mengzhx/semantica
1
0
Fork 0
mirror of https://github.com/semantica-agi/semantica.git synced 2026-08-29 04:26:20 +00:00
Code Issues Packages Projects Releases Wiki Activity

feat(provenance): close PROV-O compliance gaps and high-stakes trust blockers (closes #825)

Browse Source 
Part A - high-stakes trust blockers:
- Invalidation tombstones via ProvenanceManager.invalidate() (archive-then-append,
  never mutates or deletes) instead of hard delete
- Hash-chained integrity: sequence_id/previous_checksum chain every entry to its
  predecessor; new verify_chain() detects wholesale row deletion that a lone
  per-row checksum cannot
- Typed Agent (AgentRecord: agent_type/is_automated) and Activity (ActivityRecord:
  start/end timing), wired through all 18 *_provenance.py wrappers
- Split parent_entity_id into previous_version_id (correction) vs derived_from_id
  (cross-source derivation), additive alongside the legacy combined field
- Downstream/descendant lineage traversal (get_descendants/trace_descendants,
  reverse BFS) closing the dead direction="downstream" code path in the
  Explorer's provenance route
- Qualified Association+hadRole and Invalidation in export_prov()
- New CLI: provenance invalidate|verify-chain|descendants

Part B - general PROV-O spec completeness:
- Qualified Generation/Usage/Derivation in export_prov()
- wasAssociatedWith, actedOnBehalfOf, wasInformedBy relations
- Bitemporal fields (valid_from/valid_until/revision_type/supersedes) plus
  revision_history()/query_recorded_between(), closing the deprecated
  kg.ProvenanceTracker's "no direct equivalent yet" migration gaps
- prov:Bundle/hadMember membership via bundle_id
- Configurable base_uri (--base-uri CLI flag), shared by RDFExporter's
  NamespaceManager and OWLExporter's default ontology_uri so KG/OWL/PROV
  exports co-resolve under one namespace instead of three hardcoded ones

Bugs fixed along the way:
- agent_id was a dead field: no track_* method read it from kwargs
- track_entities_batch silently absorbed typed kwargs into the metadata blob
- compute_checksum() had to exclude entity_id itself: hashing it made
  track_entity's versioning-archive relabel permanently orphan any entry
  already chained from the pre-relabel checksum, a false-positive "broken
  chain" for a legitimate rename
- InMemoryStorage.get_chain_head() ignored the committed head whenever the
  current transaction had staged entries, corrupting the next chain link
- several new ProvenanceEntry fields were wired into the dataclass and
  export_prov() but not into SQLiteStorage's DDL/INSERT/row-mapping;
  InMemoryStorage masked the gap. Added a permanent round-trip regression
  test to catch this class of bug for future field additions

Flagged, not fixed (separate pre-existing issues, out of scope for #825):
- pipeline/pipeline_provenance.py imports a nonexistent module and wraps a
  Pipeline dataclass with no run() method
- most *_provenance.py wrappers' backing classes are themselves missing or
  incomplete (context_manager, deduplicator, normalizer, etc.)
- kg_provenance.py passes entity_type inside metadata={} instead of as a
  top-level track_entity() kwarg across most of its call sites
This commit is contained in:
KaifAhmad1
2026-08-03 22:28:48 +05:30
committed by Mohd Kaif
parent b59211ea7f
commit db4361ad46
33 changed files with 3104 additions and 313 deletions
Download Patch File Download Diff File Expand all files Collapse all files
CHANGELOG.md
+17
View File
@@ -11,6 +11,23 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
### Added
- **PROV-O trust blockers and general spec completeness for `ProvenanceManager`** (#825) by @KaifAhmad1
- **Invalidation instead of hard delete**: new `ProvenanceManager.invalidate(entity_id, agent_id, reason=None)` tombstones an entry — archives its pre-invalidation state under a stable versioned key, then appends the invalidated entry (`invalidated`, `invalidated_at_time`, `invalidated_by`, `invalidation_reason`) — instead of mutating or deleting it, so an audit can prove a fact existed, was reviewed, and was retracted. `ProvenanceManager.clear()` remains the bulk dev/test store-reset utility it always was; it was not repurposed
- **Hash-chained integrity**: every entry now carries `sequence_id`/`previous_checksum`, chaining it to the entry immediately before it in insertion order. New `ProvenanceManager.verify_chain()` walks the chain and reports any break, including a row hard-deleted directly from the underlying table — something a lone per-row SHA-256 checksum can never detect on its own. `compute_checksum()` now also covers `agent_id`/`agent_type`, the lineage-link fields, and the invalidation fields, closing several fields that previously weren't tamper-evident
- **Typed Agent/Activity**: `agent_id` was a dead field — no `track_*` method read it from kwargs, so it was always the `"semantica"` default regardless of what callers passed; fixed, and paired with new `AgentRecord(id, agent_type, is_automated)` / `ActivityRecord(id, activity_type, started_at_time, ended_at_time)` dataclasses (pass via `agent=`/`activity=` kwargs) so a human reviewer, an LLM call, and an automated pipeline stage are now distinguishable, and activities carry real start/end timing. Wired through all 18 `*_provenance.py` wrapper modules and `track_entity`/`track_relationship`/`track_chunk`/`track_property_source`
- **Versioning vs. derivation split**: new `previous_version_id` ("this corrects a prior version of the same fact") and `derived_from_id` ("this was derived from a different source entity") fields, additive alongside the legacy combined `parent_entity_id` so existing readers are unaffected
- **Downstream lineage traversal**: new `get_descendants()`/`trace_descendants()` (reverse BFS in both `InMemoryStorage` and `SQLiteStorage`), closing the gap flagged in `semantica/explorer/routes/provenance.py` where `direction="downstream"` was dead code with no reverse lookup to feed it; the Explorer's `/api/provenance` lineage response now merges both directions
- **W3C PROV-O qualified relations**: `export_prov()` now emits `prov:qualifiedAssociation`/`hadRole` (distinguishing "approved by" from "generated by" for sign-off workflows), `qualifiedGeneration`/`Generation`, `qualifiedUsage`/`Usage`, `qualifiedDerivation`/`Derivation`, `qualifiedInvalidation`/`Invalidation`, `wasAssociatedWith` (Activity→Agent), `actedOnBehalfOf` (Agent→Agent delegation), and `wasInformedBy` (Activity→Activity, via a new `informed_by=[...]` kwarg), alongside the existing plain triples
- **Bitemporal + Bundle support**: `revision_type`/`supersedes`/`valid_from`/`valid_until` fields (plain caller-supplied passthrough, matching the deprecated `kg.ProvenanceTracker`'s actual contract) plus new `revision_history()` and `query_recorded_between()` methods, closing the two "no direct equivalent yet" rows in `docs/migration/kg-provenance-tracker.md`; `bundle_id` emits `prov:Bundle`/`hadMember` membership triples to partition provenance by source/dataset/ingestion-run
- **Configurable, interlinked namespace**: `export_prov(base_uri=...)` / `--base-uri` CLI flag, defaulting to a new `ProvenanceManager.DEFAULT_BASE_URI` (`https://semantica.dev/ns#`) that `RDFExporter`'s `NamespaceManager` and `OWLExporter`'s default `ontology_uri` now both reuse, so KG-exported, OWL-exported, and PROV-exported URIs for the same `entity_id` co-resolve instead of three independently-hardcoded placeholder domains
- New CLI commands: `semantica provenance invalidate|verify-chain|descendants`
- **Fixed along the way**: `track_entities_batch()` silently absorbed batch-level typed kwargs (`agent_id`, `entity_type`, `activity_id`) into the opaque `metadata` JSON blob instead of forwarding them, so the documented banking example in `docs/guides/provenance.md` never actually worked as written
- **Fixed along the way**: `compute_checksum()` had to exclude `entity_id` itself from the hash — `track_entity()`'s versioning archives a prior value by copying it to a new key (`"X"` → `"X:v:<timestamp>"`), and hashing `entity_id` meant that legitimate relabel permanently orphaned any other entry that had already chained its `previous_checksum` from the pre-relabel value, surfacing as a false-positive "broken chain." Archival and invalidation are now always a pure relabel (unchanged checksum/sequence position) followed by a fresh chained append, never an in-place mutation of an already-chained entry
- **Fixed along the way**: `InMemoryStorage.get_chain_head()` ignored the already-committed chain head whenever the current transaction had staged any entries, understating the head and corrupting the next append's chain link
- **Fixed along the way**: several new `ProvenanceEntry` fields were initially wired into the dataclass and `export_prov()` but not into `SQLiteStorage`'s DDL/INSERT/row-mapping — `InMemoryStorage` stores the dataclass directly so it masked the gap. Added a permanent regression test (`test_all_fields_round_trip_through_sqlite`) asserting every field survives a SQLite round trip, to catch this class of bug for any future field additions
- Flagged, not fixed (separate, pre-existing issues independent of #825): `semantica/pipeline/pipeline_provenance.py` imports a nonexistent module and wraps a `Pipeline` dataclass with no `run()` method, so `PipelineWithProvenance` has never worked; most of the 18 wrapper modules' backing classes are themselves missing or incomplete (e.g. `context.context_manager`, `deduplication.deduplicator`, `normalize.normalizer` don't exist; `EmbeddingGenerator` exists but has no `.embed()`); `kg_provenance.py` passes `entity_type` inside its `metadata={}` dict instead of as a top-level `track_entity()` kwarg across most of its ~30 call sites, so it never actually populates the real field
- Extensive new test coverage across `tests/provenance/test_manager.py`, `test_schemas.py`, and `test_storage.py` (invalidation, hash-chain verification including a simulated hard-delete-detection case and an interleaved-chaining stress test, agent/activity typing, versioning/derivation split, downstream lineage, qualified export triples, bitemporal methods, Bundle export, and namespace interlinking)
- **Altair Anzo triplet store backend** (#813) by @KaifAhmad1
- Added `AnzoStore` (`semantica/triplet_store/anzo_store.py`), a fourth peer to `BlazegraphStore`/`RDF4JStore`/`JenaStore` speaking plain SPARQL 1.1 over HTTP — no new dependency, since Anzo has no official Python SDK but needs none
- The one structural difference from the existing backends: Anzo addresses data by a dataset/graphmart **URI** (`dataset_uri`, required) rather than a short namespace/repository name, so the endpoint path (`<endpoint>/sparql/<store_type>/<url-encoded_dataset_uri>`) percent-encodes it; `store_type` defaults to `"graphmart"` and can be set to `"dataset"`
docs/guides/provenance.md
+20 -3
View File
@@ -633,12 +633,29 @@ Every `ProvenanceEntry` maps directly to W3C PROV-O terms. If your compliance te
| :--- | :--- | :--- |
| `prov:Entity` | `entity_id` | The tracked object — entity, chunk, relationship, or property |
| `prov:Activity` | `activity_id` | The process that produced it — `"ner_extraction"`, `"bureau_parsing"` |
| `prov:Agent` | `agent_id` | Who ran the activity — pipeline name, analyst ID |
| `prov:wasDerivedFrom` | `parent_entity_id` | The previous version of this entity — enables version chaining |
| `prov:Agent` / `prov:Person` / `prov:SoftwareAgent` / `prov:Organization` | `agent_id`, `agent_type`, `is_automated` | Who — or what — ran the activity, and whether a human was directly accountable |
| `prov:qualifiedAssociation` + `prov:hadRole` | `role` | The agent's role for this specific entity — `"generator"` (default), `"approver"`, `"reviewer"` — for sign-off/four-eyes workflows |
| `prov:wasDerivedFrom` | `parent_entity_id` (legacy combined field) | The previous version or source of this entity |
| — | `previous_version_id` | This entry corrects/replaces a prior version of the *same* fact |
| `prov:wasDerivedFrom` | `derived_from_id` | This entry was derived from a *different* source entity |
| `prov:used` | `used_entities` | Entity IDs consumed to produce this one |
| `prov:generatedAtTime` | `timestamp` | ISO datetime, auto-set to `datetime.utcnow()` at write time |
| `prov:qualifiedInvalidation` | `invalidated`, `invalidated_at_time`, `invalidated_by`, `invalidation_reason` | A retraction/correction recorded as a tombstone via `ProvenanceManager.invalidate()`, never a hard delete |
| `prov:startedAtTime` / `prov:endedAtTime` | `activity_started_at_time`, `activity_ended_at_time` | Typed Activity timing — pass an `ActivityRecord` via the `activity=` kwarg to set these together with `activity_id` |
| `prov:qualifiedGeneration`/`Generation`, `qualifiedUsage`/`Usage`, `qualifiedDerivation`/`Derivation` | (derived from the fields above) | Additive qualified forms of `wasGeneratedBy`/`used`/`wasDerivedFrom`, emitted automatically alongside the plain triples |
| `prov:wasAssociatedWith` | (derived from `agent_id`) | Direct Activity→Agent link, distinct from the Entity→Agent `wasAttributedTo` |
| `prov:actedOnBehalfOf` | `acted_on_behalf_of` | Agent→Agent delegation — e.g. an automated agent acting on behalf of the human/organization that authorized it |
| `prov:wasInformedBy` | `informed_by_activities` (pass as `informed_by=[...]`) | Chains this entry's activity to prior activities it was informed by (e.g. a pipeline stage informed by the stage before it) |
| `prov:Bundle` + `prov:hadMember` | `bundle_id` | Groups entries by source/dataset/ingestion-run (membership triples, not true RDF named-graph partitioning) |
| — | `valid_from`, `valid_until`, `revision_type`, `supersedes` | Bitemporal fields merged from the deprecated `kg.ProvenanceTracker` — always caller-supplied (never auto-computed), surfaced via `ProvenanceManager.revision_history()`, which falls back to timestamp-based derivation for entries that don't set them explicitly |
The `checksum` field is not part of the PROV-O standard — it is Semantica's tamper-detection extension. Every entry's SHA-256 is computed from its content fields at write time and can be recomputed at any time to verify the record has not been modified.
`previous_version_id` and `derived_from_id` are additive alongside `parent_entity_id` — existing code reading `parent_entity_id` keeps working unchanged, while new code gets the two relations disambiguated.
The `checksum` field is not part of the PROV-O standard — it is Semantica's tamper-detection extension. Every entry's SHA-256 now also incorporates `previous_checksum` (the prior entry's checksum, by insertion order via `sequence_id`), chaining every entry to the one before it. `ProvenanceManager.verify_chain()` walks the full chain and reports any break — including a row that was hard-deleted from the underlying table, which a lone per-row checksum can't detect on its own.
Note: the banking example above passes `agent_id="credit_data_service_v2"` to `track_entities_batch()` — this now actually populates the entry's `agent_id` field (previously a bug caused batch-level typed kwargs like `agent_id`/`entity_type`/`activity_id` to be silently absorbed into the opaque `metadata` blob instead).
`export_prov()` mints entity/agent/activity URIs under `ProvenanceManager.DEFAULT_BASE_URI` (`https://semantica.dev/ns#` by default — the same namespace `RDFExporter`'s `NamespaceManager` uses for its `"semantica"` prefix, so KG-exported and PROV-exported URIs for the same `entity_id` co-resolve) unless overridden via `export_prov(base_uri=...)` or the CLI's `--base-uri` option.
## Related Guides
docs/migration/kg-provenance-tracker.md
+2 -2
View File
@@ -17,8 +17,8 @@ Every method on `kg.ProvenanceTracker` now emits a `DeprecationWarning` on use,
| `track_entity(entity_id, source, metadata)` | `track_entity(entity_id, source, metadata)` | Same call shape. `ProvenanceManager` additionally auto-links each update to its prior version via `parent_entity_id`. |
| `get_all_sources(entity_id)` | `get_all_sources(entity_id)` | Field name differs: the `kg` tracker returns each record's time under `"recorded_at"`; `ProvenanceManager` returns `"timestamp"`. |
| `clear(entity_id=None)` | `clear()` | `ProvenanceManager.clear()` clears all provenance data; there is no per-entity clear yet. |
| `query_recorded_between(start, end)` | *No direct equivalent yet* | Filter the entries returned by `get_lineage()` / `trace_lineage()` client-side in the meantime. |
| `revision_history(fact_id)` | *No direct equivalent yet* | `get_lineage(fact_id)["lineage_chain"]` returns the full chain of `ProvenanceEntry` records but not in the same versioned shape. |
| `query_recorded_between(start, end)` | `query_recorded_between(start, end)` | Same call shape; filters by `timestamp` (ISO 8601 string comparison) across all tracked entries, not just one entity. |
| `revision_history(fact_id)` | `revision_history(fact_id)` | Same call shape and return shape (`version`, `valid_from`, `valid_until`, `recorded_at`, `author`, optional `revision_type`/`supersedes`) — walks the entity's `previous_version_id` chain rather than a flat per-entity dict. |
| `export_audit_log(fact_ids, format)` | *No direct equivalent yet* | Build the export from `get_lineage()` output, or serialize `get_statistics()` for a summary view. |
Methods with no direct equivalent are not planned to be reimplemented on `kg.ProvenanceTracker` — they will need a small adapter in caller code, or a feature request against `ProvenanceManager` if you rely on them heavily.
semantica/cli.py
+117 -4
View File
@@ -2559,27 +2559,31 @@ def provenance_audit(cli_ctx: CLIContext, since: Optional[str], fmt: str,
@provenance.command("export")
@click.option("--format", "fmt", type=click.Choice(["turtle", "ntriples", "jsonld"]),
default="turtle", show_default=True)
@click.option("--base-uri", "base_uri", default=None,
help="Namespace URI entities/agents/activities are minted under "
"(default: ProvenanceManager.DEFAULT_BASE_URI).")
@click.option("--output", default=None, type=click.Path())
@click.option("--dry-run", "local_dry", is_flag=True, default=False)
@click.pass_obj
def provenance_export(cli_ctx: CLIContext, fmt: str, output: Optional[str],
local_dry: bool) -> None:
def provenance_export(cli_ctx: CLIContext, fmt: str, base_uri: Optional[str],
output: Optional[str], local_dry: bool) -> None:
"""Export provenance as W3C PROV-O RDF.
\b
Example:
semantica provenance export --format turtle --output prov.ttl
semantica provenance export --base-uri https://example.org/kg# --output prov.ttl
"""
cli_ctx = _require_ctx(cli_ctx)
def _action() -> None:
if _is_dry(cli_ctx, local_dry):
_dry(cli_ctx, "export provenance", format=fmt, output=output)
_dry(cli_ctx, "export provenance", format=fmt, base_uri=base_uri, output=output)
return
try:
from .provenance import ProvenanceManager
pm = ProvenanceManager(config=cli_ctx.config.to_dict())
data = pm.export_prov(format=fmt)
data = pm.export_prov(format=fmt, base_uri=base_uri)
except ImportError as exc:
raise click.ClickException(f"Provenance module not available: {exc}") from exc
if output:
@@ -2621,6 +2625,115 @@ def provenance_check(cli_ctx: CLIContext, strict: bool, local_json: bool) -> Non
_run_with_error_handling(_action)
@provenance.command("invalidate")
@click.argument("entity_id")
@click.option("--by", "agent_id", required=True, help="Agent responsible for the invalidation.")
@click.option("--reason", default=None, help="Human-readable reason for the invalidation.")
@click.option("--json", "local_json", is_flag=True, default=False)
@click.option("--dry-run", "local_dry", is_flag=True, default=False)
@click.pass_obj
def provenance_invalidate(cli_ctx: CLIContext, entity_id: str, agent_id: str,
reason: Optional[str], local_json: bool,
local_dry: bool) -> None:
"""Mark a tracked entity as invalidated (tombstone, not a hard delete).
