block removed.
• extraction_method parameter added; typed path now correctly sets "llm_typed".
• Cross-test cache pollution in test_retry_logic.py — _result_cache.clear() added to setUp().
• 14 tests in tests/context/test_cross_graph_navigation.py; 85 real-world tests in tests/test_030_realworld_comprehensive.py.

---

[0.3.0-beta] - 2026-03-07

Added

• Multi-Founder LLM Extraction (PR #354, @KaifAhmad1):
  • _parse_relation_result: unmatched subjects/objects produce a synthetic UNKNOWN entity instead of being silently dropped.
  • _match_pattern rewritten: splits on ?var placeholders, pre-bound variable resolution, repeated-variable backreferences.
• TTL Export Aliases (PR #355, @KaifAhmad1) — format="ttl"/"nt"/"xml"/"rdf"/"json-ld" resolve correctly before format validation; 8 tests in tests/export/test_rdf_exporter.py.
• Incremental/Delta Processing (PR #349, @ZohaibHassan16) — native delta computation between graph snapshots via SPARQL, delta-aware pipeline execution (delta_mode), snapshot retention with prune_versions(), significant performance improvements for near real-time pipelines.
• Deduplication v2:
  • Candidate Generation v2 (PR #338, @ZohaibHassan16) — multi-key blocking, phonetic (Soundex) blocking, deterministic candidate budgeting; 63.6% faster (0.259 s → 0.094 s for 100 entities).
  • Two-Stage Scoring Prefilter (PR #339, @ZohaibHassan16) — type mismatch, length ratio, token overlap gates; 18–25% faster batch processing.
  • Semantic Relationship Deduplication v2 (PR #340, @ZohaibHassan16) — predicate synonym mapping (works_for → employed_by), O(1) hash matching, weighted scoring (60% predicate + 40% object); 6.98x speedup (~83 ms vs ~579 ms).
  • Migration Guide (PR #344, @ZohaibHassan16) — comprehensive MIGRATION_V2.md; critical infinite recursion bug in dedup_triplets() fixed.
• ArangoDB AQL Export (PR #342, @tibisabau) — AQL INSERT generation, configurable collections, batch processing (default 1 000), .aql auto-detection, 17 tests.
• Apache Parquet Export (PR #343, @tibisabau) — columnar storage, configurable compression (snappy/gzip/brotli/zstd/lz4/none), explicit Arrow schemas, .parquet auto-detection, 25 tests.

Fixed

• Test Suite Fixes (@KaifAhmad1):
  • Context: entity extraction gated on use_hybrid_search=True; _extract_entities_from_query uses word[0].isupper(); added expand_context BFS method; hybrid_retrieval and multi_hop_context_assembly corrected; vector result fallback to metadata["content"].
  • KG: calculate_pagerank aliases; community_detector._to_networkx no longer silently loses edges; _build_adjacency handles both "edges" and "relationships" keys; 9 tracking methods added to AlgorithmTrackerWithProvenance.
  • Pipeline: retry loop honours max_retries; FailureHandler.handle_failure() added; add_step return type fixed; validate alias added; error message standardized.
  • Tests: emoji replaced with ASCII for Windows cp1252 compatibility.
• NameError: missing Type import in utils/helpers.py.