\b
Example:
semantica provenance invalidate entity_alice --by reviewer_jane --reason "Source retracted"
"""
cli_ctx = _require_ctx(cli_ctx)
def _action() -> None:
if _is_dry(cli_ctx, local_dry):
_dry(cli_ctx, "invalidate provenance entity", entity_id=entity_id, by=agent_id, reason=reason)
return
try:
from .provenance import ProvenanceManager
pm = ProvenanceManager(config=cli_ctx.config.to_dict())
result = pm.invalidate(entity_id, agent_id=agent_id, reason=reason)
except ImportError as exc:
raise click.ClickException(f"Provenance module not available: {exc}") from exc
except ValueError as exc:
raise click.ClickException(str(exc)) from exc
if _is_json(cli_ctx, local_json):
_jecho(result.to_dict())
else:
_ok(cli_ctx, f"Invalidated {entity_id} (by {agent_id})")
_run_with_error_handling(_action)
@provenance.command("verify-chain")
@click.option("--json", "local_json", is_flag=True, default=False)
@click.pass_obj
def provenance_verify_chain(cli_ctx: CLIContext, local_json: bool) -> None:
"""Verify the hash chain across all provenance entries.
Detects wholesale row deletion: per-row checksums alone only prove a
surviving row wasn't edited in place, not that no row is missing.
\b
Example:
semantica provenance verify-chain
"""
cli_ctx = _require_ctx(cli_ctx)
def _action() -> None:
try:
from .provenance import ProvenanceManager
pm = ProvenanceManager(config=cli_ctx.config.to_dict())
result = pm.verify_chain()
except ImportError as exc:
raise click.ClickException(f"Provenance module not available: {exc}") from exc
if _is_json(cli_ctx, local_json):
_jecho(result)
elif result.get("valid"):
_ok(cli_ctx, f"Chain verified: {result.get('total_entries')} entries, no breaks")
else:
_warn(cli_ctx, f"Chain verification failed: {len(result.get('broken_links', []))} broken link(s)")
for link in result.get("broken_links", []):
click.echo(f" {link}")
_run_with_error_handling(_action)
@provenance.command("descendants")
@click.argument("entity_id")
@click.option("--depth", default=None, type=int, show_default=True)
@click.option("--json", "local_json", is_flag=True, default=False)
@click.pass_obj
def provenance_descendants(cli_ctx: CLIContext, entity_id: str, depth: Optional[int],
local_json: bool) -> None:
"""Show downstream descendants (reverse lineage) for an entity.
The counterpart to `provenance lineage`, which only traces upstream
ancestors. Answers "entity X was wrong — what downstream facts used it?"
\b
Example:
semantica provenance descendants entity_alice --depth 3
"""
cli_ctx = _require_ctx(cli_ctx)
def _action() -> None:
try:
from .provenance import ProvenanceManager
pm = ProvenanceManager(config=cli_ctx.config.to_dict())
if depth is not None:
entries = [e.to_dict() for e in pm.trace_descendants(entity_id, max_depth=depth)]
result = {"entity_id": entity_id, "depth": depth, "entries": entries}
else:
result = pm.get_descendants(entity_id) or {"entity_id": entity_id, "entries": []}
except ImportError as exc:
raise click.ClickException(f"Provenance module not available: {exc}") from exc
if _is_json(cli_ctx, local_json):
_jecho(result)
else:
_pprint(cli_ctx, result)
_run_with_error_handling(_action)
@main.group(invoke_without_command=True)
@click.pass_context
def validate(ctx: click.Context) -> None:
semantica/conflicts/conflicts_provenance.py
+17 -1
View File
@@ -15,15 +15,24 @@ License: MIT
"""
from typing import Optional, Dict, Any, List
from datetime import datetime
class SourceTrackerWithUnifiedBackend:
"""SourceTracker using unified provenance backend."""
def __init__(self, **config):
def __init__(
self,
agent_id: Optional[str] = None,
is_automated: bool = True,
**config,
):
"""Initialize with unified backend or fallback to legacy."""
from .source_tracker import SourceTracker
self._agent_id = agent_id or self.__class__.__name__
self._is_automated = is_automated
try:
from semantica.provenance import ProvenanceManager
self._unified_manager = ProvenanceManager()
@@ -35,6 +44,7 @@ class SourceTrackerWithUnifiedBackend:
def track_property_source(self, entity_id: str, property_name: str, value: Any, source: Any, **metadata):
"""Track property source with unified backend."""
activity_started_at_time = datetime.utcnow().isoformat()
if self._use_unified:
from semantica.provenance import SourceReference
@@ -45,6 +55,12 @@ class SourceTrackerWithUnifiedBackend:
confidence=getattr(source, 'confidence', 1.0)
)
metadata.setdefault("agent_id", self._agent_id)
metadata.setdefault("agent_type", "software_agent")
metadata.setdefault("is_automated", self._is_automated)
metadata.setdefault("activity_started_at_time", activity_started_at_time)
metadata.setdefault("activity_ended_at_time", datetime.utcnow().isoformat())
self._unified_manager.track_property_source(
entity_id=entity_id,
property_name=property_name,
semantica/context/context_provenance.py
+17 -1
View File
@@ -12,19 +12,28 @@ License: MIT
"""
from typing import Optional, Any
from datetime import datetime
import uuid
class ContextManagerWithProvenance:
"""Context manager with provenance tracking."""
def __init__(self, provenance: bool = False, **config):
def __init__(
self,
provenance: bool = False,
agent_id: Optional[str] = None,
is_automated: bool = True,
**config,
):
"""Initialize context manager with optional provenance."""
from .context_manager import ContextManager
self.provenance = provenance
self._context_manager = ContextManager(**config)
self._prov_manager = None
self._agent_id = agent_id or self.__class__.__name__
self._is_automated = is_automated
if provenance:
try:
@@ -35,13 +44,20 @@ class ContextManagerWithProvenance:
def add_context(self, context: Any, source: Optional[str] = None, **kwargs):
"""Add context with provenance tracking."""
activity_started_at_time = datetime.utcnow().isoformat()
result = self._context_manager.add_context(context, **kwargs)
activity_ended_at_time = datetime.utcnow().isoformat()
if self.provenance and self._prov_manager:
self._prov_manager.track_entity(
entity_id=f"context_{uuid.uuid4().hex[:8]}",
source=source or "context_manager",
entity_type="context",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=activity_started_at_time,
activity_ended_at_time=activity_ended_at_time,
metadata={"context_preview": str(context)[:100]}
)
semantica/deduplication/deduplication_provenance.py
+18 -2
View File
@@ -13,19 +13,28 @@ Author: Semantica Contributors
License: MIT
"""
from typing import List, Any
from typing import List, Any, Optional
from datetime import datetime
import uuid
class DeduplicatorWithProvenance:
"""Deduplicator with provenance tracking."""
def __init__(self, provenance: bool = False, **config):
def __init__(
self,
provenance: bool = False,
agent_id: Optional[str] = None,
is_automated: bool = True,
**config,
):
from .deduplicator import Deduplicator
self.provenance = provenance
self._deduplicator = Deduplicator(**config)
self._prov_manager = None
self._agent_id = agent_id or self.__class__.__name__
self._is_automated = is_automated
if provenance:
try:
@@ -36,7 +45,9 @@ class DeduplicatorWithProvenance:
def deduplicate(self, items: List[Any], source: str = None, **kwargs):
"""Deduplicate items with provenance tracking."""
activity_started_at_time = datetime.utcnow().isoformat()
unique_items = self._deduplicator.deduplicate(items, **kwargs)
activity_ended_at_time = datetime.utcnow().isoformat()
if self.provenance and self._prov_manager:
duplicates_found = len(items) - len(unique_items)
@@ -44,6 +55,11 @@ class DeduplicatorWithProvenance:
entity_id=f"dedup_{uuid.uuid4().hex[:8]}",
source=source or "deduplication",
entity_type="deduplication_operation",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=activity_started_at_time,
activity_ended_at_time=activity_ended_at_time,
metadata={
"input_count": len(items),
"output_count": len(unique_items),
semantica/embeddings/embeddings_provenance.py
+18 -2
View File
@@ -13,19 +13,28 @@ Author: Semantica Contributors
License: MIT
"""
from typing import List
from typing import List, Optional
from datetime import datetime
import uuid
class EmbeddingGeneratorWithProvenance:
"""Embedding generator with provenance tracking."""
def __init__(self, provenance: bool = False, **config):
def __init__(
self,
provenance: bool = False,
agent_id: Optional[str] = None,
is_automated: bool = True,
**config,
):
from .embedding_generator import EmbeddingGenerator
self.provenance = provenance
self._generator = EmbeddingGenerator(**config)
self._prov_manager = None
self._agent_id = agent_id or self.__class__.__name__
self._is_automated = is_automated
if provenance:
try:
@@ -36,13 +45,20 @@ class EmbeddingGeneratorWithProvenance:
def embed(self, texts: List[str], source: str = None, **kwargs):
"""Generate embeddings with provenance tracking."""
activity_started_at_time = datetime.utcnow().isoformat()
embeddings = self._generator.embed(texts, **kwargs)
activity_ended_at_time = datetime.utcnow().isoformat()
if self.provenance and self._prov_manager:
self._prov_manager.track_entity(
entity_id=f"embed_{uuid.uuid4().hex[:8]}",
source=source or "embedding_generation",
entity_type="embeddings",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=activity_started_at_time,
activity_ended_at_time=activity_ended_at_time,
metadata={
"model": getattr(self._generator, 'model', 'unknown'),
"dimensions": len(embeddings[0]) if embeddings else 0,
semantica/explorer/routes/provenance.py
+55 -30
View File
@@ -32,21 +32,7 @@ def _classify_prov(node_type: str) -> tuple[str, str]:
return "Entity", "group_entity"
def _transform_audit_lineage(lineage: Dict[str, Any], node_id: str) -> Dict[str, Any]:
# Mapping decision (W3C PROV-O to frontend swim-lanes):
# Every ProvenanceEntry maps to a node classified by _classify_prov(entry["entity_type"]),
# placing documents/chunks/entities in 'group_entity' (prov_type='Entity'), persons/systems
# in 'group_agent', and actions/processes in 'group_activity'.
# For derivation relationships (parent_entity_id and used_entities), we connect
# parent -> child directly with edge label set to activity_id (or 'wasDerivedFrom'),
# keeping the lineage graph scannable without cluttering it with intermediate activity
# nodes when activity_id is an operational label.
nodes: List[Dict[str, Any]] = []
edges: List[Dict[str, Any]] = []
seen_nodes = set()
seen_edges = set()
chain = lineage.get("lineage_chain") or lineage.get("entries") or []
def _add_chain_nodes(chain: List[Any], nodes: List[Dict[str, Any]], seen_nodes: set) -> None:
for entry in chain:
if not isinstance(entry, dict):
continue
@@ -69,6 +55,10 @@ def _transform_audit_lineage(lineage: Dict[str, Any], node_id: str) -> Dict[str,
"checksum": entry.get("checksum") or None,
})
def _add_chain_edges(
chain: List[Any], edges: List[Dict[str, Any]], seen_edges: set, direction: str
) -> None:
for entry in chain:
if not isinstance(entry, dict):
continue
@@ -86,21 +76,11 @@ def _transform_audit_lineage(lineage: Dict[str, Any], node_id: str) -> Dict[str,
activity = str(entry.get("activity_id") or "wasDerivedFrom")
for src in parents:
edge_key = (src, eid)
edge_key = (src, eid, direction)
if edge_key in seen_edges:
continue
seen_edges.add(edge_key)
# NOTE: ProvenanceManager.get_lineage() currently only traces upstream
# ancestor chains via parent_entity_id and used_entities. It does not perform
# reverse lookups for downstream descendants. Consequently, 'direction = "downstream"'
# is unreachable in practice for this audit path until reverse lookup is supported
# by ProvenanceManager. All ancestor derivation edges are upstream lineage.
if src == node_id:
direction = "downstream"
else:
direction = "upstream"
edges.append({
"id": f"{src}-{eid}",
"source": src,
@@ -109,6 +89,44 @@ def _transform_audit_lineage(lineage: Dict[str, Any], node_id: str) -> Dict[str,
"direction": direction,
})
def _transform_audit_lineage(
lineage: Dict[str, Any],
node_id: str,
descendants: Optional[Dict[str, Any]] = None,
) -> Dict[str, Any]:
# Mapping decision (W3C PROV-O to frontend swim-lanes):
# Every ProvenanceEntry maps to a node classified by _classify_prov(entry["entity_type"]),
# placing documents/chunks/entities in 'group_entity' (prov_type='Entity'), persons/systems
# in 'group_agent', and actions/processes in 'group_activity'.
# For derivation relationships (parent_entity_id and used_entities), we connect
# parent -> child directly with edge label set to activity_id (or 'wasDerivedFrom'),
# keeping the lineage graph scannable without cluttering it with intermediate activity
# nodes when activity_id is an operational label.
#
# Upstream edges come from lineage's ancestor chain (parent_entity_id/
# used_entities, via ProvenanceManager.get_lineage()); downstream edges
# come from ProvenanceManager.get_descendants()'s descendant chain (issue
# #825, Part A item 5). Previously 'direction="downstream"' was dead code
# here since no reverse lookup existed.
nodes: List[Dict[str, Any]] = []
edges: List[Dict[str, Any]] = []
seen_nodes: set = set()
seen_edges: set = set()
ancestor_chain = lineage.get("lineage_chain") or lineage.get("entries") or []
descendant_chain = (
(descendants or {}).get("descendant_chain")
or (descendants or {}).get("entries")
or []
)
_add_chain_nodes(ancestor_chain, nodes, seen_nodes)
_add_chain_nodes(descendant_chain, nodes, seen_nodes)
_add_chain_edges(ancestor_chain, edges, seen_edges, "upstream")
_add_chain_edges(descendant_chain, edges, seen_edges, "downstream")
for edge in edges:
for endpoint in (edge["source"], edge["target"]):
if endpoint not in seen_nodes:
@@ -132,9 +150,9 @@ def _transform_audit_lineage(lineage: Dict[str, Any], node_id: str) -> Dict[str,
def _build_provenance(session: GraphSession, node_id: Optional[str] = None) -> dict:
"""Build provenance lineage for a node, attempting the audit-grade store first.
NOTE: The audit path (source='audit') shows verified upstream lineage only.
For descendant/downstream relationships, the naive graph-traversal fallback
remains the only source until ProvenanceManager gains a reverse lookup.
Combines ProvenanceManager.get_lineage() (upstream ancestors) with
get_descendants() (downstream — issue #825, Part A item 5) so both
directions are populated from the audit-grade store, not just upstream.
"""
if not node_id:
return {"nodes": [], "edges": [], "source": "graph_traversal"}
@@ -149,7 +167,14 @@ def _build_provenance(session: GraphSession, node_id: Optional[str] = None) -> d
entries = lineage.get("lineage_chain") or lineage.get("entries") or []
integrity_ok = all(verify_checksum(entry) for entry in entries)
if integrity_ok:
return _transform_audit_lineage(lineage, node_id)
descendants = {}
try:
descendants = manager.get_descendants(node_id) or {}
except Exception as exc:
logger.warning(
f"ProvenanceManager get_descendants failed for {node_id}: {exc}"
)
return _transform_audit_lineage(lineage, node_id, descendants)
logger.warning(
f"Provenance integrity verification failed for {node_id}, falling back to graph traversal"
)
semantica/export/export_provenance.py
+18 -2
View File
@@ -13,19 +13,28 @@ Author: Semantica Contributors
License: MIT
"""
from typing import Any
from typing import Any, Optional
from datetime import datetime
import uuid
class ExporterWithProvenance:
"""Base exporter with provenance tracking."""
def __init__(self, provenance: bool = False, **config):
def __init__(
self,
provenance: bool = False,
agent_id: Optional[str] = None,
is_automated: bool = True,
**config,
):
from .exporter import Exporter
self.provenance = provenance
self._exporter = Exporter(**config)
self._prov_manager = None
self._agent_id = agent_id or self.__class__.__name__
self._is_automated = is_automated
if provenance:
try:
@@ -36,13 +45,20 @@ class ExporterWithProvenance:
def export(self, data: Any, destination: str, **kwargs):
"""Export data with provenance tracking."""
activity_started_at_time = datetime.utcnow().isoformat()
result = self._exporter.export(data, destination, **kwargs)
activity_ended_at_time = datetime.utcnow().isoformat()
if self.provenance and self._prov_manager:
self._prov_manager.track_entity(
entity_id=f"export_{uuid.uuid4().hex[:8]}",
source="export_operation",
entity_type="export",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=activity_started_at_time,
activity_ended_at_time=activity_ended_at_time,
metadata={
"destination": destination,
"format": kwargs.get('format', 'unknown')
semantica/export/owl_exporter.py
+9 -2
View File
@@ -31,6 +31,12 @@ from ..utils.helpers import ensure_directory
from ..utils.logging import get_logger
from ..utils.progress_tracker import get_progress_tracker
# Issue #825, Part B Tier 3 — exporter interlinking. Reuses the same default
# namespace as ProvenanceManager.export_prov() and RDFExporter's
# NamespaceManager "semantica" entry, so ontology URIs, KG instance URIs, and
# PROV-exported URIs co-resolve under one shared namespace by default.
from ..provenance.manager import DEFAULT_BASE_URI
class OWLExporter:
"""
@@ -58,7 +64,7 @@ class OWLExporter:
def __init__(
self,
ontology_uri: str = "https://semantica.dev/ontology/",
ontology_uri: str = DEFAULT_BASE_URI,
version: str = "1.0",
format: str = "owl-xml",
config: Optional[Dict[str, Any]] = None,
@@ -70,7 +76,8 @@ class OWLExporter:
Sets up the exporter with ontology URI, version, and format configuration.