---

[0.3.0-alpha] - 2026-02-19

Added

• Decision Tracking System — complete lifecycle management (record → analyze → query → precedent → influence) with audit trails and provenance tracking.
• Advanced KG Algorithms — Node2Vec embeddings, centrality analysis, community detection for decision insights.
• Enhanced Context Module — unified AgentContext with granular feature flags for decision tracking, KG algorithms, and vector store features.
• Vector Store Features — hybrid search combining semantic, structural, and category similarity.
• Policy Management — versioning, compliance checking, and exception handling.
• Context Engineering Enhancement (PR #307, @KaifAhmad1) — full decision tracking, hybrid search, PolicyException model, GraphStore validation, explainable AI features, 9 critical bug fixes, 100% test coverage (9/9).
• PgVector Store Support (PR #303, @Sameer6305 @KaifAhmad1) — HNSW/IVFFlat indexing, JSONB metadata filtering, psycopg3/psycopg2 fallback, SQL injection protection via psycopg_sql.SQL(), 36+ tests.
• Apache AGE Backend (PR #311, @Sameer6305) — AgeStore with GraphStore API compatibility, SQL injection protection.
• Improved Vector Store for Decision Tracking (PR #293, @KaifAhmad1) — DecisionEmbeddingPipeline, HybridSimilarityCalculator (0.7 semantic + 0.3 structural), DecisionContext, ContextRetriever with multi-hop reasoning; 34+ tests.
• Improved Graph Algorithms (PR #292, @KaifAhmad1) — 30+ algorithms across 7 categories (Node2Vec, Dijkstra, A*, PageRank, Louvain, Leiden, etc.), unified provenance tracking with GraphBuilderWithProvenance / AlgorithmTrackerWithProvenance.
• ResourceScheduler Deadlock Fix (PRs #299 #301, @d4ndr4d3 @KaifAhmad1) — threading.Lock → threading.RLock; allocation validation; leak prevention on failure; 6 regression tests.
• Dependabot & Security Automation — bi-weekly security updates, automated Bandit/Safety/Semgrep scans, security-critical package grouping.

Fixed

• Context Graphs decision tracking bugs (PR #315, @KaifAhmad1): empty/None decision ID, None metadata, causal chain depth logic, nonexistent node handling, to_dict/from_dict round-trip.
• PolicyEngine latest version selection; AgentContext fallback robustness and secure logging.
• Import issues in test suite (ProvenanceTracker location); causal analyzer max_depth bounds.

---

[0.2.7] - 2026-02-09

Added

• Snowflake Connector (PR #276, @Sameer6305) — multi-auth (password/OAuth/key-pair/SSO), table and query ingestion, SQL injection prevention, progress tracking, 24 tests. pip install semantica[db-snowflake].
• Apache Arrow Export (PR #273, @Sameer6305) — explicit Arrow schemas, entity/relationship export, Pandas/DuckDB compatible, 20 tests.
• Benchmark Suite (PR #289, @ZohaibHassan16 @KaifAhmad1) — 137+ benchmarks across all 10 modules, Z-score statistical regression detection, GitHub Actions workflow. CLI: python benchmarks/benchmark_runner.py.

---

[0.2.6] - 2026-02-03

Added

• W3C PROV-O Provenance Tracking (Issues #254 #246, @KaifAhmad1):
  • Comprehensive provenance across all 17 Semantica modules; InMemory/SQLite backends; SHA-256 integrity.
  • FDA 21 CFR Part 11, SOX, HIPAA, TNFD compliance infrastructure.
  • 237 tests; opt-in (provenance=False by default).
• Enhanced Change Management (Issues #248 #243, @KaifAhmad1):
  • TemporalVersionManager and OntologyVersionManager with SQLite/in-memory backends; SHA-256 checksums; detailed diffs.
  • 104 tests; 17.6 ms for 10 k entities; 510+ ops/sec concurrent.
• CSV Ingestion Enhancements (PR #244, @saloni0318) — auto-detect encoding (chardet) and delimiter (csv.Sniffer); tolerant decoding; optional chunked reading.
• Ingest Unit Tests (Issues #239 #232, @Mohammed2372) — file, web, and feed ingestors; 998 lines of tests; 80–86% coverage.
• TextNormalizer comprehensive unit tests (PR #242, @ZohaibHassan16).

Fixed

• Temperature Compatibility (Issues #256 #252, @F0rt1s @IGES-Institut) — temperature=None now omits parameter so APIs use model defaults; _add_if_set helper applied to all 5 providers; 10 tests.
• JenaStore Empty Graph (Issues #257 #258, @ZohaibHassan16) — if self.graph is None: replaces implicit falsy check in 5 methods.

---

[0.2.5] - 2026-01-27

Added

• Pinecone Vector Store (closes #219 #220) — serverless and pod-based indexes, namespace support, metadata filtering, unified VectorStore integration.
• Configurable LLM Retry Logic — max_retries parameter (default 3) in NERExtractor, RelationExtractor, TripletExtractor, and all extract_*_llm methods.
• Bring Your Own Model (BYOM) — custom HuggingFace models in all extractors; custom tokenizer support; runtime model= overrides config defaults.
• Enhanced NER — configurable aggregation strategies (simple/first/average/max); IOB/BILOU parsing for raw model outputs; confidence scoring.
• Relation Extraction — entity marker technique (<subj>/<obj> tags) for sequence classification models; structured output parsing.
• Triplet Extraction — Seq2Seq model support (REBEL) for direct structured triplet generation from text.

Fixed

• LLM extraction: strict max_retries enforcement prevents infinite retry loops.
• Model parameter precedence: runtime arguments now correctly override config defaults in HuggingFace extractors.
• Circular imports in test suites.

---

[0.2.4] - 2026-01-22

Added

• Ontology Ingestion Module — OntologyIngestor for Turtle/RDF-XML/JSON-LD/N3 files; ingest_ontology() convenience function; recursive directory scanning; OntologyData dataclass; integrated into ingest(source_type="ontology").

---

[0.2.3] - 2026-01-20

Added

• Amazon Neptune dev environment — CloudFormation template; cfn-lint in pre-commit.
• Vector Store high-performance ingestion — VectorStore.add_documents() with batching and parallel processing (max_workers=6); VectorStore.embed_batch() helper.
• LLM relation extraction tests (mocked and Groq integration).

Changed

• Simplified relation extraction parameter interface; improved error handling and verbose logging.
• Standardized VectorStore concurrency defaults; implicit max_workers=6 in examples.

Fixed

• LLM Relation Extraction Parsing — normalized typed responses to consistent dict format before parsing; structured JSON fallback; extra kwargs removed from internals.
• Pipeline Circular Import (Issues #192 #193) — lazy-loaded PipelineValidator inside PipelineBuilder.__init__; TYPE_CHECKING guard.
• JupyterLab Progress (Issue #181) — SEMANTICA_DISABLE_JUPYTER_PROGRESS env var suppresses rich progress tables.

---

[0.2.2] - 2026-01-15

Added

• Parallel Extraction Engine — concurrent.futures.ThreadPoolExecutor across all extractors (NERExtractor, RelationExtractor, TripletExtractor, EventDetector, SemanticNetworkExtractor); max_workers parameter; thread-safe ProgressTracker.
• Semantic extract regression suite; real-use-case benchmark script.

Changed

• Gemini SDK Migration — google-genai SDK with google.generativeai fallback.
• Pinned opentelemetry-api/-sdk to 1.37.0; updated protobuf/grpcio constraints.
• Entity filtering applied only to LLM prompt construction, not non-LLM flows.
• Raised global optimization.max_workers default to 8.

Security

• Credential sanitization — hardcoded API keys removed from 8 notebooks; ExtractionCache excludes api_key/token/password from cache keys; cache key hashing upgraded MD5 → SHA-256.

Performance

• ~1.89× speedup via parallel extraction (Groq llama-3.3-70b-versatile, standard datasets).
• Optimized entity matching: exact/substring/word-boundary fast paths before embedding similarity.

---

[0.2.1] - 2026-01-12

Fixed

• LLM Output Stability (Bug #176) — correct max_tokens propagation; automatic chunk-halving and retry on context/output limit errors.
• Removed hardcoded max_length constraints from Entity, Relation, Triplet.
• Orchestrator lazy property initialization and configuration normalization.
• AssertionError in orchestrator tests (mock alignment).
• Pinned protobuf>=5.29.1,<7.0, grpcio>=1.71.2; added GitPython and chardet to pyproject.toml.

Changed

• Increased default max_text_length to 64 000 characters for all major providers.
• Standardized Groq defaults: llama-3.3-70b-versatile, 64 k context, native max_tokens/max_completion_tokens.

---

[0.2.0] - 2026-01-10

Added

• Amazon Neptune Support — AmazonNeptuneStore via Bolt/OpenCypher; NeptuneAuthTokenManager with AWS IAM SigV4 signing; retry/backoff. pip install semantica[graph-amazon-neptune].
• Docling Integration — DoclingParser for PDF/DOCX/PPTX/XLSX/HTML/image parsing; OCR support; Markdown/HTML/JSON export.
• Robust Extraction Fallbacks — ML/LLM → Pattern → Last Resort chains across all extractors.
• Provenance & Tracking — batch_index and document_id metadata on all extracted items.
• Semantic Extract — auto-chunking for long text; silent_fail parameter; JSON parsing with 3-attempt exponential backoff.
• End-to-end KG pipeline integration tests; TextEmbedder model switching tests.