Args:
ontology_uri: Base URI for the ontology (default: "https://semantica.dev/ontology/")
ontology_uri: Base URI for the ontology (default: ProvenanceManager.DEFAULT_BASE_URI,
shared with RDFExporter's NamespaceManager and export_prov() so URIs co-resolve)
version: Ontology version string (default: "1.0")
format: Default export format - 'owl-xml' or 'turtle' (default: 'owl-xml')
config: Optional configuration dictionary (merged with kwargs)
semantica/export/rdf_exporter.py
+8 -2
View File
@@ -71,13 +71,19 @@ class NamespaceManager:
"""
self.logger = get_logger("namespace_manager")
# Standard RDF namespaces
# Standard RDF namespaces. "semantica" reuses ProvenanceManager's
# DEFAULT_BASE_URI (issue #825, Part B Tier 3 — exporter
# interlinking) so KG-exported and PROV-exported URIs for the same
# entity_id co-resolve to the same namespace instead of two
# independently-hardcoded placeholder domains.
from ..provenance.manager import DEFAULT_BASE_URI
self.namespaces: Dict[str, str] = {
"rdf": "http://www.w3.org/1999/02/22-rdf-syntax-ns#",
"rdfs": "http://www.w3.org/2000/01/rdf-schema#",
"owl": "http://www.w3.org/2002/07/owl#",
"xsd": "http://www.w3.org/2001/XMLSchema#",
"semantica": "https://semantica.dev/ns#",
"semantica": DEFAULT_BASE_URI,
}
self.config = config or {}
semantica/graph_store/graph_store_provenance.py
+18 -2
View File
@@ -13,19 +13,28 @@ Author: Semantica Contributors
License: MIT
"""
from typing import Any
from typing import Any, Optional
from datetime import datetime
import uuid
class GraphStoreWithProvenance:
"""Graph store with provenance tracking."""
def __init__(self, provenance: bool = False, **config):
def __init__(
self,
provenance: bool = False,
agent_id: Optional[str] = None,
is_automated: bool = True,
**config,
):
from .graph_store import GraphStore
self.provenance = provenance
self._store = GraphStore(**config)
self._prov_manager = None
self._agent_id = agent_id or self.__class__.__name__
self._is_automated = is_automated
if provenance:
try:
@@ -36,7 +45,9 @@ class GraphStoreWithProvenance:
def add_node(self, node: Any, source: str = None, **kwargs):
"""Add node with provenance tracking."""
activity_started_at_time = datetime.utcnow().isoformat()
result = self._store.add_node(node, **kwargs)
activity_ended_at_time = datetime.utcnow().isoformat()
if self.provenance and self._prov_manager:
node_id = getattr(node, 'id', f"node_{uuid.uuid4().hex[:8]}")
@@ -44,6 +55,11 @@ class GraphStoreWithProvenance:
entity_id=node_id,
source=source or "graph_store",
entity_type="graph_node",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=activity_started_at_time,
activity_ended_at_time=activity_ended_at_time,
metadata={"properties": getattr(node, 'properties', {})}
)
semantica/ingest/ingest_provenance.py
+27 -3
View File
@@ -14,15 +14,24 @@ License: MIT
"""
from typing import Optional, List
from datetime import datetime
import uuid
class IngestProvenanceMixin:
"""Mixin for ingest provenance tracking."""
def __init__(self, provenance: bool = False, **kwargs):
def __init__(
self,
provenance: bool = False,
agent_id: Optional[str] = None,
is_automated: bool = True,
**kwargs,
):
self.provenance = provenance
self._prov_manager = None
self._agent_id = agent_id or self.__class__.__name__
self._is_automated = is_automated
if provenance:
try:
@@ -35,15 +44,25 @@ class IngestProvenanceMixin:
class PDFIngestorWithProvenance(IngestProvenanceMixin):
"""PDF ingestor with provenance tracking."""
def __init__(self, provenance: bool = False, **config):
def __init__(
self,
provenance: bool = False,
agent_id: Optional[str] = None,
is_automated: bool = True,
**config,
):
from .pdf_ingestor import PDFIngestor
IngestProvenanceMixin.__init__(self, provenance=provenance)
IngestProvenanceMixin.__init__(
self, provenance=provenance, agent_id=agent_id, is_automated=is_automated
)
self._ingestor = PDFIngestor(**config)
def ingest(self, file_path: str, **kwargs):
"""Ingest PDF with provenance tracking."""
activity_started_at_time = datetime.utcnow().isoformat()
docs = self._ingestor.ingest(file_path, **kwargs)
activity_ended_at_time = datetime.utcnow().isoformat()
if self.provenance and self._prov_manager:
for doc in docs:
@@ -52,6 +71,11 @@ class PDFIngestorWithProvenance(IngestProvenanceMixin):
entity_id=doc_id,
source=file_path,
entity_type="document",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=activity_started_at_time,
activity_ended_at_time=activity_ended_at_time,
metadata={
"file_type": "pdf",
"pages": getattr(doc, 'page_count', None)
semantica/kg/kg_provenance.py
+177 -5
View File
@@ -54,6 +54,7 @@ Version: 1.0.0
"""
from typing import Any, Dict, List, Optional
from datetime import datetime
import uuid
import time
@@ -78,12 +79,20 @@ class GraphBuilderWithProvenance:
})
"""
def __init__(self, provenance: bool = False, **config):
def __init__(
self,
provenance: bool = False,
agent_id: Optional[str] = None,
is_automated: bool = True,
**config,
):
from .graph_builder import GraphBuilder
self.provenance = provenance
self._builder = GraphBuilder(**config)
self._prov_manager = None
self._agent_id = agent_id or self.__class__.__name__
self._is_automated = is_automated
if provenance:
try:
@@ -94,12 +103,18 @@ class GraphBuilderWithProvenance:
def build(self, sources, **kwargs):
"""Build graph with provenance tracking."""
# Track the build operation
activity_started_at_time = datetime.utcnow().isoformat()
# Track the build operation (recorded before the build runs, so it
# has no end time yet — this is the "in progress" marker).
if self.provenance and self._prov_manager:
build_id = f"graph_build_{uuid.uuid4().hex[:8]}"
self._prov_manager.track_entity(
entity_id=build_id,
source="graph_construction",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=activity_started_at_time,
metadata={
"entity_type": "graph_build_operation",
"operation": "build_graph",
@@ -109,6 +124,7 @@ class GraphBuilderWithProvenance:
)
result = self._builder.build(sources, **kwargs)
activity_ended_at_time = datetime.utcnow().isoformat()
# Track individual entities and relationships if available
if self.provenance and self._prov_manager and hasattr(result, 'get'):
@@ -124,6 +140,11 @@ class GraphBuilderWithProvenance:
entity_id=entity_id,
source="graph_construction",
entity_type="graph_entity",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=activity_started_at_time,
activity_ended_at_time=activity_ended_at_time,
metadata={
"operation": "build_entity",
"entity_type": entity.get('type'),
@@ -139,6 +160,11 @@ class GraphBuilderWithProvenance:
entity_id=rel_id,
source="graph_construction",
entity_type="graph_relationship",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=activity_started_at_time,
activity_ended_at_time=activity_ended_at_time,
metadata={
"operation": "build_relationship",
"relationship_type": relationship.get('type'),
@@ -154,12 +180,18 @@ class GraphBuilderWithProvenance:
def build_single_source(self, kg_data, **kwargs):
"""Build graph from single source with provenance tracking."""
# Track the build operation
activity_started_at_time = datetime.utcnow().isoformat()
# Track the build operation (recorded before the build runs, so it
# has no end time yet — this is the "in progress" marker).
if self.provenance and self._prov_manager:
build_id = f"graph_build_single_{uuid.uuid4().hex[:8]}"
self._prov_manager.track_entity(
entity_id=build_id,
source="graph_construction",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=activity_started_at_time,
metadata={
"entity_type": "graph_build_operation",
"operation": "build_single_source",
@@ -168,6 +200,7 @@ class GraphBuilderWithProvenance:
)
result = self._builder.build_single_source(kg_data, **kwargs)
activity_ended_at_time = datetime.utcnow().isoformat()
# Track entities and relationships if available
if self.provenance and self._prov_manager and isinstance(result, dict):
@@ -182,6 +215,11 @@ class GraphBuilderWithProvenance:
entity_id=entity_id,
source="graph_construction",
entity_type="graph_entity",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=activity_started_at_time,
activity_ended_at_time=activity_ended_at_time,
metadata={
"operation": "build_entity",
"entity_type": entity.get('type'),
@@ -197,6 +235,11 @@ class GraphBuilderWithProvenance:
entity_id=rel_id,
source="graph_construction",
entity_type="graph_relationship",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=activity_started_at_time,
activity_ended_at_time=activity_ended_at_time,
metadata={
"operation": "build_relationship",
"relationship_type": relationship.get('type'),
@@ -247,9 +290,17 @@ class AlgorithmTrackerWithProvenance:
)
"""
def __init__(self, provenance: bool = False, **config):
def __init__(
self,
provenance: bool = False,
agent_id: Optional[str] = None,
is_automated: bool = True,
**config,
):
self.provenance = provenance
self._prov_manager = None
self._agent_id = agent_id or self.__class__.__name__
self._is_automated = is_automated
if provenance:
try:
@@ -264,7 +315,8 @@ class AlgorithmTrackerWithProvenance:
algorithm: str,
embeddings: Dict[str, List[float]],
parameters: Dict[str, Any],
source: str = None
source: str = None,
**kwargs
):
"""
Track node embedding algorithm computation with provenance.
@@ -286,6 +338,11 @@ class AlgorithmTrackerWithProvenance:
self._prov_manager.track_entity(
entity_id=execution_id,
source=source or "algorithm_execution",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=kwargs.get("activity_started_at_time"),
activity_ended_at_time=kwargs.get("activity_ended_at_time"),
metadata={
"entity_type": "embedding_computation",
"algorithm": algorithm,
@@ -303,6 +360,11 @@ class AlgorithmTrackerWithProvenance:
self._prov_manager.track_entity(
entity_id=f"embedding_{node_id}",
source=source or "algorithm_execution",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=kwargs.get("activity_started_at_time"),
activity_ended_at_time=kwargs.get("activity_ended_at_time"),
metadata={
"entity_type": "node_embedding",
"algorithm": algorithm,
@@ -345,6 +407,11 @@ class AlgorithmTrackerWithProvenance:
self._prov_manager.track_entity(
entity_id=execution_id,
source=source or "algorithm_execution",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=kwargs.get("activity_started_at_time"),
activity_ended_at_time=kwargs.get("activity_ended_at_time"),
metadata={
"entity_type": "similarity_calculation",
"algorithm": f"similarity_{method}",
@@ -363,6 +430,11 @@ class AlgorithmTrackerWithProvenance:
self._prov_manager.track_entity(
entity_id=f"similarity_{node_id}_{execution_id}",
source=source or "algorithm_execution",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=kwargs.get("activity_started_at_time"),
activity_ended_at_time=kwargs.get("activity_ended_at_time"),
metadata={
"entity_type": "similarity_result",
"method": method,
@@ -393,6 +465,11 @@ class AlgorithmTrackerWithProvenance:
self._prov_manager.track_entity(
entity_id=execution_id,
source=source or "algorithm_execution",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=kwargs.get("activity_started_at_time"),
activity_ended_at_time=kwargs.get("activity_ended_at_time"),
metadata={
"entity_type": "link_prediction",
"algorithm": f"link_prediction_{method}",
@@ -410,6 +487,11 @@ class AlgorithmTrackerWithProvenance:
self._prov_manager.track_entity(
entity_id=f"prediction_{execution_id}_{i}",
source=source or "algorithm_execution",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=kwargs.get("activity_started_at_time"),
activity_ended_at_time=kwargs.get("activity_ended_at_time"),
metadata={
"entity_type": "link_prediction_result",
"method": method,
@@ -453,6 +535,11 @@ class AlgorithmTrackerWithProvenance:
self._prov_manager.track_entity(
entity_id=execution_id,
source=source or "algorithm_execution",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=kwargs.get("activity_started_at_time"),
activity_ended_at_time=kwargs.get("activity_ended_at_time"),
metadata={
"entity_type": "centrality_calculation",
"algorithm": f"centrality_{method}",
@@ -470,6 +557,11 @@ class AlgorithmTrackerWithProvenance:
self._prov_manager.track_entity(
entity_id=f"centrality_{node_id}_{execution_id}",
source=source or "algorithm_execution",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=kwargs.get("activity_started_at_time"),
activity_ended_at_time=kwargs.get("activity_ended_at_time"),
metadata={
"entity_type": "centrality_score",
"method": method,
@@ -500,6 +592,11 @@ class AlgorithmTrackerWithProvenance:
self._prov_manager.track_entity(
entity_id=execution_id,
source=source or "algorithm_execution",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=kwargs.get("activity_started_at_time"),
activity_ended_at_time=kwargs.get("activity_ended_at_time"),
metadata={
"entity_type": "community_detection",
"algorithm": f"community_detection_{method}",
@@ -517,6 +614,11 @@ class AlgorithmTrackerWithProvenance:
self._prov_manager.track_entity(
entity_id=f"community_{execution_id}_{i}",
source=source or "algorithm_execution",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=kwargs.get("activity_started_at_time"),
activity_ended_at_time=kwargs.get("activity_ended_at_time"),
metadata={
"entity_type": "community",
"method": method,
@@ -547,6 +649,11 @@ class AlgorithmTrackerWithProvenance:
self._prov_manager.track_entity(
entity_id=execution_id,
source=source or "graph_construction",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=kwargs.get("activity_started_at_time"),
activity_ended_at_time=kwargs.get("activity_ended_at_time"),
metadata={
"entity_type": "graph_construction",
"entities_count": entities_count,
@@ -573,6 +680,11 @@ class AlgorithmTrackerWithProvenance:
self._prov_manager.track_entity(
entity_id=result_id,
source=source or "similarity_result",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=kwargs.get("activity_started_at_time"),
activity_ended_at_time=kwargs.get("activity_ended_at_time"),
metadata={
"entity_type": "similarity_result",
"node_id": node_id,
@@ -599,6 +711,11 @@ class AlgorithmTrackerWithProvenance:
self._prov_manager.track_entity(
entity_id=result_id,
source=source or "similarity_threshold",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=kwargs.get("activity_started_at_time"),
activity_ended_at_time=kwargs.get("activity_ended_at_time"),
metadata={
"entity_type": "similarity_threshold_analysis",
"execution_id": execution_id,
@@ -624,6 +741,11 @@ class AlgorithmTrackerWithProvenance:
self._prov_manager.track_entity(
entity_id=result_id,
source=source or "entity_processing",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=kwargs.get("activity_started_at_time"),
activity_ended_at_time=kwargs.get("activity_ended_at_time"),
metadata={
"entity_type": "entity_processing",
"processed_entity_id": entity_id,
@@ -648,6 +770,11 @@ class AlgorithmTrackerWithProvenance:
self._prov_manager.track_entity(
entity_id=result_id,
source=source or "relationship_processing",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=kwargs.get("activity_started_at_time"),
activity_ended_at_time=kwargs.get("activity_ended_at_time"),
metadata={
"entity_type": "relationship_processing",
"processed_relationship_id": relationship_id,
@@ -672,6 +799,11 @@ class AlgorithmTrackerWithProvenance:
self._prov_manager.track_entity(
entity_id=result_id,
source=source or "path_analysis",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=kwargs.get("activity_started_at_time"),
activity_ended_at_time=kwargs.get("activity_ended_at_time"),
metadata={
"entity_type": "path_analysis",
"paths_count": len(paths) if paths else 0,
@@ -700,6 +832,11 @@ class AlgorithmTrackerWithProvenance:
self._prov_manager.track_entity(
entity_id=result_id,
source=source or "path_finding",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=kwargs.get("activity_started_at_time"),
activity_ended_at_time=kwargs.get("activity_ended_at_time"),
metadata={
"entity_type": "path_finding",
"source_node": source_node,
@@ -724,6 +861,11 @@ class AlgorithmTrackerWithProvenance:
self._prov_manager.track_entity(
entity_id=result_id,
source=source or "embedding_analysis",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=kwargs.get("activity_started_at_time"),
activity_ended_at_time=kwargs.get("activity_ended_at_time"),
metadata={
"entity_type": "embedding_analysis",
"embeddings_count": len(embeddings),
@@ -746,6 +888,11 @@ class AlgorithmTrackerWithProvenance:
self._prov_manager.track_entity(
entity_id=result_id,
source=source or "connectivity_analysis",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=kwargs.get("activity_started_at_time"),
activity_ended_at_time=kwargs.get("activity_ended_at_time"),
metadata={
"entity_type": "connectivity_analysis",
"components_count": len(components),
@@ -768,6 +915,11 @@ class AlgorithmTrackerWithProvenance:
self._prov_manager.track_entity(
entity_id=result_id,
source=source or "cross_layer_analysis",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=kwargs.get("activity_started_at_time"),
activity_ended_at_time=kwargs.get("activity_ended_at_time"),
metadata={
"entity_type": "cross_layer_analysis",
"layers_count": len(cross_layer_results) if cross_layer_results else 0,
@@ -793,6 +945,11 @@ class AlgorithmTrackerWithProvenance:
self._prov_manager.track_entity(
entity_id=result_id,
source=source or "pipeline_summary",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=kwargs.get("activity_started_at_time"),
activity_ended_at_time=kwargs.get("activity_ended_at_time"),
metadata={
"entity_type": "pipeline_summary",
"pipeline_id": pipeline_id,
@@ -820,6 +977,11 @@ class AlgorithmTrackerWithProvenance:
self._prov_manager.track_entity(
entity_id=summary_id,
source=source or "workflow_summary",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=kwargs.get("activity_started_at_time"),
activity_ended_at_time=kwargs.get("activity_ended_at_time"),
metadata={
"entity_type": "workflow_summary",
"master_workflow_id": master_workflow_id,
@@ -848,6 +1010,11 @@ class AlgorithmTrackerWithProvenance:
self._prov_manager.track_entity(
entity_id=result_id,
source=source or "link_prediction_result",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=kwargs.get("activity_started_at_time"),
activity_ended_at_time=kwargs.get("activity_ended_at_time"),
metadata={
"entity_type": "link_prediction_result",
"source_node": source_node,
@@ -869,6 +1036,11 @@ class AlgorithmTrackerWithProvenance:
self._prov_manager.track_entity(
entity_id=result_id,
source=source or analysis_type,
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=kwargs.get("activity_started_at_time"),
activity_ended_at_time=kwargs.get("activity_ended_at_time"),
metadata={"entity_type": analysis_type, "timestamp": time.time(), **{k: str(v)[:100] for k, v in kwargs.items() if not callable(v)}},
)
return result_id
semantica/llms/llms_provenance.py
+80 -9
View File
@@ -42,6 +42,7 @@ License: MIT
"""
from typing import Optional, Dict, Any
from datetime import datetime
import time
import uuid
@@ -54,16 +55,27 @@ class LLMProvenanceMixin:
LLM API calls including tokens, costs, and performance metrics.
"""
def __init__(self, provenance: bool = False, **kwargs):
def __init__(
self,
provenance: bool = False,
agent_id: Optional[str] = None,
is_automated: bool = True,
**kwargs,
):
"""
Initialize LLM provenance tracking.
Args:
provenance: Enable provenance tracking (default: False)
agent_id: Agent identifier for accountability (issue #825); defaults
to the wrapping class name
is_automated: Whether this agent acted without direct human review
**kwargs: Additional arguments passed to parent class
"""
self.provenance = provenance
self._prov_manager = None
self._agent_id = agent_id or self.__class__.__name__
self._is_automated = is_automated
if provenance:
try:
@@ -99,10 +111,21 @@ class LLMProvenanceMixin:
elif not isinstance(response, str):
response_text = str(response)
# Typed Activity timing (issue #825, Part B Tier 1): popped out so
# it populates real fields, not the opaque metadata blob.
activity_started_at_time = metadata.pop("activity_started_at_time", None)
activity_ended_at_time = metadata.pop("activity_ended_at_time", None)
self._prov_manager.track_entity(
entity_id=call_id,
source=f"{self.__class__.__name__}_api",
entity_type="llm_generation",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_id=call_id,
activity_started_at_time=activity_started_at_time,
activity_ended_at_time=activity_ended_at_time,
metadata={
"model": getattr(self, 'model', 'unknown'),
"prompt_preview": prompt[:200] if len(prompt) > 200 else prompt,
@@ -124,7 +147,13 @@ class GroqLLMWithProvenance(LLMProvenanceMixin):
>>> # API call is tracked with model, tokens, cost, latency
"""
def __init__(self, provenance: bool = False, **config):
def __init__(
self,
provenance: bool = False,
agent_id: Optional[str] = None,
is_automated: bool = True,
**config,
):
"""
Initialize Groq LLM with optional provenance.
@@ -134,7 +163,9 @@ class GroqLLMWithProvenance(LLMProvenanceMixin):
"""
from .groq_llm import GroqLLM
LLMProvenanceMixin.__init__(self, provenance=provenance)
LLMProvenanceMixin.__init__(
self, provenance=provenance, agent_id=agent_id, is_automated=is_automated
)
self._llm = GroqLLM(**config)
self.model = getattr(self._llm, 'model', 'groq')
@@ -150,8 +181,10 @@ class GroqLLMWithProvenance(LLMProvenanceMixin):
LLM response (same format as original GroqLLM)
"""
start_time = time.time()
activity_started_at_time = datetime.utcnow().isoformat()
response = self._llm.generate(prompt, **kwargs)
elapsed = time.time() - start_time
activity_ended_at_time = datetime.utcnow().isoformat()
if self.provenance:
# Extract token counts if available
@@ -175,6 +208,8 @@ class GroqLLMWithProvenance(LLMProvenanceMixin):
total_tokens=(prompt_tokens + completion_tokens) if (prompt_tokens and completion_tokens) else None,
total_cost=total_cost,
latency_seconds=elapsed,
activity_started_at_time=activity_started_at_time,
activity_ended_at_time=activity_ended_at_time,
temperature=kwargs.get('temperature'),
max_tokens=kwargs.get('max_tokens'),
top_p=kwargs.get('top_p')
@@ -194,7 +229,13 @@ class OpenAILLMWithProvenance(LLMProvenanceMixin):
Wraps the original OpenAILLM and tracks all API calls.
"""
def __init__(self, provenance: bool = False, **config):
def __init__(
self,
provenance: bool = False,
agent_id: Optional[str] = None,
is_automated: bool = True,
**config,
):
"""
Initialize OpenAI LLM with optional provenance.
@@ -204,7 +245,9 @@ class OpenAILLMWithProvenance(LLMProvenanceMixin):
"""
from .openai_llm import OpenAILLM
LLMProvenanceMixin.__init__(self, provenance=provenance)
LLMProvenanceMixin.__init__(
self, provenance=provenance, agent_id=agent_id, is_automated=is_automated
)
self._llm = OpenAILLM(**config)
self.model = getattr(self._llm, 'model', 'openai')
@@ -220,8 +263,10 @@ class OpenAILLMWithProvenance(LLMProvenanceMixin):
LLM response
"""
start_time = time.time()
activity_started_at_time = datetime.utcnow().isoformat()
response = self._llm.generate(prompt, **kwargs)
elapsed = time.time() - start_time
activity_ended_at_time = datetime.utcnow().isoformat()
if self.provenance:
# Extract token counts if available
@@ -245,6 +290,8 @@ class OpenAILLMWithProvenance(LLMProvenanceMixin):
total_tokens=(prompt_tokens + completion_tokens) if (prompt_tokens and completion_tokens) else None,
total_cost=total_cost,
latency_seconds=elapsed,
activity_started_at_time=activity_started_at_time,
activity_ended_at_time=activity_ended_at_time,
temperature=kwargs.get('temperature'),
max_tokens=kwargs.get('max_tokens')
)
@@ -263,7 +310,13 @@ class HuggingFaceLLMWithProvenance(LLMProvenanceMixin):
Wraps the original HuggingFaceLLM and tracks all generations.
"""
def __init__(self, provenance: bool = False, **config):
def __init__(
self,
provenance: bool = False,
agent_id: Optional[str] = None,
is_automated: bool = True,
**config,
):
"""
Initialize HuggingFace LLM with optional provenance.
@@ -273,7 +326,9 @@ class HuggingFaceLLMWithProvenance(LLMProvenanceMixin):
"""
from .huggingface_llm import HuggingFaceLLM
LLMProvenanceMixin.__init__(self, provenance=provenance)
LLMProvenanceMixin.__init__(
self, provenance=provenance, agent_id=agent_id, is_automated=is_automated
)
self._llm = HuggingFaceLLM(**config)
self.model = getattr(self._llm, 'model', 'huggingface')
@@ -289,8 +344,10 @@ class HuggingFaceLLMWithProvenance(LLMProvenanceMixin):
LLM response
"""
start_time = time.time()
activity_started_at_time = datetime.utcnow().isoformat()
response = self._llm.generate(prompt, **kwargs)
elapsed = time.time() - start_time
activity_ended_at_time = datetime.utcnow().isoformat()
if self.provenance:
self._track_llm_call(
@@ -298,6 +355,8 @@ class HuggingFaceLLMWithProvenance(LLMProvenanceMixin):
prompt=prompt,
response=response,
latency_seconds=elapsed,
activity_started_at_time=activity_started_at_time,
activity_ended_at_time=activity_ended_at_time,
max_length=kwargs.get('max_length'),
temperature=kwargs.get('temperature')
)
@@ -316,7 +375,13 @@ class LiteLLMWithProvenance(LLMProvenanceMixin):
Wraps the original LiteLLM and tracks all API calls across providers.
"""
def __init__(self, provenance: bool = False, **config):
def __init__(
self,
provenance: bool = False,
agent_id: Optional[str] = None,
is_automated: bool = True,
**config,
):
"""
Initialize LiteLLM with optional provenance.
@@ -326,7 +391,9 @@ class LiteLLMWithProvenance(LLMProvenanceMixin):
"""
from .lite_llm import LiteLLM
LLMProvenanceMixin.__init__(self, provenance=provenance)
LLMProvenanceMixin.__init__(
self, provenance=provenance, agent_id=agent_id, is_automated=is_automated
)
self._llm = LiteLLM(**config)
self.model = getattr(self._llm, 'model', 'litellm')
@@ -342,8 +409,10 @@ class LiteLLMWithProvenance(LLMProvenanceMixin):
LLM response
"""
start_time = time.time()
activity_started_at_time = datetime.utcnow().isoformat()
response = self._llm.generate(prompt, **kwargs)
elapsed = time.time() - start_time
activity_ended_at_time = datetime.utcnow().isoformat()
if self.provenance:
# LiteLLM provides unified response format
@@ -366,6 +435,8 @@ class LiteLLMWithProvenance(LLMProvenanceMixin):
completion_tokens=completion_tokens,
total_cost=total_cost,
latency_seconds=elapsed,
activity_started_at_time=activity_started_at_time,
activity_ended_at_time=activity_ended_at_time,
provider=kwargs.get('provider')
)
semantica/normalize/normalize_provenance.py
+18 -2
View File
@@ -11,19 +11,28 @@ Author: Semantica Contributors
License: MIT
"""
from typing import Any
from typing import Any, Optional
from datetime import datetime
import uuid
class NormalizerWithProvenance:
"""Normalizer with provenance tracking."""
def __init__(self, provenance: bool = False, **config):
def __init__(
self,
provenance: bool = False,
agent_id: Optional[str] = None,
is_automated: bool = True,
**config,
):
from .normalizer import Normalizer
self.provenance = provenance
self._normalizer = Normalizer(**config)
self._prov_manager = None
self._agent_id = agent_id or self.__class__.__name__
self._is_automated = is_automated
if provenance:
try:
@@ -34,13 +43,20 @@ class NormalizerWithProvenance:
def normalize(self, data: Any, source: str = None, **kwargs):
"""Normalize data with provenance tracking."""
activity_started_at_time = datetime.utcnow().isoformat()
result = self._normalizer.normalize(data, **kwargs)
activity_ended_at_time = datetime.utcnow().isoformat()
if self.provenance and self._prov_manager:
self._prov_manager.track_entity(
entity_id=f"normalize_{uuid.uuid4().hex[:8]}",
source=source or "normalization",
entity_type="normalized_data",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=activity_started_at_time,
activity_ended_at_time=activity_ended_at_time,
metadata={"method": kwargs.get('method', 'default')}
)
semantica/ontology/ontology_provenance.py
+18 -2
View File
@@ -11,19 +11,28 @@ Author: Semantica Contributors
License: MIT
"""
from typing import Any
from typing import Any, Optional
from datetime import datetime
import uuid
class OntologyManagerWithProvenance:
"""Ontology manager with provenance tracking."""
def __init__(self, provenance: bool = False, **config):
def __init__(
self,
provenance: bool = False,
agent_id: Optional[str] = None,
is_automated: bool = True,
**config,
):
from .ontology_manager import OntologyManager
self.provenance = provenance
self._manager = OntologyManager(**config)
self._prov_manager = None
self._agent_id = agent_id or self.__class__.__name__
self._is_automated = is_automated
if provenance:
try:
@@ -34,13 +43,20 @@ class OntologyManagerWithProvenance:
def add_concept(self, concept: Any, source: str = None, **kwargs):
"""Add concept with provenance tracking."""
activity_started_at_time = datetime.utcnow().isoformat()
result = self._manager.add_concept(concept, **kwargs)
activity_ended_at_time = datetime.utcnow().isoformat()
if self.provenance and self._prov_manager:
self._prov_manager.track_entity(
entity_id=f"concept_{uuid.uuid4().hex[:8]}",
source=source or "ontology",
entity_type="ontology_concept",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=activity_started_at_time,
activity_ended_at_time=activity_ended_at_time,
metadata={"concept_name": str(concept)}
)
semantica/parse/parse_provenance.py
+18 -2
View File
@@ -13,19 +13,28 @@ Author: Semantica Contributors
License: MIT
"""
from typing import Any
from typing import Any, Optional
from datetime import datetime
import uuid
class ParserWithProvenance:
"""Base parser with provenance tracking."""
def __init__(self, provenance: bool = False, **config):
def __init__(
self,
provenance: bool = False,
agent_id: Optional[str] = None,
is_automated: bool = True,
**config,
):
from .parser import Parser
self.provenance = provenance
self._parser = Parser(**config)
self._prov_manager = None
self._agent_id = agent_id or self.__class__.__name__
self._is_automated = is_automated
if provenance:
try:
@@ -36,13 +45,20 @@ class ParserWithProvenance:
def parse(self, file_path: str, **kwargs):
"""Parse file with provenance tracking."""
activity_started_at_time = datetime.utcnow().isoformat()
data = self._parser.parse(file_path, **kwargs)
activity_ended_at_time = datetime.utcnow().isoformat()
if self.provenance and self._prov_manager:
self._prov_manager.track_entity(
entity_id=f"parse_{uuid.uuid4().hex[:8]}",
source=file_path,
entity_type="parsed_data",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=activity_started_at_time,
activity_ended_at_time=activity_ended_at_time,
metadata={
"file_path": file_path,
"format": kwargs.get('format', 'unknown')
semantica/pipeline/pipeline_provenance.py
+18 -1
View File
@@ -16,6 +16,7 @@ License: MIT
"""
from typing import Optional, Any, Dict, List
from datetime import datetime
import uuid
import time
@@ -23,13 +24,21 @@ import time
class PipelineWithProvenance:
"""Pipeline executor with complete provenance tracking."""
def __init__(self, provenance: bool = False, **config):
def __init__(
self,
provenance: bool = False,
agent_id: Optional[str] = None,
is_automated: bool = True,
**config,
):
"""Initialize pipeline with optional provenance."""
from .pipeline import Pipeline
self.provenance = provenance
self._pipeline = Pipeline(**config)
self._prov_manager = None
self._agent_id = agent_id or self.__class__.__name__
self._is_automated = is_automated
if provenance:
try:
@@ -42,15 +51,23 @@ class PipelineWithProvenance:
"""Run pipeline with provenance tracking."""
pipeline_id = f"pipeline_{uuid.uuid4().hex[:8]}"
start_time = time.time()
activity_started_at_time = datetime.utcnow().isoformat()
result = self._pipeline.run(data, **kwargs)
elapsed = time.time() - start_time
activity_ended_at_time = datetime.utcnow().isoformat()
if self.provenance and self._prov_manager:
self._prov_manager.track_entity(
entity_id=pipeline_id,
source=source or "pipeline_execution",
entity_type="pipeline_run",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_id=pipeline_id,
activity_started_at_time=activity_started_at_time,
activity_ended_at_time=activity_ended_at_time,
metadata={
"steps": len(self._pipeline.steps) if hasattr(self._pipeline, 'steps') else 0,
"duration_seconds": elapsed,
semantica/provenance/__init__.py
+10 -1
View File
@@ -41,7 +41,13 @@ Author: Semantica Contributors
License: MIT
"""
from .schemas import ProvenanceEntry, SourceReference
from .schemas import (
ProvenanceEntry,
SourceReference,
AgentRecord,
ActivityRecord,
Invalidation,
)
from .storage import ProvenanceStorage, InMemoryStorage, SQLiteStorage
from .manager import ProvenanceManager, default_storage_path
from .integrity import compute_checksum, verify_checksum
@@ -50,6 +56,9 @@ __all__ = [
# Core schemas
"ProvenanceEntry",
"SourceReference",
"AgentRecord",
"ActivityRecord",
"Invalidation",
# Storage backends
"ProvenanceStorage",
semantica/provenance/integrity.py
+48 -3
View File
@@ -31,6 +31,29 @@ def compute_checksum(entry: Any) -> str:
Creates a deterministic checksum based on critical provenance fields
to detect any tampering or corruption of provenance data.
Includes `previous_checksum` (issue #825, Part A item 2), which chains
each entry to the prior entry in insertion order (see
ProvenanceStorage.get_chain_head / ProvenanceManager.verify_chain) —
wholesale deletion of a row breaks the chain for the entry that used to
follow it, making the deletion detectable even though per-row checksums
only prove a surviving row wasn't edited in place. Also includes
agent_id/agent_type and the lineage-link fields (parent_entity_id,
previous_version_id, derived_from_id, used_entities) so tampering with
attribution or lineage is detected, not just tampering with the six
original fields.
Deliberately excludes `entity_id` itself: entity_id is the storage
primary key, and ProvenanceManager.track_entity()'s versioning archives
a prior value by copying it to a new entity_id (e.g. "X" -> "X:v:...").
If entity_id were hashed, that relabeling would change the archived
copy's checksum, permanently orphaning any later entry whose
previous_checksum had already chained from the pre-relabel value —
a false-positive "broken chain" for a legitimate rename, not tampering.
Tampering that swaps a row's entity_id while keeping its content is a
narrower threat than content tampering, already partially caught by
verify_chain() (it requires a delete+insert, which breaks the chain for
whatever the deleted row's successor was).
Args:
entry: ProvenanceEntry or dict to compute checksum for
@@ -48,24 +71,46 @@ def compute_checksum(entry: Any) -> str:
>>> print(checksum)
'a3b2c1d4e5f6...'
"""
# Concatenate critical fields for checksum
# Concatenate critical fields for checksum (entity_id intentionally excluded, see docstring)
if isinstance(entry, dict):
used_entities = entry.get("used_entities") or []
data = (
f"{entry.get('entity_id') or ''}"
f"{entry.get('entity_type') or ''}"
f"{entry.get('activity_id') or ''}"
f"{entry.get('agent_id') or ''}"
f"{entry.get('agent_type') or ''}"
f"{entry.get('source_document') or ''}"
f"{entry.get('timestamp') or ''}"
f"{entry.get('confidence') if entry.get('confidence') is not None else 1.0}"
f"{entry.get('parent_entity_id') or ''}"
f"{entry.get('previous_version_id') or ''}"
f"{entry.get('derived_from_id') or ''}"
f"{','.join(used_entities)}"
f"{entry.get('previous_checksum') or ''}"
f"{bool(entry.get('invalidated'))}"
f"{entry.get('invalidated_at_time') or ''}"
f"{entry.get('invalidated_by') or ''}"
f"{entry.get('invalidation_reason') or ''}"
)
else:
used_entities = getattr(entry, "used_entities", None) or []
data = (
f"{entry.entity_id}"
f"{entry.entity_type}"
f"{entry.activity_id}"
f"{getattr(entry, 'agent_id', '') or ''}"
f"{getattr(entry, 'agent_type', '') or ''}"
f"{entry.source_document}"
f"{entry.timestamp}"
f"{entry.confidence}"
f"{getattr(entry, 'parent_entity_id', '') or ''}"
f"{getattr(entry, 'previous_version_id', '') or ''}"
f"{getattr(entry, 'derived_from_id', '') or ''}"
f"{','.join(used_entities)}"
f"{getattr(entry, 'previous_checksum', '') or ''}"
f"{bool(getattr(entry, 'invalidated', False))}"
f"{getattr(entry, 'invalidated_at_time', '') or ''}"
f"{getattr(entry, 'invalidated_by', '') or ''}"
f"{getattr(entry, 'invalidation_reason', '') or ''}"
)
return hashlib.sha256(data.encode('utf-8')).hexdigest()
semantica/provenance/manager.py
+588 -22
View File
@@ -33,11 +33,18 @@ import inspect
import json
import threading
from .schemas import ProvenanceEntry, SourceReference
from .schemas import ProvenanceEntry, SourceReference, AgentRecord, ActivityRecord
from .storage import ProvenanceStorage, InMemoryStorage, SQLiteStorage
from .integrity import compute_checksum, verify_checksum
from ..utils.logging import get_logger
# Issue #825, Part B Tier 3 — configurable base URI for export_prov(), shared
# with RDFExporter's NamespaceManager "semantica" entry (semantica/export/
# rdf_exporter.py) so KG-exported and PROV-exported URIs for the same
# entity_id co-resolve to the same namespace instead of two different
# placeholder domains.
DEFAULT_BASE_URI = "https://semantica.dev/ns#"
@contextmanager
def default_storage_path(path: Optional[str]):
@@ -146,6 +153,16 @@ class ProvenanceManager:
_raise_on_error: bool = False,
) -> Optional[ProvenanceEntry]:
"""Compute checksum, store entry persistently, and log/handle storage errors (#783)."""
# Hash-chain linkage (issue #825, Part A item 2): link this entry to
# the previous entry in global insertion order before hashing, so
# deleting a row later breaks the chain for whatever followed it.
try:
head = self.storage.get_chain_head(_conn)
except Exception:
head = None
entry.sequence_id = (head[0] + 1) if head else 1
entry.previous_checksum = head[1] if head else None
entry.checksum = compute_checksum(entry)
try:
@@ -178,6 +195,69 @@ class ProvenanceManager:
with self.storage.transaction() as conn:
yield conn
# Recognized typed kwargs for track_entity, used both directly and to
# split track_entities_batch's **metadata into real kwargs vs. free-form
# metadata (issue #825, Part A item 3 — fixes a bug where agent_id/
# entity_type/activity_id passed to track_entities_batch were silently
# absorbed into the opaque metadata blob instead of populating fields).
_TRACK_ENTITY_KWARGS = frozenset({
"entity_type", "activity_id", "agent_id", "agent_type", "is_automated",
"role", "agent", "source_location", "source_quote", "confidence",
"parent_entity_id", "used_entities",
# Part B Tier 1/2/3 (issue #825)
"activity", "activity_started_at_time", "activity_ended_at_time",
"acted_on_behalf_of", "informed_by", "valid_from", "valid_until",
"revision_type", "supersedes", "bundle_id",
})
@staticmethod
def _resolve_agent_kwargs(kwargs: Dict[str, Any]) -> Dict[str, Any]:
"""
Resolve agent_id/agent_type/is_automated/role from kwargs.
Accepts either an `agent=AgentRecord(...)` kwarg (takes precedence)
or individual `agent_id`/`agent_type`/`is_automated` scalar kwargs.
Fixes issue #825's dead agent_id field: previously no track_*
method read agent_id from kwargs at all, so it was always the
dataclass default "semantica" regardless of what callers passed.