Changed

• Removed internal dedup logic from extractors (deferred to semantica/conflicts).
• Standardized batch processing across all extractors using unified extract/analyze/resolve pattern.
• Clarified weighted confidence scoring (50% Method Confidence + 50% Type Similarity).

Fixed

• NameError in extraction_validator.py (missing Union import).
• Extractors returning empty lists for valid input when primary methods fail.
• Model switching bug in TextEmbedder (state not cleared on model switch). (Issue #160)
• TypeError: unhashable type: 'Entity' in GraphAnalyzer. (Issue #159)
• Pinned protobuf==4.25.3, grpcio==1.67.1.
• TripletExtractor.validate_triplets shadowed by internal attribute.
• Incorrect TextSplitter import path.

---

[0.1.1] - 2026-01-05

Added

• Exported DoclingParser and DoclingMetadata from semantica.parse.
• Windows-specific troubleshooting note for PyTorch DLL issues.

Fixed

• DoclingParser import/export across platforms (Windows, Linux, Google Colab).
• Error messaging when optional docling dependency is missing.
• Versioning inconsistencies across the framework.

---

[0.1.0] - 2025-12-31

Added

• Command-line interface (semantica CLI) with knowledge base building and info commands.
• FastAPI-based REST API server for remote access.
• Background worker component for scalable task processing.
• Framework-level versioning configuration for PyPI distribution.
• Automated release workflow with Trusted Publishing support.

Changed

• Updated versioning across the framework to 0.1.0.
• Refined entry point configurations in pyproject.toml.
• Improved lazy module loading for core components.

---

[0.0.5] - 2025-11-26

Changed

• Configured Trusted Publishing for secure automated PyPI deployments.

---

[0.0.4] - 2025-11-26

Changed

• Fixed PyPI deployment issues from v0.0.3.

---

[0.0.3] - 2025-11-25

Added

• Comprehensive issue templates (Bug, Feature, Documentation, Support, Grant/Partnership).
• Updated pull request template with clear guidelines.
• Community support documentation (SUPPORT.md).
• Funding and sponsorship configuration (FUNDING.yml).
• 10+ domain-specific cookbook examples (Finance, Healthcare, Cybersecurity, etc.).

Changed

• Simplified CI/CD workflows — removed failing tests and strict linting.
• Combined release and PyPI publishing into single workflow.
• Simplified security scanning to weekly pip-audit only.

Removed

• Redundant scripts folder (8 shell/PowerShell scripts).
• Unnecessary automation workflows (label-issues, mark-answered).
• Excessive issue templates.

---

[0.0.2] - 2025-11-25

Changed

• Updated README with streamlined content and better examples.
• Added more notebooks to cookbook.
• Improved documentation structure.

---

[0.0.1] - 2024-01-XX

Added

• Core framework architecture.
• Universal data ingestion (multiple file formats).
• Semantic intelligence engine (NER, relation extraction, event detection).
• Knowledge graph construction with entity resolution.
• 6-stage ontology generation pipeline.
• GraphRAG engine for hybrid retrieval.
• Multi-agent system infrastructure.
• Production-ready quality assurance modules.
• Comprehensive documentation with MkDocs.
• Cookbook with interactive tutorials.
• Multiple vector store backends (Weaviate, Qdrant, FAISS).
• Multiple graph database backends (Neo4j, NetworkX, RDFLib).
• Temporal knowledge graph support.
• Conflict detection and resolution; deduplication and entity merging.
• Schema template enforcement; seed data management.
• Multi-format export (RDF, JSON-LD, CSV, GraphML).
• Visualization tools; pipeline orchestration.
• Streaming support (Kafka, RabbitMQ, Kinesis).
• Context engineering for AI agents; reasoning and inference engine.

---

Types of Changes

Label	Meaning
Added	New features
Changed	Changes in existing functionality
Deprecated	Soon-to-be removed features
Removed	Removed features
Fixed	Bug fixes
Security	Vulnerability fixes
Performance	Performance improvements

---

For detailed release notes, see GitHub Releases.