"""
agent = kwargs.get("agent")
if isinstance(agent, AgentRecord):
return {
"agent_id": agent.id,
"agent_type": agent.agent_type,
"is_automated": agent.is_automated,
"role": kwargs.get("role"),
}
return {
"agent_id": kwargs.get("agent_id", "semantica"),
"agent_type": kwargs.get("agent_type", "software_agent"),
"is_automated": kwargs.get("is_automated", True),
"role": kwargs.get("role"),
}
@staticmethod
def _resolve_activity_kwargs(kwargs: Dict[str, Any], default_activity_id: str) -> Dict[str, Any]:
"""
Resolve activity_id/activity_started_at_time/activity_ended_at_time
from kwargs (issue #825, Part B Tier 1 — typed Activity).
Accepts either an `activity=ActivityRecord(...)` kwarg (takes
precedence) or individual scalar kwargs.
"""
activity = kwargs.get("activity")
if isinstance(activity, ActivityRecord):
return {
"activity_id": activity.id,
"activity_started_at_time": activity.started_at_time,
"activity_ended_at_time": activity.ended_at_time,
}
return {
"activity_id": kwargs.get("activity_id", default_activity_id),
"activity_started_at_time": kwargs.get("activity_started_at_time"),
"activity_ended_at_time": kwargs.get("activity_ended_at_time"),
}
# === Entity Tracking (from kg.ProvenanceTracker) ===
def track_entity(
@@ -253,15 +333,31 @@ class ProvenanceManager:
history_entry.entity_id = history_id
# Pure relabel: checksum/sequence_id/previous_checksum are
# left exactly as they were. compute_checksum() excludes
# entity_id specifically so this is safe — recomputing it
# here (or assigning a fresh sequence slot) would either
# invalidate any later entry that already chained from
# this row's checksum, or strand the original sequence
# position and make verify_chain() see a phantom gap
# (issue #825: an archival relabel is not tampering and
# must not look like it to the hash chain).
self.storage._store_with_conn(conn, history_entry)
archived_history_id = history_id
if not explicit_parent_supplied:
parent_id = history_id
agent_info = self._resolve_agent_kwargs(kwargs)
activity_info = self._resolve_activity_kwargs(kwargs, "entity_tracking")
entry = ProvenanceEntry(
entity_id=entity_id,
entity_type=kwargs.get("entity_type", "entity"),
activity_id=kwargs.get("activity_id", "entity_tracking"),
activity_id=activity_info["activity_id"],
agent_id=agent_info["agent_id"],
agent_type=agent_info["agent_type"],
is_automated=agent_info["is_automated"],
role=agent_info["role"],
source_document=source,
source_location=kwargs.get("source_location"),
source_quote=kwargs.get("source_quote"),
@@ -271,8 +367,29 @@ class ProvenanceManager:
last_updated=datetime.utcnow().isoformat(),
parent_entity_id=parent_id,
used_entities=list(kwargs.get("used_entities", [])),
activity_started_at_time=activity_info["activity_started_at_time"],
activity_ended_at_time=activity_info["activity_ended_at_time"],
acted_on_behalf_of=kwargs.get("acted_on_behalf_of"),
informed_by_activities=list(kwargs.get("informed_by", [])),
valid_from=kwargs.get("valid_from"),
valid_until=kwargs.get("valid_until"),
revision_type=kwargs.get("revision_type"),
supersedes=kwargs.get("supersedes"),
bundle_id=kwargs.get("bundle_id"),
)
# Versioning vs. derivation (issue #825, Part A item 4):
# previous_version_id always captures "this corrects a prior
# version of the same fact" when one was archived, independent
# of whether an explicit cross-source parent was also given.
# derived_from_id captures "this fact was derived from a
# different source entity" — only set when a parent was
# explicitly resolved (kwarg/metadata['derived_from']/source
# heuristic), never from the automatic archival link.
entry.previous_version_id = archived_history_id
if explicit_parent_supplied:
entry.derived_from_id = parent_id
if archived_history_id and explicit_parent_supplied:
entry.used_entities.append(archived_history_id)
@@ -323,16 +440,32 @@ class ProvenanceManager:
... metadata={"type": "founded"}
... )
"""
agent_info = self._resolve_agent_kwargs(kwargs)
activity_info = self._resolve_activity_kwargs(kwargs, "relationship_tracking")
entry = ProvenanceEntry(
entity_id=relationship_id,
entity_type="relationship",
activity_id=kwargs.get("activity_id", "relationship_tracking"),
activity_id=activity_info["activity_id"],
agent_id=agent_info["agent_id"],
agent_type=agent_info["agent_type"],
is_automated=agent_info["is_automated"],
role=agent_info["role"],
source_document=source,
source_location=kwargs.get("source_location"),
confidence=kwargs.get("confidence", 1.0),
metadata=metadata or {},
first_seen=datetime.utcnow().isoformat(),
last_updated=datetime.utcnow().isoformat()
last_updated=datetime.utcnow().isoformat(),
activity_started_at_time=activity_info["activity_started_at_time"],
activity_ended_at_time=activity_info["activity_ended_at_time"],
acted_on_behalf_of=kwargs.get("acted_on_behalf_of"),
informed_by_activities=list(kwargs.get("informed_by", [])),
valid_from=kwargs.get("valid_from"),
valid_until=kwargs.get("valid_until"),
revision_type=kwargs.get("revision_type"),
supersedes=kwargs.get("supersedes"),
bundle_id=kwargs.get("bundle_id"),
)
return self._save_entry(entry)
@@ -374,17 +507,36 @@ class ProvenanceManager:
... end_index=500
... )
"""
agent_info = self._resolve_agent_kwargs(metadata)
activity_info = self._resolve_activity_kwargs(metadata, "chunking")
for key in (
"agent_id", "agent_type", "is_automated", "role", "agent",
"activity", "activity_started_at_time", "activity_ended_at_time",
):
metadata.pop(key, None)
entry = ProvenanceEntry(
entity_id=chunk_id,
entity_type="chunk",
activity_id="chunking",
activity_id=activity_info["activity_id"],
agent_id=agent_info["agent_id"],
agent_type=agent_info["agent_type"],
is_automated=agent_info["is_automated"],
role=agent_info["role"],
source_document=source_document,
source_location=source_path,
start_index=start_index,
end_index=end_index,
parent_entity_id=parent_chunk_id,
# A chunk split from a parent chunk is a derivation (a new entity
# produced from an existing one), not a correction of the same
# fact — see track_entity's previous_version_id/derived_from_id
# split (issue #825, Part A item 4).
derived_from_id=parent_chunk_id,
metadata=metadata,
timestamp=datetime.utcnow().isoformat()
timestamp=datetime.utcnow().isoformat(),
activity_started_at_time=activity_info["activity_started_at_time"],
activity_ended_at_time=activity_info["activity_ended_at_time"],
)
return self._save_entry(entry, _conn=_conn, _raise_on_error=(_conn is not None))
@@ -425,10 +577,22 @@ class ProvenanceManager:
... source=source
... )
"""
agent_info = self._resolve_agent_kwargs(metadata)
activity_info = self._resolve_activity_kwargs(metadata, "property_tracking")
for key in (
"agent_id", "agent_type", "is_automated", "role", "agent",
"activity", "activity_started_at_time", "activity_ended_at_time",
):
metadata.pop(key, None)
entry = ProvenanceEntry(
entity_id=f"{entity_id}_{property_name}",
entity_type="property",
activity_id="property_tracking",
activity_id=activity_info["activity_id"],
agent_id=agent_info["agent_id"],
agent_type=agent_info["agent_type"],
is_automated=agent_info["is_automated"],
role=agent_info["role"],
source_document=source.document,
source_location=f"page_{source.page}" if source.page else source.section,
confidence=source.confidence,
@@ -440,7 +604,9 @@ class ProvenanceManager:
**metadata,
**source.metadata
},
timestamp=datetime.utcnow().isoformat()
timestamp=datetime.utcnow().isoformat(),
activity_started_at_time=activity_info["activity_started_at_time"],
activity_ended_at_time=activity_info["activity_ended_at_time"],
)
return self._save_entry(entry)
@@ -474,6 +640,18 @@ class ProvenanceManager:
tracked_count = 0
batch_size = 1000 # Justification (#807): 1,000 items per transaction bounds SQLite WAL frame growth and reduces lock contention during multi-thousand-row imports while achieving a 1000x reduction in connection/commit overhead.
# Split batch-level **metadata into recognized typed track_entity
# kwargs (entity_type, activity_id, agent_id, ...) vs. free-form data.
# Previously ALL of **metadata was merged into the metadata dict and
# passed as track_entity's positional `metadata` arg, so typed kwargs
# like agent_id/entity_type/activity_id were silently absorbed into
# the opaque metadata JSON blob instead of populating real fields
# (issue #825, Part A item 3 — this is what made agent_id a "dead"
# field for every batch caller, including the documented example in
# docs/guides/provenance.md).
batch_kwargs = {k: v for k, v in metadata.items() if k in self._TRACK_ENTITY_KWARGS}
free_metadata = {k: v for k, v in metadata.items() if k not in self._TRACK_ENTITY_KWARGS}
for i in range(0, len(entities), batch_size):
batch = entities[i : i + batch_size]
batch_count = 0
@@ -484,11 +662,13 @@ class ProvenanceManager:
if not entity_id:
continue
entity_metadata = {**metadata, **entity.get("metadata", {})}
entity_metadata = {**free_metadata, **entity.get("metadata", {})}
try:
with self.storage.savepoint(conn):
self.track_entity(entity_id, source, entity_metadata, _conn=conn)
self.track_entity(
entity_id, source, entity_metadata, _conn=conn, **batch_kwargs
)
batch_count += 1
except Exception:
pass # Continue with other entities in this batch
@@ -655,6 +835,137 @@ class ProvenanceManager:
return self.storage.trace_lineage(entity_id, max_depth=max_depth)
return self.storage.trace_lineage(entity_id)
def trace_descendants(
self, entity_id: str, max_depth: Optional[int] = None
) -> List[ProvenanceEntry]:
"""
Trace downstream descendants (reverse lineage) and return raw entries.
This is the counterpart to trace_lineage()/get_lineage(), which only
ever trace upstream ancestors via parent_entity_id/used_entities.
Downstream traceability answers the incident-response question "entity
X was wrong — what downstream facts used it?" (issue #825, Part A
item 5).
Args:
entity_id: Entity identifier
max_depth: Optional maximum BFS depth
Returns:
List of ProvenanceEntry objects that (transitively) reference
entity_id, in BFS order.
"""
return self.storage.trace_descendants(entity_id, max_depth=max_depth)
def get_descendants(self, entity_id: str) -> Dict[str, Any]:
"""
Get downstream descendants for an entity, mirroring get_lineage()'s
return shape but for the reverse direction.
Args:
entity_id: Entity identifier
Returns:
Dictionary containing descendant entries, or {} if none found.
"""
descendant_entries = self.trace_descendants(entity_id)
if not descendant_entries:
return {}
integrity_verified = all(verify_checksum(entry) for entry in descendant_entries)
chain_dicts = [entry.to_dict() for entry in descendant_entries]
return {
"entity_id": entity_id,
"descendant_chain": chain_dicts,
"entries": chain_dicts,
"entity_count": len(descendant_entries),
"integrity_verified": integrity_verified,
}
def revision_history(self, entity_id: str) -> List[Dict[str, Any]]:
"""
Return the version history for entity_id in ascending order.
Issue #825, Part B Tier 3 — closes the "no direct equivalent yet"
gap for kg.ProvenanceTracker.revision_history() documented in
docs/migration/kg-provenance-tracker.md.
Walks the entity's own previous_version_id chain (not the full
upstream lineage via get_lineage(), which also pulls in unrelated
derived_from_id/used_entities links from other source entities).
Args:
entity_id: Entity identifier
Returns:
List of {version, valid_from, valid_until, recorded_at, author,
revision_type, supersedes} dicts, oldest first. valid_from/
valid_until use the entry's own explicit fields when set (the
fact's asserted validity window); otherwise valid_from defaults
to when this version was recorded, and valid_until to the next
version's timestamp (None for the current/most recent version).
Empty list if entity_id was never tracked.
"""
current = self.storage.retrieve(entity_id)
if current is None:
return []
chain: List[ProvenanceEntry] = [current]
visited = {entity_id}
cursor = current
while getattr(cursor, "previous_version_id", None):
prev_id = cursor.previous_version_id
if prev_id in visited:
break
prev_entry = self.storage.retrieve(prev_id)
if prev_entry is None:
break
chain.append(prev_entry)
visited.add(prev_id)
cursor = prev_entry
chain.reverse() # oldest first
history = []
for i, entry in enumerate(chain):
default_valid_until = chain[i + 1].timestamp if i + 1 < len(chain) else None
version_dict: Dict[str, Any] = {
"version": i + 1,
"valid_from": entry.valid_from or entry.timestamp,
"valid_until": entry.valid_until or default_valid_until,
"recorded_at": entry.timestamp,
"author": entry.agent_id,
}
if entry.revision_type:
version_dict["revision_type"] = entry.revision_type
if entry.supersedes:
version_dict["supersedes"] = entry.supersedes
history.append(version_dict)
return history
def query_recorded_between(self, start: str, end: str) -> List[Dict[str, Any]]:
"""
Return all provenance entries whose timestamp falls within [start, end].
Issue #825, Part B Tier 3 — closes the "no direct equivalent yet"
gap for kg.ProvenanceTracker.query_recorded_between() documented in
docs/migration/kg-provenance-tracker.md.
Args:
start: Start of range, ISO 8601 string (inclusive)
end: End of range, ISO 8601 string (inclusive)
Returns:
List of matching entries as dicts, sorted by timestamp ascending.
"""
matches = [
e for e in self.storage.retrieve_all()
if e.timestamp and start <= e.timestamp <= end
]
matches.sort(key=lambda e: e.timestamp)
return [e.to_dict() for e in matches]
def get_all_sources(self, entity_id: str) -> List[Dict[str, Any]]:
"""
Get all sources for an entity (kg.ProvenanceTracker compatible).
@@ -695,12 +1006,92 @@ class ProvenanceManager:
return entry.to_dict()
return None
# === Invalidation (tombstone, not hard delete) ===
def invalidate(
self,
entity_id: str,
agent_id: str,
reason: Optional[str] = None,
metadata: Optional[Dict[str, Any]] = None,
) -> ProvenanceEntry:
"""
Mark a tracked entity as invalidated instead of deleting it
(issue #825, Part A item 1 — prov:Invalidation).
An audit needs "this was deleted/corrected, by whom, when, why" to
itself be provable. Rather than mutating the row in place, this
archives the pre-invalidation state under a stable versioned key
(the same pattern track_entity() uses for corrections) and then
writes the invalidated entry as a fresh, chained append. Mutating
the existing row's checksum in place would silently invalidate any
later entry that had already chained its previous_checksum from
this row's pre-invalidation value — turning a legitimate
invalidation into a false-positive "broken chain" report.
The entry remains visible via retrieve()/retrieve_all()/lineage
traversal, but callers can filter on `invalidated` to exclude
retracted facts.
Args:
entity_id: Entity identifier to invalidate
agent_id: Agent responsible for the invalidation (prov:Agent)
reason: Optional human-readable reason
metadata: Optional metadata to merge into the entry
Returns:
The updated (invalidated) ProvenanceEntry
Raises:
ValueError: If no provenance entry exists for entity_id
Example:
>>> prov_mgr.invalidate("entity_1", agent_id="reviewer_jane",
... reason="Source document retracted")
"""
with self.storage.transaction() as conn:
existing = self.storage._retrieve_with_conn(conn, entity_id)
if existing is None:
raise ValueError(
f"Cannot invalidate: no provenance entry found for entity_id={entity_id!r}"
)
# Archive the pre-invalidation state under a stable key — a pure
# relabel (see track_entity's identical pattern), so its
# checksum/sequence_id/previous_checksum are left untouched.
history_entry = copy.deepcopy(existing)
base_history_id = f"{entity_id}:v:{existing.last_updated}"
history_id = base_history_id
counter = 1
while self.storage._retrieve_with_conn(conn, history_id):
history_id = f"{base_history_id}:{counter}"
counter += 1
history_entry.entity_id = history_id
self.storage._store_with_conn(conn, history_entry)
entry = copy.deepcopy(existing)
entry.invalidated = True
entry.invalidated_at_time = datetime.utcnow().isoformat()
entry.invalidated_by = agent_id
entry.invalidation_reason = reason
entry.previous_version_id = history_id
if metadata:
entry.metadata = {**entry.metadata, **metadata}
self._save_entry(entry, _conn=conn, _raise_on_error=True)
return entry
# === Utility Methods ===
def clear(self) -> int:
"""
Clear all provenance data.
Note: this is a bulk storage reset (used for dev/test teardown), not
a single-fact retraction — use invalidate() to retract/correct an
individual tracked entity while preserving its audit trail.
Returns:
Number of entries cleared
"""
@@ -801,28 +1192,44 @@ class ProvenanceManager:
)
return "\n".join(lines)
def export_prov(self, format: str = "turtle") -> str:
# Agent-type refinement for export_prov (issue #825, Part A item 6 —
# cheap, opportunistic PROV-O typing on top of the generic prov:Agent).
_AGENT_TYPE_PROV_CLASS = {
"person": "Person",
"software_agent": "SoftwareAgent",
"organization": "Organization",
}
def export_prov(self, format: str = "turtle", base_uri: Optional[str] = None) -> str:
"""
Export provenance as W3C PROV-O RDF.
Args:
format: RDF format ('turtle', 'ntriples', 'jsonld')
base_uri: Namespace URI entities/agents/activities are minted
under (issue #825, Part B Tier 3). Defaults to
DEFAULT_BASE_URI, which matches RDFExporter's NamespaceManager
"semantica" entry so KG-exported and PROV-exported URIs for
the same entity_id co-resolve.
Returns:
Serialized RDF string
"""
from rdflib import Graph, Literal, Namespace, URIRef
from rdflib import BNode, Graph, Literal, Namespace, URIRef
from rdflib.namespace import RDF, XSD
PROV = Namespace("http://www.w3.org/ns/prov#")
EX = Namespace("http://example.org/ns/")
EX = Namespace(base_uri or DEFAULT_BASE_URI)
g = Graph()
g.bind("prov", PROV)
g.bind("ex", EX)
def uri(entity_id: Any) -> URIRef:
return URIRef(EX[str(entity_id)])
for e in self.storage.retrieve_all():
ent_uri = URIRef(EX[str(e.entity_id)])
ent_uri = uri(e.entity_id)
g.add((ent_uri, RDF.type, PROV.Entity))
if getattr(e, "timestamp", None):
@@ -834,29 +1241,125 @@ class ProvenanceManager:
)
)
ag_uri = None
if getattr(e, "agent_id", None) and e.agent_id != "unknown":
ag_uri = URIRef(EX[str(e.agent_id)])
ag_uri = uri(e.agent_id)
g.add((ag_uri, RDF.type, PROV.Agent))
prov_subclass = self._AGENT_TYPE_PROV_CLASS.get(
getattr(e, "agent_type", None)
)
if prov_subclass:
g.add((ag_uri, RDF.type, PROV[prov_subclass]))
g.add((ent_uri, PROV.wasAttributedTo, ag_uri))
# Qualified Association with hadRole (issue #825, Part A item
# 6): distinguishes "approved by" from "generated by" from
# "reviewed by" for the same agent/entity pair, which the
# plain wasAttributedTo triple above cannot express.
association = BNode()
g.add((ent_uri, PROV.qualifiedAssociation, association))
g.add((association, RDF.type, PROV.Association))
g.add((association, PROV.agent, ag_uri))
role = getattr(e, "role", None) or "generator"
g.add((association, PROV.hadRole, uri(f"role_{role}")))
# prov:actedOnBehalfOf (issue #825, Part B Tier 2) — agent
# delegation, e.g. an automated agent acting on behalf of the
# human/organization that authorized it.
delegate_id = getattr(e, "acted_on_behalf_of", None)
if delegate_id:
delegate_uri = uri(delegate_id)
g.add((delegate_uri, RDF.type, PROV.Agent))
g.add((ag_uri, PROV.actedOnBehalfOf, delegate_uri))
act_uri = None
if getattr(e, "activity_id", None) and e.activity_id != "unknown":
act_uri = URIRef(EX[str(e.activity_id)])
act_uri = uri(e.activity_id)
g.add((act_uri, RDF.type, PROV.Activity))
g.add((ent_uri, PROV.wasGeneratedBy, act_uri))
# Typed Activity timing (issue #825, Part B Tier 1)
if getattr(e, "activity_started_at_time", None):
g.add((act_uri, PROV.startedAtTime,
Literal(e.activity_started_at_time, datatype=XSD.dateTime)))
if getattr(e, "activity_ended_at_time", None):
g.add((act_uri, PROV.endedAtTime,
Literal(e.activity_ended_at_time, datatype=XSD.dateTime)))
# Qualified Generation (issue #825, Part B Tier 1)
generation = BNode()
g.add((ent_uri, PROV.qualifiedGeneration, generation))
g.add((generation, RDF.type, PROV.Generation))
g.add((generation, PROV.activity, act_uri))
if getattr(e, "timestamp", None):
g.add((generation, PROV.atTime, Literal(e.timestamp, datatype=XSD.dateTime)))
# prov:wasAssociatedWith (issue #825, Part B Tier 2) — direct
# Activity->Agent link, distinct from the Entity->Agent
# wasAttributedTo/qualifiedAssociation triples above.
if ag_uri is not None:
g.add((act_uri, PROV.wasAssociatedWith, ag_uri))
# prov:wasInformedBy (issue #825, Part B Tier 2) — chains this
# activity to prior activities it was informed by (e.g. a
# pipeline stage informed by the stage before it).
for informing_id in getattr(e, "informed_by_activities", []):
g.add((act_uri, PROV.wasInformedBy, uri(informing_id)))
def emit_derivation(source_id: Any) -> None:
"""Emit plain + qualified wasDerivedFrom for a source entity."""
s_uri = uri(source_id)
g.add((ent_uri, PROV.wasDerivedFrom, s_uri))
derivation = BNode()
g.add((ent_uri, PROV.qualifiedDerivation, derivation))
g.add((derivation, RDF.type, PROV.Derivation))
g.add((derivation, PROV.entity, s_uri))
if act_uri is not None:
g.add((derivation, PROV.hadActivity, act_uri))
if getattr(e, "parent_entity_id", None):
p_uri = URIRef(EX[str(e.parent_entity_id)])
g.add((ent_uri, PROV.wasDerivedFrom, p_uri))
emit_derivation(e.parent_entity_id)
for u_id in getattr(e, "used_entities", []):
u_uri = URIRef(EX[str(u_id)])
u_uri = uri(u_id)
# Emit wasDerivedFrom only when this used entity is not the same
# as parent_entity_id — which already carries that triple above.
if u_id != getattr(e, "parent_entity_id", None):
g.add((ent_uri, PROV.wasDerivedFrom, u_uri))
if getattr(e, "activity_id", None) and e.activity_id != "unknown":
act_uri = URIRef(EX[str(e.activity_id)])
emit_derivation(u_id)
if act_uri is not None:
g.add((act_uri, PROV.used, u_uri))
# Qualified Usage (issue #825, Part B Tier 1)
usage = BNode()
g.add((act_uri, PROV.qualifiedUsage, usage))
g.add((usage, RDF.type, PROV.Usage))
g.add((usage, PROV.entity, u_uri))
# Qualified Invalidation (issue #825, Part A item 1): records the
# tombstone as provable RDF rather than a silent hard delete.
if getattr(e, "invalidated", False):
invalidation = BNode()
g.add((ent_uri, PROV.qualifiedInvalidation, invalidation))
g.add((invalidation, RDF.type, PROV.Invalidation))
if getattr(e, "invalidated_at_time", None):
g.add(
(
invalidation,
PROV.invalidatedAtTime,
Literal(e.invalidated_at_time, datatype=XSD.dateTime),
)
)
if getattr(e, "invalidated_by", None):
inv_ag_uri = uri(e.invalidated_by)
g.add((inv_ag_uri, RDF.type, PROV.Agent))
g.add((invalidation, PROV.agent, inv_ag_uri))
# prov:Collection / prov:Bundle membership (issue #825, Part B
# Tier 3) — partitions provenance by source/dataset/ingestion-run.
# Membership triples, not true RDF named-graph partitioning.
if getattr(e, "bundle_id", None):
bundle_uri = uri(f"bundle_{e.bundle_id}")
g.add((bundle_uri, RDF.type, PROV.Bundle))
g.add((bundle_uri, PROV.hadMember, ent_uri))
rdf_format = "json-ld" if format == "jsonld" else format
return g.serialize(format=rdf_format)
@@ -873,6 +1376,7 @@ class ProvenanceManager:
"""
entries = self.storage.retrieve_all()
all_ids = {e.entity_id for e in entries}
all_activity_ids = {e.activity_id for e in entries if getattr(e, "activity_id", None)}
missing_refs = []
for e in entries:
@@ -882,15 +1386,77 @@ class ProvenanceManager:
for u_id in getattr(e, "used_entities", []):
if u_id not in all_ids:
missing_refs.append(f"{e.entity_id} -> {u_id}")
for ref_id in (
getattr(e, "previous_version_id", None),
getattr(e, "derived_from_id", None),
getattr(e, "supersedes", None),
):
if ref_id and ref_id not in all_ids:
missing_refs.append(f"{e.entity_id} -> {ref_id}")
# informed_by_activities references activity_ids, a distinct
# ID space from entity_id (issue #825, Part B Tier 2).
for act_id in getattr(e, "informed_by_activities", []):
if act_id not in all_activity_ids:
missing_refs.append(f"{e.entity_id} (activity) -> {act_id}")
valid = len(missing_refs) == 0
errors = len(missing_refs)
invalidated_count = sum(1 for e in entries if getattr(e, "invalidated", False))
return {
"valid": valid,
"total_entries": len(entries),
"missing_references": missing_refs,
"invalidated_count": invalidated_count,
"strict": strict,
"errors": errors,
}
def verify_chain(self) -> Dict[str, Any]:
"""
Verify the hash chain across all provenance entries (issue #825,
Part A item 2).
Sorts entries by sequence_id (global insertion order) and checks
both that each entry's own checksum matches its content, and that
each entry's previous_checksum matches the checksum of the entry
that precedes it. A gap in the chain — a surviving entry whose
previous_checksum doesn't match its predecessor's checksum — is
exactly what wholesale row deletion produces, since per-row
checksums alone can't detect that a row is simply missing.
Returns:
Dictionary with "valid", "total_entries", and "broken_links"
(list of {"entity_id", "sequence_id", "reason"} dicts).
"""
entries = sorted(
(e for e in self.storage.retrieve_all() if e.sequence_id is not None),
key=lambda e: e.sequence_id,
)
broken_links: List[Dict[str, Any]] = []
expected_previous: Optional[str] = None
for entry in entries:
if not verify_checksum(entry):
broken_links.append({
"entity_id": entry.entity_id,
"sequence_id": entry.sequence_id,
"reason": "checksum_mismatch",
})
continue
if entry.previous_checksum != expected_previous:
broken_links.append({
"entity_id": entry.entity_id,
"sequence_id": entry.sequence_id,
"reason": "chain_break",
"expected_previous_checksum": expected_previous,
"actual_previous_checksum": entry.previous_checksum,
})
expected_previous = entry.checksum
return {
"valid": len(broken_links) == 0,
"total_entries": len(entries),
"broken_links": broken_links,
}
semantica/provenance/schemas.py
+221 -3
View File
@@ -13,10 +13,14 @@ Consolidates:
W3C PROV-O Mapping:
- ProvenanceEntry.entity_id → prov:Entity
- ProvenanceEntry.activity_id → prov:Activity
- ProvenanceEntry.agent_id → prov:Agent
- ProvenanceEntry.parent_entity_id → prov:wasDerivedFrom
- ProvenanceEntry.agent_id / agent_type → prov:Agent (Person|SoftwareAgent|Organization)
- ProvenanceEntry.role → prov:hadRole (via prov:qualifiedAssociation)
- ProvenanceEntry.parent_entity_id → prov:wasDerivedFrom (legacy combined field)
- ProvenanceEntry.derived_from_id → prov:wasDerivedFrom (true cross-source derivation)
- ProvenanceEntry.previous_version_id → prior version of the same fact (correction/versioning)
- ProvenanceEntry.used_entities → prov:used
- ProvenanceEntry.timestamp → prov:generatedAtTime
- ProvenanceEntry.invalidated* → prov:Invalidation (tombstone, not a hard delete)
Author: Semantica Contributors
License: MIT
@@ -76,6 +80,11 @@ class ProvenanceEntry:
activity_id: str
agent_id: str = "semantica"
# Accountability-scoped agent typing (issue #825, Part A item 3)
agent_type: str = "software_agent" # "person" | "software_agent" | "organization"
is_automated: bool = True
role: Optional[str] = None # prov:hadRole, e.g. "generator", "approver", "reviewer"
# Audit-grade source tracking
source_document: str = ""
source_location: Optional[str] = None
@@ -90,10 +99,58 @@ class ProvenanceEntry:
confidence: float = 1.0
checksum: Optional[str] = None
# Chain of custody (W3C PROV-O)
# Hash-chain linkage (issue #825, Part A item 2): previous_checksum links
# this entry to the prior entry in global insertion order (sequence_id),
# so wholesale row deletion breaks the chain and becomes detectable via
# ProvenanceManager.verify_chain().
sequence_id: Optional[int] = None
previous_checksum: Optional[str] = None
# Chain of custody (W3C PROV-O) — kept for backward compatibility.
parent_entity_id: Optional[str] = None
used_entities: List[str] = field(default_factory=list)
# Versioning vs. derivation (issue #825, Part A item 4): previous_version_id
# is "this corrects/replaces a prior version of the same fact"
# (prov:specializationOf-flavored); derived_from_id is "this fact was
# derived from a different source entity" (prov:wasDerivedFrom). Both are
# additive alongside parent_entity_id, which remains the legacy combined
# field for existing readers.
previous_version_id: Optional[str] = None
derived_from_id: Optional[str] = None
# Typed Activity (issue #825, Part B Tier 1): prov:startedAtTime/endedAtTime
# for the activity that generated this entry.
activity_started_at_time: Optional[str] = None
activity_ended_at_time: Optional[str] = None
# prov:actedOnBehalfOf / prov:wasInformedBy (issue #825, Part B Tier 2)
acted_on_behalf_of: Optional[str] = None
informed_by_activities: List[str] = field(default_factory=list)
# Bitemporal fields merged from the deprecated kg.ProvenanceTracker
# (issue #825, Part B Tier 3) — in the deprecated tracker these were
# always caller-supplied optional metadata keys, never auto-computed;
# same contract here. valid_from/valid_until express the fact's own
# asserted validity window (domain/business time), independent of
# sequence_id/timestamp which track when the system recorded it.
valid_from: Optional[str] = None
valid_until: Optional[str] = None
revision_type: Optional[str] = None
supersedes: Optional[str] = None
# prov:Collection / prov:Bundle membership (issue #825, Part B Tier 3):
# partitions provenance by source/dataset/ingestion-run.
bundle_id: Optional[str] = None
# Invalidation / tombstone tracking (issue #825, Part A item 1): set via
# ProvenanceManager.invalidate() instead of deleting the row, so an audit
# can prove a fact existed, was reviewed, and was retracted/corrected.
invalidated: bool = False
invalidated_at_time: Optional[str] = None
invalidated_by: Optional[str] = None
invalidation_reason: Optional[str] = None
# Chunk-specific fields (from split.ProvenanceInfo)
start_index: Optional[int] = None
end_index: Optional[int] = None
@@ -117,6 +174,9 @@ class ProvenanceEntry:
"entity_type": self.entity_type,
"activity_id": self.activity_id,
"agent_id": self.agent_id,
"agent_type": self.agent_type,
"is_automated": self.is_automated,
"role": self.role,
"source_document": self.source_document,
"source_location": self.source_location,
"source_quote": self.source_quote,
@@ -125,8 +185,25 @@ class ProvenanceEntry:
"last_updated": self.last_updated,
"confidence": self.confidence,
"checksum": self.checksum,
"sequence_id": self.sequence_id,
"previous_checksum": self.previous_checksum,
"parent_entity_id": self.parent_entity_id,
"used_entities": self.used_entities,
"previous_version_id": self.previous_version_id,
"derived_from_id": self.derived_from_id,
"activity_started_at_time": self.activity_started_at_time,
"activity_ended_at_time": self.activity_ended_at_time,
"acted_on_behalf_of": self.acted_on_behalf_of,
"informed_by_activities": self.informed_by_activities,
"valid_from": self.valid_from,
"valid_until": self.valid_until,
"revision_type": self.revision_type,
"supersedes": self.supersedes,
"bundle_id": self.bundle_id,
"invalidated": self.invalidated,
"invalidated_at_time": self.invalidated_at_time,
"invalidated_by": self.invalidated_by,
"invalidation_reason": self.invalidation_reason,
"start_index": self.start_index,
"end_index": self.end_index,
"credibility": self.credibility,
@@ -275,3 +352,144 @@ class PropertySource:
for s in data["sources"]
]
return cls(**data)
@dataclass
class AgentRecord:
"""
Minimum-viable typed W3C PROV-O agent (prov:Agent).
Narrower than the full PROV-O Person/SoftwareAgent/Organization taxonomy —
just enough to answer "was a human accountable here" for high-stakes
provenance review. Pass an AgentRecord to ProvenanceManager.track_entity()
(etc.) via the `agent=` kwarg to set agent_id/agent_type/is_automated
together.
Attributes:
id: Agent identifier (prov:Agent)
agent_type: One of "person", "software_agent", "organization"
is_automated: Whether this agent acted without direct human review
name: Optional human-readable name
metadata: Additional metadata dictionary
Example:
>>> agent = AgentRecord(id="reviewer_jane", agent_type="person", is_automated=False)
>>> prov_mgr.track_entity("entity_1", source="doc_1", agent=agent, role="approver")
"""
id: str
agent_type: str = "software_agent"
is_automated: bool = True
name: Optional[str] = None
metadata: Dict[str, Any] = field(default_factory=dict)
def to_dict(self) -> Dict[str, Any]:
"""Convert agent record to dictionary."""
return {
"id": self.id,
"agent_type": self.agent_type,
"is_automated": self.is_automated,
"name": self.name,
"metadata": self.metadata,
}
@classmethod
def from_dict(cls, data: Dict[str, Any]) -> "AgentRecord":
"""Create agent record from dictionary."""
return cls(**data)
@dataclass
class ActivityRecord:
"""
Minimum-viable typed W3C PROV-O activity (prov:Activity).
Issue #825, Part B Tier 1 — promotes activity_id from a bare string to
a typed object carrying start/end timing, mirroring AgentRecord. Pass an
ActivityRecord to ProvenanceManager.track_entity() (etc.) via the
`activity=` kwarg to set activity_id/activity_started_at_time/
activity_ended_at_time together.
Attributes:
id: Activity identifier (prov:Activity)
activity_type: Free-form activity classification (e.g. "process", "extraction")
started_at_time: prov:startedAtTime, ISO datetime string
ended_at_time: prov:endedAtTime, ISO datetime string
metadata: Additional metadata dictionary
Example:
>>> activity = ActivityRecord(
... id="bureau_parsing_run_42",
... started_at_time="2026-01-01T00:00:00",
... ended_at_time="2026-01-01T00:00:03",
... )
>>> prov_mgr.track_entity("entity_1", source="doc_1", activity=activity)
"""
id: str
activity_type: str = "process"
started_at_time: Optional[str] = None
ended_at_time: Optional[str] = None
metadata: Dict[str, Any] = field(default_factory=dict)
def to_dict(self) -> Dict[str, Any]:
"""Convert activity record to dictionary."""
return {
"id": self.id,
"activity_type": self.activity_type,
"started_at_time": self.started_at_time,
"ended_at_time": self.ended_at_time,
"metadata": self.metadata,
}
@classmethod
def from_dict(cls, data: Dict[str, Any]) -> "ActivityRecord":
"""Create activity record from dictionary."""
return cls(**data)
@dataclass
class Invalidation:
"""
W3C PROV-O invalidation record (prov:Invalidation).
Represents the tombstone for a retracted/corrected provenance entry:
"this was deleted/corrected, by whom, when, why" — recorded in place of
a hard delete so the fact that an entity existed remains provable.
Attributes:
entity_id: The entity that was invalidated (prov:Entity)
invalidated_at_time: When invalidation occurred (prov:invalidatedAtTime)
invalidated_by: Agent responsible for the invalidation (prov:Agent)
reason: Optional human-readable reason
metadata: Additional metadata dictionary
Example:
>>> Invalidation(
... entity_id="entity_123",
... invalidated_at_time="2026-01-01T00:00:00",
... invalidated_by="reviewer_jane",
... reason="Source document retracted",
... )
"""
entity_id: str
invalidated_at_time: str
invalidated_by: str
reason: Optional[str] = None
metadata: Dict[str, Any] = field(default_factory=dict)
def to_dict(self) -> Dict[str, Any]:
"""Convert invalidation record to dictionary."""
return {
"entity_id": self.entity_id,
"invalidated_at_time": self.invalidated_at_time,
"invalidated_by": self.invalidated_by,
"reason": self.reason,
"metadata": self.metadata,
}
@classmethod
def from_dict(cls, data: Dict[str, Any]) -> "Invalidation":
"""Create invalidation record from dictionary."""
return cls(**data)
semantica/provenance/storage.py
+304 -5
View File
@@ -20,7 +20,7 @@ License: MIT
"""
from abc import ABC, abstractmethod
from typing import List, Optional, Dict, Any
from typing import List, Optional, Dict, Any, Tuple
import sqlite3
import json
import uuid
@@ -123,6 +123,70 @@ class ProvenanceStorage(ABC):
"""Internal retrieve method using an active connection/transaction."""
return self.retrieve(entity_id)
def get_chain_head(self, conn: Any = None) -> Optional[Tuple[int, str]]:
"""
Return (sequence_id, checksum) of the most recently stored entry, or
None if the store is empty or this backend doesn't support chaining.
Used by ProvenanceManager._save_entry to link each new entry to its
predecessor (issue #825, Part A item 2 — hash-chained integrity).
Default implementation reports no chaining support; override in
backends that maintain sequence_id.
"""
return None
def trace_descendants(
self, entity_id: str, max_depth: Optional[int] = None
) -> List[ProvenanceEntry]:
"""
Trace downstream descendants (reverse lineage) of an entity.
Generic default implemented purely against retrieve_all(): builds a
reverse adjacency map from parent_entity_id/previous_version_id/
derived_from_id/used_entities, then BFS from entity_id. Correct for
any backend; SQLiteStorage overrides this with indexed queries.
Args:
entity_id: Entity identifier to find descendants of
max_depth: Optional maximum BFS depth
Returns:
List of ProvenanceEntry objects that (transitively) reference
entity_id, in BFS order.
"""
reverse_index: Dict[str, List[ProvenanceEntry]] = {}
for entry in self.retrieve_all():
referenced = set(
filter(
None,
[
entry.parent_entity_id,
entry.previous_version_id,
entry.derived_from_id,
],
)
)
referenced.update(entry.used_entities or [])
for ref_id in referenced:
reverse_index.setdefault(ref_id, []).append(entry)
descendants = []
visited = {entity_id}
queue = deque([(entity_id, 0)])
while queue:
current_id, depth = queue.popleft()
if max_depth is not None and depth >= max_depth:
continue
for child in reverse_index.get(current_id, []):
if child.entity_id in visited:
continue
visited.add(child.entity_id)
descendants.append(child)
queue.append((child.entity_id, depth + 1))
return descendants
class InMemoryStorage(ProvenanceStorage):
"""
@@ -150,6 +214,11 @@ class InMemoryStorage(ProvenanceStorage):
"""Initialize in-memory storage."""
self._entries: Dict[str, ProvenanceEntry] = {}
self._local = threading.local()
# Hash-chain state (issue #825, Part A item 2). Best-effort: this is
# the dev/test backend, not the durable audit backend (SQLiteStorage).
self._seq_lock = threading.Lock()
self._seq_counter = 0
self._chain_head: Optional[str] = None
def _get_pending_stack(self) -> list:
if not hasattr(self._local, "pending_stack"):
@@ -163,6 +232,10 @@ class InMemoryStorage(ProvenanceStorage):
Args:
entry: ProvenanceEntry to store
"""
with self._seq_lock:
if entry.sequence_id is not None:
self._seq_counter = max(self._seq_counter, entry.sequence_id)
self._chain_head = entry.checksum
self._entries[entry.entity_id] = entry
def retrieve(self, entity_id: str) -> Optional[ProvenanceEntry]:
@@ -286,6 +359,31 @@ class InMemoryStorage(ProvenanceStorage):
return pending[entity_id]
return self.retrieve(entity_id)
def get_chain_head(self, conn: Any = None) -> Optional[Tuple[int, str]]:
"""Return (sequence_id, checksum) of the most recent entry, including
entries staged but not yet committed within the current transaction.
Must compare against the *committed* head too, not just the pending
stack: a transaction can stage an unchanged relabel (track_entity's
archival, or invalidate()'s pre-invalidation archive) whose
sequence_id is older than the true head from a prior, already-
committed transaction. Blindly preferring "last staged entry"
understates the head and corrupts the chain for the next append.
"""
with self._seq_lock:
best_seq = self._seq_counter
best_checksum = self._chain_head
for pending in self._get_pending_stack():
for staged_entry in pending.values():
if staged_entry.sequence_id is not None and (
best_checksum is None or staged_entry.sequence_id > best_seq
):
best_seq = staged_entry.sequence_id
best_checksum = staged_entry.checksum
if best_checksum is None:
return None
return (best_seq, best_checksum)
class SQLiteStorage(ProvenanceStorage):
"""
@@ -461,7 +559,27 @@ class SQLiteStorage(ProvenanceStorage):
end_index INTEGER,
credibility REAL,
metadata TEXT,
version TEXT DEFAULT '1.0'
version TEXT DEFAULT '1.0',
agent_type TEXT DEFAULT 'software_agent',
is_automated INTEGER DEFAULT 1,
role TEXT,
sequence_id INTEGER,
previous_checksum TEXT,
previous_version_id TEXT,
derived_from_id TEXT,
invalidated INTEGER DEFAULT 0,
invalidated_at_time TEXT,
invalidated_by TEXT,
invalidation_reason TEXT,
activity_started_at_time TEXT,
activity_ended_at_time TEXT,
acted_on_behalf_of TEXT,
informed_by_activities TEXT,
valid_from TEXT,
valid_until TEXT,
revision_type TEXT,
supersedes TEXT,
bundle_id TEXT
)
""")
@@ -481,6 +599,31 @@ class SQLiteStorage(ProvenanceStorage):
ON provenance(parent_entity_id)
""")
cursor.execute("""
CREATE INDEX IF NOT EXISTS idx_previous_version_id
ON provenance(previous_version_id)
""")
cursor.execute("""
CREATE INDEX IF NOT EXISTS idx_derived_from_id
ON provenance(derived_from_id)
""")
cursor.execute("""
CREATE INDEX IF NOT EXISTS idx_sequence_id
ON provenance(sequence_id)
""")
cursor.execute("""
CREATE INDEX IF NOT EXISTS idx_invalidated
ON provenance(invalidated)
""")
cursor.execute("""
CREATE INDEX IF NOT EXISTS idx_bundle_id
ON provenance(bundle_id)
""")
conn.commit()
finally:
conn.close()
@@ -500,7 +643,9 @@ class SQLiteStorage(ProvenanceStorage):
cursor = conn.cursor()
cursor.execute("""
INSERT OR REPLACE INTO provenance VALUES (
?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?
?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?,
?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?,
?, ?, ?, ?, ?, ?, ?, ?, ?
)
""", (
entry.entity_id,
@@ -521,7 +666,27 @@ class SQLiteStorage(ProvenanceStorage):
entry.end_index,
entry.credibility,
json.dumps(entry.metadata),
entry.version
entry.version,
entry.agent_type,
1 if entry.is_automated else 0,
entry.role,
entry.sequence_id,
entry.previous_checksum,
entry.previous_version_id,
entry.derived_from_id,
1 if entry.invalidated else 0,
entry.invalidated_at_time,
entry.invalidated_by,
entry.invalidation_reason,
entry.activity_started_at_time,
entry.activity_ended_at_time,
entry.acted_on_behalf_of,
json.dumps(entry.informed_by_activities),
entry.valid_from,
entry.valid_until,
entry.revision_type,
entry.supersedes,
entry.bundle_id,
))
def retrieve(self, entity_id: str) -> Optional[ProvenanceEntry]:
@@ -637,6 +802,120 @@ class SQLiteStorage(ProvenanceStorage):
return lineage
def get_chain_head(self, conn: Any = None) -> Optional[Tuple[int, str]]:
"""Return (sequence_id, checksum) of the most recently stored entry.
Reuses the caller's transaction connection when given, so the head
read and the subsequent insert happen atomically under the existing
BEGIN IMMEDIATE writer lock (issue #825, Part A item 2).
"""
# Tie-break on rowid (SQLite's implicit row-order column) in addition
# to sequence_id: during track_entity's archival, the just-relabeled
# history row briefly shares its sequence_id with the still-present
# (about-to-be-overwritten) live row it was copied from. rowid DESC
# deterministically picks the most-recently-written row of the two,
# which is the correct chain head — relying on sequence_id ordering
# alone leaves the tie-break implementation-defined.
query = (
"SELECT sequence_id, checksum FROM provenance "
"WHERE sequence_id IS NOT NULL "
"ORDER BY sequence_id DESC, rowid DESC LIMIT 1"
)
if conn is not None:
row = conn.cursor().execute(query).fetchone()
return (row[0], row[1]) if row else None
with self._read_connection() as read_conn:
row = read_conn.cursor().execute(query).fetchone()
return (row[0], row[1]) if row else None
def trace_descendants(
self, entity_id: str, max_depth: Optional[int] = None
) -> List[ProvenanceEntry]:
"""
Trace downstream descendants (reverse lineage) using indexed lookups
on parent_entity_id/previous_version_id/derived_from_id, plus an
in-memory reverse index over used_entities (a JSON column, so not
directly indexable in SQLite).
Args:
entity_id: Entity identifier to find descendants of
max_depth: Optional maximum BFS depth
Returns:
List of ProvenanceEntry objects that (transitively) reference
entity_id, in BFS order.
"""
descendants: List[ProvenanceEntry] = []
visited = {entity_id}
frontier = [entity_id]
depth = 0
with self._read_connection() as conn:
cursor = conn.cursor()
cursor.execute(
"SELECT entity_id, used_entities FROM provenance "
"WHERE used_entities IS NOT NULL AND used_entities != '[]'"
)
used_reverse: Dict[str, List[str]] = {}
for child_id, used_json in cursor.fetchall():
try:
used_ids = json.loads(used_json) if used_json else []
except (json.JSONDecodeError, TypeError):
used_ids = []
for uid in used_ids:
used_reverse.setdefault(uid, []).append(child_id)
while frontier and (max_depth is None or depth < max_depth):
current_frontier = list(dict.fromkeys(frontier))
if not current_frontier:
break
child_ids = set()
for i in range(0, len(current_frontier), 999):
chunk = current_frontier[i : i + 999]
placeholders = ",".join("?" * len(chunk))
cursor.execute(
f"""
SELECT entity_id FROM provenance
WHERE parent_entity_id IN ({placeholders})
OR previous_version_id IN ({placeholders})
OR derived_from_id IN ({placeholders})
""",
chunk * 3,
)
child_ids.update(row[0] for row in cursor.fetchall())
for ref_id in current_frontier:
child_ids.update(used_reverse.get(ref_id, []))
next_frontier = [cid for cid in child_ids if cid not in visited]
if not next_frontier:
break
visited.update(next_frontier)
entries_map = {}
for i in range(0, len(next_frontier), 999):
chunk = next_frontier[i : i + 999]
placeholders = ",".join("?" * len(chunk))
cursor.execute(
f"SELECT * FROM provenance WHERE entity_id IN ({placeholders})",
chunk,
)
for row in cursor.fetchall():
entry = self._row_to_entry(row)
entries_map[entry.entity_id] = entry
for cid in next_frontier:
entry = entries_map.get(cid)
if entry:
descendants.append(entry)
frontier = next_frontier
depth += 1
return descendants
def clear(self) -> int:
"""
Clear all provenance data.
@@ -687,5 +966,25 @@ class SQLiteStorage(ProvenanceStorage):
end_index=row[15],
credibility=row[16],
metadata=json.loads(row[17]) if row[17] else {},
version=row[18]
version=row[18],
agent_type=row[19] or "software_agent",
is_automated=bool(row[20]) if row[20] is not None else True,
role=row[21],
sequence_id=row[22],
previous_checksum=row[23],
previous_version_id=row[24],
derived_from_id=row[25],
invalidated=bool(row[26]) if row[26] is not None else False,
invalidated_at_time=row[27],
invalidated_by=row[28],
invalidation_reason=row[29],
activity_started_at_time=row[30],
activity_ended_at_time=row[31],
acted_on_behalf_of=row[32],
informed_by_activities=json.loads(row[33]) if row[33] else [],
valid_from=row[34],
valid_until=row[35],
revision_type=row[36],
supersedes=row[37],
bundle_id=row[38],
)
semantica/reasoning/reasoning_provenance.py
+18 -2
View File
@@ -13,19 +13,28 @@ Author: Semantica Contributors
License: MIT
"""
from typing import Any
from typing import Any, Optional
from datetime import datetime
import uuid
class ReasoningEngineWithProvenance:
"""Reasoning engine with provenance tracking."""
def __init__(self, provenance: bool = False, **config):
def __init__(
self,
provenance: bool = False,
agent_id: Optional[str] = None,
is_automated: bool = True,
**config,
):
from .reasoning_engine import ReasoningEngine
self.provenance = provenance
self._engine = ReasoningEngine(**config)
self._prov_manager = None
self._agent_id = agent_id or self.__class__.__name__
self._is_automated = is_automated
if provenance:
try:
@@ -36,13 +45,20 @@ class ReasoningEngineWithProvenance:
def infer(self, premises: Any, source: str = None, **kwargs):
"""Perform inference with provenance tracking."""
activity_started_at_time = datetime.utcnow().isoformat()
result = self._engine.infer(premises, **kwargs)
activity_ended_at_time = datetime.utcnow().isoformat()
if self.provenance and self._prov_manager:
self._prov_manager.track_entity(
entity_id=f"inference_{uuid.uuid4().hex[:8]}",
source=source or "reasoning_engine",
entity_type="inference",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=activity_started_at_time,
activity_ended_at_time=activity_ended_at_time,
metadata={
"premises_count": len(premises) if hasattr(premises, '__len__') else 1,
"confidence": getattr(result, 'confidence', None)
semantica/semantic_extract/semantic_extract_provenance.py
+95 -16
View File
@@ -37,6 +37,7 @@ License: MIT
"""
from typing import Optional, List, Dict, Any
from datetime import datetime
import uuid
@@ -48,16 +49,27 @@ class ProvenanceMixin:
added to any extraction class without modifying its core functionality.
"""
def __init__(self, provenance: bool = False, **kwargs):
def __init__(
self,
provenance: bool = False,
agent_id: Optional[str] = None,
is_automated: bool = True,
**kwargs,
):
"""
Initialize provenance tracking.
Args:
provenance: Enable provenance tracking (default: False)
agent_id: Agent identifier for accountability (issue #825); defaults
to the wrapping class name
is_automated: Whether this agent acted without direct human review
**kwargs: Additional arguments passed to parent class
"""
self.provenance = provenance
self._prov_manager = None
self._agent_id = agent_id or self.__class__.__name__
self._is_automated = is_automated
if provenance:
try:
@@ -84,10 +96,17 @@ class ProvenanceMixin:
**metadata: Additional metadata to track
"""
if self.provenance and self._prov_manager:
activity_started_at_time = metadata.pop("activity_started_at_time", None)
activity_ended_at_time = metadata.pop("activity_ended_at_time", None)
self._prov_manager.track_entity(
entity_id=entity_id,
source=source,
entity_type=entity_type,
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=activity_started_at_time,
activity_ended_at_time=activity_ended_at_time,
metadata=metadata
)
@@ -105,7 +124,13 @@ class NERExtractorWithProvenance(ProvenanceMixin):
>>> # Each entity is tracked with source, confidence, and metadata
"""
def __init__(self, provenance: bool = False, **config):
def __init__(
self,
provenance: bool = False,
agent_id: Optional[str] = None,
is_automated: bool = True,
**config,
):
"""
Initialize NER extractor with optional provenance.
@@ -115,7 +140,9 @@ class NERExtractorWithProvenance(ProvenanceMixin):
"""
from .ner_extractor import NERExtractor
ProvenanceMixin.__init__(self, provenance=provenance)
ProvenanceMixin.__init__(
self, provenance=provenance, agent_id=agent_id, is_automated=is_automated
)
self._extractor = NERExtractor(**config)
def extract(self, text: str, source: Optional[str] = None, **kwargs):
@@ -130,7 +157,9 @@ class NERExtractorWithProvenance(ProvenanceMixin):
Returns:
List of extracted entities (same as original NERExtractor)
"""
activity_started_at_time = datetime.utcnow().isoformat()
entities = self._extractor.extract(text, **kwargs)
activity_ended_at_time = datetime.utcnow().isoformat()
if self.provenance:
for entity in entities:
@@ -150,7 +179,9 @@ class NERExtractorWithProvenance(ProvenanceMixin):
label=entity.label,
confidence=getattr(entity, 'confidence', 1.0),
start=entity.start,
end=entity.end
end=entity.end,
activity_started_at_time=activity_started_at_time,
activity_ended_at_time=activity_ended_at_time,
)
return entities
@@ -167,7 +198,13 @@ class RelationExtractorWithProvenance(ProvenanceMixin):
Wraps the original RelationExtractor and tracks all extracted relations.
"""
def __init__(self, provenance: bool = False, **config):
def __init__(
self,
provenance: bool = False,
agent_id: Optional[str] = None,
is_automated: bool = True,
**config,
):
"""
Initialize relation extractor with optional provenance.
@@ -177,7 +214,9 @@ class RelationExtractorWithProvenance(ProvenanceMixin):
"""
from .relation_extractor import RelationExtractor
ProvenanceMixin.__init__(self, provenance=provenance)
ProvenanceMixin.__init__(
self, provenance=provenance, agent_id=agent_id, is_automated=is_automated
)
self._extractor = RelationExtractor(**config)
def extract(self, text: str, source: Optional[str] = None, **kwargs):
@@ -192,7 +231,9 @@ class RelationExtractorWithProvenance(ProvenanceMixin):
Returns:
List of extracted relations
"""
activity_started_at_time = datetime.utcnow().isoformat()
relations = self._extractor.extract(text, **kwargs)
activity_ended_at_time = datetime.utcnow().isoformat()
if self.provenance:
for relation in relations:
@@ -211,7 +252,9 @@ class RelationExtractorWithProvenance(ProvenanceMixin):
subject=relation.subject,
predicate=relation.predicate,
object=relation.object,
confidence=getattr(relation, 'confidence', 1.0)
confidence=getattr(relation, 'confidence', 1.0),
activity_started_at_time=activity_started_at_time,
activity_ended_at_time=activity_ended_at_time,
)
return relations
@@ -228,7 +271,13 @@ class EventDetectorWithProvenance(ProvenanceMixin):
Wraps the original EventDetector and tracks all detected events.
"""
def __init__(self, provenance: bool = False, **config):
def __init__(
self,
provenance: bool = False,
agent_id: Optional[str] = None,
is_automated: bool = True,
**config,
):
"""
Initialize event detector with optional provenance.
@@ -238,7 +287,9 @@ class EventDetectorWithProvenance(ProvenanceMixin):
"""
from .event_detector import EventDetector
ProvenanceMixin.__init__(self, provenance=provenance)
ProvenanceMixin.__init__(
self, provenance=provenance, agent_id=agent_id, is_automated=is_automated
)
self._detector = EventDetector(**config)
def detect(self, text: str, source: Optional[str] = None, **kwargs):
@@ -253,7 +304,9 @@ class EventDetectorWithProvenance(ProvenanceMixin):
Returns:
List of detected events
"""
activity_started_at_time = datetime.utcnow().isoformat()
events = self._detector.detect(text, **kwargs)
activity_ended_at_time = datetime.utcnow().isoformat()
if self.provenance:
for event in events:
@@ -271,7 +324,9 @@ class EventDetectorWithProvenance(ProvenanceMixin):
entity_type="event",
event_type=event.type,
trigger=event.trigger,
confidence=getattr(event, 'confidence', 1.0)
confidence=getattr(event, 'confidence', 1.0),
activity_started_at_time=activity_started_at_time,
activity_ended_at_time=activity_ended_at_time,
)
return events
@@ -288,7 +343,13 @@ class CoreferenceResolverWithProvenance(ProvenanceMixin):
Wraps the original CoreferenceResolver and tracks coreference chains.
"""
def __init__(self, provenance: bool = False, **config):
def __init__(
self,
provenance: bool = False,
agent_id: Optional[str] = None,
is_automated: bool = True,
**config,
):
"""
Initialize coreference resolver with optional provenance.
@@ -298,7 +359,9 @@ class CoreferenceResolverWithProvenance(ProvenanceMixin):
"""
from .coreference_resolver import CoreferenceResolver
ProvenanceMixin.__init__(self, provenance=provenance)
ProvenanceMixin.__init__(
self, provenance=provenance, agent_id=agent_id, is_automated=is_automated
)
self._resolver = CoreferenceResolver(**config)
def resolve(self, text: str, source: Optional[str] = None, **kwargs):
@@ -313,7 +376,9 @@ class CoreferenceResolverWithProvenance(ProvenanceMixin):
Returns:
Coreference chains
"""
activity_started_at_time = datetime.utcnow().isoformat()
chains = self._resolver.resolve(text, **kwargs)
activity_ended_at_time = datetime.utcnow().isoformat()
if self.provenance:
for chain in chains:
@@ -329,7 +394,9 @@ class CoreferenceResolverWithProvenance(ProvenanceMixin):
entity_id=chain_id,
source=source or text[:100],
entity_type="coreference_chain",
mentions=len(chain.mentions) if hasattr(chain, 'mentions') else 0
mentions=len(chain.mentions) if hasattr(chain, 'mentions') else 0,
activity_started_at_time=activity_started_at_time,
activity_ended_at_time=activity_ended_at_time,
)
return chains
@@ -346,7 +413,13 @@ class TripletExtractorWithProvenance(ProvenanceMixin):
Wraps the original TripletExtractor and tracks all extracted triplets.
"""
def __init__(self, provenance: bool = False, **config):
def __init__(
self,
provenance: bool = False,
agent_id: Optional[str] = None,
is_automated: bool = True,
**config,
):
"""
Initialize triplet extractor with optional provenance.
@@ -356,7 +429,9 @@ class TripletExtractorWithProvenance(ProvenanceMixin):
"""
from .triplet_extractor import TripletExtractor
ProvenanceMixin.__init__(self, provenance=provenance)
ProvenanceMixin.__init__(
self, provenance=provenance, agent_id=agent_id, is_automated=is_automated
)
self._extractor = TripletExtractor(**config)
def extract(self, text: str, source: Optional[str] = None, **kwargs):
@@ -371,7 +446,9 @@ class TripletExtractorWithProvenance(ProvenanceMixin):
Returns:
List of extracted triplets
"""
activity_started_at_time = datetime.utcnow().isoformat()
triplets = self._extractor.extract(text, **kwargs)
activity_ended_at_time = datetime.utcnow().isoformat()
if self.provenance:
for triplet in triplets:
@@ -390,7 +467,9 @@ class TripletExtractorWithProvenance(ProvenanceMixin):
subject=triplet.subject,
predicate=triplet.predicate,
object=triplet.object,
confidence=getattr(triplet, 'confidence', 1.0)
confidence=getattr(triplet, 'confidence', 1.0),
activity_started_at_time=activity_started_at_time,
activity_ended_at_time=activity_ended_at_time,
)
return triplets
semantica/triplet_store/triplet_store_provenance.py
+18 -2
View File
@@ -13,19 +13,28 @@ Author: Semantica Contributors
License: MIT
"""
from typing import Any
from typing import Any, Optional
from datetime import datetime
import uuid
class TripletStoreWithProvenance:
"""Triplet store with provenance tracking."""
def __init__(self, provenance: bool = False, **config):
def __init__(
self,
provenance: bool = False,
agent_id: Optional[str] = None,
is_automated: bool = True,
**config,
):
from .triplet_store import TripletStore
self.provenance = provenance
self._store = TripletStore(**config)
self._prov_manager = None
self._agent_id = agent_id or self.__class__.__name__
self._is_automated = is_automated
if provenance:
try:
@@ -36,13 +45,20 @@ class TripletStoreWithProvenance:
def add_triplet(self, subject: Any, predicate: Any, obj: Any, source: str = None, **kwargs):
"""Add triplet with provenance tracking."""
activity_started_at_time = datetime.utcnow().isoformat()
result = self._store.add_triplet(subject, predicate, obj, **kwargs)
activity_ended_at_time = datetime.utcnow().isoformat()
if self.provenance and self._prov_manager:
self._prov_manager.track_entity(
entity_id=f"triplet_{uuid.uuid4().hex[:8]}",
source=source or "triplet_store",
entity_type="triplet",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=activity_started_at_time,
activity_ended_at_time=activity_ended_at_time,
metadata={
"subject": str(subject),
"predicate": str(predicate),
semantica/vector_store/vector_store_provenance.py
+18 -2
View File
@@ -13,19 +13,28 @@ Author: Semantica Contributors
License: MIT
"""
from typing import List, Any
from typing import List, Any, Optional
from datetime import datetime
import uuid
class VectorStoreWithProvenance:
"""Vector store with provenance tracking."""
def __init__(self, provenance: bool = False, **config):
def __init__(
self,
provenance: bool = False,
agent_id: Optional[str] = None,
is_automated: bool = True,
**config,
):
from .vector_store import VectorStore
self.provenance = provenance
self._store = VectorStore(**config)
self._prov_manager = None
self._agent_id = agent_id or self.__class__.__name__
self._is_automated = is_automated
if provenance:
try:
@@ -36,13 +45,20 @@ class VectorStoreWithProvenance:
def add_vectors(self, vectors: List[Any], source: str = None, **kwargs):
"""Add vectors with provenance tracking."""
activity_started_at_time = datetime.utcnow().isoformat()
result = self._store.add_vectors(vectors, **kwargs)
activity_ended_at_time = datetime.utcnow().isoformat()
if self.provenance and self._prov_manager:
self._prov_manager.track_entity(
entity_id=f"vectors_{uuid.uuid4().hex[:8]}",
source=source or "vector_store",
entity_type="vector_collection",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=activity_started_at_time,
activity_ended_at_time=activity_ended_at_time,
metadata={
"count": len(vectors),
"dimensions": len(vectors[0]) if vectors else 0
semantica/visualization/visualization_provenance.py
+18 -2
View File
@@ -11,19 +11,28 @@ Author: Semantica Contributors
License: MIT
"""
from typing import Any
from typing import Any, Optional
from datetime import datetime
import uuid
class VisualizerWithProvenance:
"""Visualizer with provenance tracking."""
def __init__(self, provenance: bool = False, **config):
def __init__(
self,
provenance: bool = False,
agent_id: Optional[str] = None,
is_automated: bool = True,
**config,
):
from .visualizer import Visualizer
self.provenance = provenance
self._visualizer = Visualizer(**config)
self._prov_manager = None
self._agent_id = agent_id or self.__class__.__name__
self._is_automated = is_automated
if provenance:
try:
@@ -34,13 +43,20 @@ class VisualizerWithProvenance:
def visualize(self, data: Any, output: str = None, **kwargs):
"""Visualize data with provenance tracking."""
activity_started_at_time = datetime.utcnow().isoformat()
result = self._visualizer.visualize(data, output=output, **kwargs)
activity_ended_at_time = datetime.utcnow().isoformat()
if self.provenance and self._prov_manager:
self._prov_manager.track_entity(
entity_id=f"viz_{uuid.uuid4().hex[:8]}",
source="visualization",
entity_type="visualization",
agent_id=self._agent_id,
agent_type="software_agent",
is_automated=self._is_automated,
activity_started_at_time=activity_started_at_time,
activity_ended_at_time=activity_ended_at_time,
metadata={"output": output, "type": kwargs.get('type', 'unknown')}
)
tests/provenance/test_manager.py
+609
View File
@@ -913,5 +913,614 @@ class TestProvenanceManager:
assert kwargs.get("exc_info") is True
class TestAgentTyping:
"""Issue #825, Part A item 3 — agent_id/agent_type/is_automated actually
populate on tracked entries (previously a dead field: no track_* method
read agent_id from kwargs at all)."""
def test_track_entity_scalar_agent_kwargs(self):
prov_mgr = ProvenanceManager()
entry = prov_mgr.track_entity(
"e1", source="doc1",
agent_id="alice", agent_type="person", is_automated=False, role="approver",
)
assert entry.agent_id == "alice"
assert entry.agent_type == "person"
assert entry.is_automated is False
assert entry.role == "approver"
def test_track_entity_agent_record_kwarg(self):
from semantica.provenance import AgentRecord
prov_mgr = ProvenanceManager()
agent = AgentRecord(id="reviewer_bob", agent_type="person", is_automated=False)
entry = prov_mgr.track_entity("e1", source="doc1", agent=agent, role="approver")
assert entry.agent_id == "reviewer_bob"
assert entry.agent_type == "person"
assert entry.is_automated is False
assert entry.role == "approver"
def test_track_entity_default_agent_unchanged(self):
"""No agent kwargs supplied should still default to 'semantica' (back-compat)."""
prov_mgr = ProvenanceManager()
entry = prov_mgr.track_entity("e1", source="doc1")
assert entry.agent_id == "semantica"
assert entry.agent_type == "software_agent"
assert entry.is_automated is True
def test_track_relationship_agent_kwargs(self):
prov_mgr = ProvenanceManager()
entry = prov_mgr.track_relationship("r1", source="doc1", agent_id="bot1")
assert entry.agent_id == "bot1"
def test_track_chunk_agent_kwargs_not_leaked_into_metadata(self):
"""agent_id/agent_type/is_automated passed to track_chunk must populate
the real fields, not leak into the opaque metadata blob."""
prov_mgr = ProvenanceManager()
entry = prov_mgr.track_chunk(
"c1", source_document="doc1", agent_id="chunker_v2", note="hello"
)
assert entry.agent_id == "chunker_v2"
assert "agent_id" not in entry.metadata
assert entry.metadata.get("note") == "hello"
def test_track_property_source_agent_kwargs(self):
prov_mgr = ProvenanceManager()
source = SourceReference(document="doc1")
entry = prov_mgr.track_property_source(
"e1", "prop", "value", source, agent_id="prop_tracker"
)
assert entry.agent_id == "prop_tracker"
assert "agent_id" not in entry.metadata
def test_track_entities_batch_agent_id_not_swallowed_into_metadata(self):
"""Regression test for the documented bug: track_entities_batch used to
merge agent_id/entity_type/activity_id into the metadata dict instead
of forwarding them as real track_entity kwargs."""
prov_mgr = ProvenanceManager()
count = prov_mgr.track_entities_batch(
[{"id": "b1"}, {"id": "b2"}],
source="doc_batch",
agent_id="batch_service_v2",
entity_type="credit_feature",
activity_id="bureau_parsing",
extra_note="kept as free-form metadata",
)
assert count == 2
for eid in ("b1", "b2"):
entry = prov_mgr.storage.retrieve(eid)
assert entry.agent_id == "batch_service_v2"
assert entry.entity_type == "credit_feature"
assert entry.activity_id == "bureau_parsing"
assert "agent_id" not in entry.metadata
assert entry.metadata.get("extra_note") == "kept as free-form metadata"
class TestVersioningVsDerivation:
"""Issue #825, Part A item 4 — previous_version_id (correction) is
additive alongside derived_from_id (cross-source derivation); both are
populated without disturbing the legacy parent_entity_id field."""
def test_retrack_without_parent_sets_previous_version_id_only(self):
prov_mgr = ProvenanceManager()
prov_mgr.track_entity("X", source="doc_1")
v2 = prov_mgr.track_entity("X", source="doc_1")
assert v2.previous_version_id is not None
assert v2.previous_version_id.startswith("X:v:")
assert v2.derived_from_id is None
# Legacy field behavior is unchanged
assert v2.parent_entity_id == v2.previous_version_id
def test_retrack_with_explicit_parent_sets_both_fields(self):
prov_mgr = ProvenanceManager()
prov_mgr.track_entity("X", source="doc_1")
v2 = prov_mgr.track_entity("X", source="doc_1", parent_entity_id="explicit_parent")
assert v2.derived_from_id == "explicit_parent"
assert v2.previous_version_id is not None
assert v2.previous_version_id.startswith("X:v:")
# Legacy field keeps explicit-wins semantics
assert v2.parent_entity_id == "explicit_parent"
def test_track_chunk_split_sets_derived_from_id(self):
prov_mgr = ProvenanceManager()
entry = prov_mgr.track_chunk(
"c2", source_document="doc1", parent_chunk_id="c1"
)
assert entry.derived_from_id == "c1"
assert entry.previous_version_id is None
assert entry.parent_entity_id == "c1"
class TestInvalidation:
"""Issue #825, Part A item 1 — tombstone instead of hard delete."""
def test_invalidate_marks_entry_without_deleting(self):
prov_mgr = ProvenanceManager()
prov_mgr.track_entity("e1", source="doc1")
result = prov_mgr.invalidate("e1", agent_id="reviewer_jane", reason="retracted")
assert result.invalidated is True
assert result.invalidated_by == "reviewer_jane"
assert result.invalidation_reason == "retracted"
assert result.invalidated_at_time is not None
# Entry remains visible via retrieve (tombstone, not delete)
stored = prov_mgr.storage.retrieve("e1")
assert stored is not None
assert stored.invalidated is True
def test_invalidate_unknown_entity_raises(self):
prov_mgr = ProvenanceManager()
with pytest.raises(ValueError):
prov_mgr.invalidate("never_tracked", agent_id="reviewer_jane")
def test_check_reports_invalidated_count(self):
prov_mgr = ProvenanceManager()
prov_mgr.track_entity("e1", source="doc1")
prov_mgr.track_entity("e2", source="doc1")
prov_mgr.invalidate("e1", agent_id="reviewer_jane")
result = prov_mgr.check()
assert result["invalidated_count"] == 1
class TestHashChain:
"""Issue #825, Part A item 2 — hash-chained integrity: chained checksums
detect wholesale row deletion, which per-row checksums alone cannot."""
def test_verify_chain_valid_on_clean_history(self):
prov_mgr = ProvenanceManager()
prov_mgr.track_entity("e1", source="doc1")
prov_mgr.track_entity("e2", source="doc1")
prov_mgr.track_entity("X", source="doc1")
prov_mgr.track_entity("X", source="doc1") # triggers archival
prov_mgr.track_entity("X", source="doc1", parent_entity_id="explicit_p")
result = prov_mgr.verify_chain()
assert result["valid"] is True
assert result["broken_links"] == []
assert result["total_entries"] == 5
def test_verify_chain_detects_deleted_row(self, tmp_path):
"""A hard delete of a row must break the chain for whatever followed it."""
import sqlite3
db_path = str(tmp_path / "chain.db")
prov_mgr = ProvenanceManager(storage_path=db_path)
prov_mgr.track_entity("e1", source="doc1")
prov_mgr.track_entity("e2", source="doc1")
prov_mgr.track_entity("e3", source="doc1")
conn = sqlite3.connect(db_path)
conn.execute("DELETE FROM provenance WHERE entity_id = 'e2'")
conn.commit()
conn.close()
prov_mgr2 = ProvenanceManager(storage_path=db_path)
result = prov_mgr2.verify_chain()
assert result["valid"] is False
assert any(link["reason"] == "chain_break" for link in result["broken_links"])
def test_sequence_ids_are_assigned_and_monotonic(self):
prov_mgr = ProvenanceManager()
e1 = prov_mgr.track_entity("e1", source="doc1")
e2 = prov_mgr.track_entity("e2", source="doc1")
assert e1.sequence_id is not None
assert e2.sequence_id is not None
assert e2.sequence_id > e1.sequence_id
assert e2.previous_checksum == e1.checksum
def test_chain_survives_interleaved_retrack_and_invalidate(self, tmp_path):
"""Regression test: an entry that already chained its previous_checksum
from another entity's checksum (Y -> X) must stay valid even after X
is later retracked (archived/relabeled) and then invalidated — both
of which write NEW entries under X's canonical key. Mutating X's
existing row in place (rather than archiving-then-appending) used to
silently orphan Y's chain link, producing a false-positive break."""
db_path = str(tmp_path / "interleaved.db")
prov_mgr = ProvenanceManager(storage_path=db_path)
prov_mgr.track_entity("e1", source="doc1")
prov_mgr.track_entity("X", source="doc1")
prov_mgr.track_entity("Y", source="doc1", parent_entity_id="X")
prov_mgr.track_entity("X", source="doc1") # retrack: archives old X
prov_mgr.track_entity("Z", source="doc1", parent_entity_id="X")
prov_mgr.invalidate("X", agent_id="reviewer") # archives again
prov_mgr.track_entity("W", source="doc1", parent_entity_id="X")
result = prov_mgr.verify_chain()
assert result["valid"] is True
assert result["broken_links"] == []
def test_invalidate_does_not_break_chain(self):
prov_mgr = ProvenanceManager()
prov_mgr.track_entity("e1", source="doc1")
prov_mgr.track_entity("e2", source="doc1", parent_entity_id="e1")
prov_mgr.invalidate("e1", agent_id="reviewer_jane")
result = prov_mgr.verify_chain()
assert result["valid"] is True
class TestDownstreamLineage:
"""Issue #825, Part A item 5 — downstream/descendant traversal (reverse
BFS), complementing the existing upstream-only trace_lineage/get_lineage."""
def test_get_descendants_finds_direct_child(self):
prov_mgr = ProvenanceManager()
prov_mgr.track_entity("parent1", source="doc1")
prov_mgr.track_entity(
"child1", source="doc1", parent_entity_id="parent1", used_entities=["parent1"]
)
result = prov_mgr.get_descendants("parent1")
entity_ids = {e["entity_id"] for e in result["entries"]}
assert "child1" in entity_ids
def test_get_descendants_transitive(self):
prov_mgr = ProvenanceManager()
prov_mgr.track_entity("a", source="doc1")
prov_mgr.track_entity("b", source="doc1", parent_entity_id="a")
prov_mgr.track_entity("c", source="doc1", parent_entity_id="b")
result = prov_mgr.get_descendants("a")
entity_ids = {e["entity_id"] for e in result["entries"]}
assert entity_ids == {"b", "c"}
def test_get_descendants_empty_for_leaf(self):
prov_mgr = ProvenanceManager()
prov_mgr.track_entity("leaf", source="doc1")
assert prov_mgr.get_descendants("leaf") == {}
def test_descendants_work_on_sqlite_backend(self, tmp_path):
db_path = str(tmp_path / "desc.db")
prov_mgr = ProvenanceManager(storage_path=db_path)
prov_mgr.track_entity("parent1", source="doc1")
prov_mgr.track_entity("child1", source="doc1", parent_entity_id="parent1")
result = prov_mgr.get_descendants("parent1")
entity_ids = {e["entity_id"] for e in result["entries"]}
assert "child1" in entity_ids
class TestQualifiedExport:
"""Issue #825, Part A item 6 — qualified Association with hadRole, plus
qualified Invalidation, in the RDF export."""
def test_export_prov_includes_qualified_association_and_role(self):
prov_mgr = ProvenanceManager()
prov_mgr.track_entity(
"e1", source="doc1", agent_id="alice", agent_type="person", role="approver"
)
ttl = prov_mgr.export_prov(format="turtle")
assert "qualifiedAssociation" in ttl
assert "hadRole" in ttl
assert "role_approver" in ttl
assert "Person" in ttl
def test_export_prov_includes_qualified_invalidation(self):
prov_mgr = ProvenanceManager()
prov_mgr.track_entity("e1", source="doc1")
prov_mgr.invalidate("e1", agent_id="reviewer_jane", reason="retracted")
ttl = prov_mgr.export_prov(format="turtle")
assert "qualifiedInvalidation" in ttl
assert "Invalidation" in ttl
class TestTypedActivity:
"""Issue #825, Part B Tier 1 — typed Activity via ActivityRecord."""
def test_track_entity_activity_record_kwarg(self):
from semantica.provenance import ActivityRecord
prov_mgr = ProvenanceManager()
activity = ActivityRecord(
id="bureau_parsing_run_42",
started_at_time="2026-01-01T00:00:00",
ended_at_time="2026-01-01T00:00:03",
)
entry = prov_mgr.track_entity("e1", source="doc1", activity=activity)
assert entry.activity_id == "bureau_parsing_run_42"
assert entry.activity_started_at_time == "2026-01-01T00:00:00"
assert entry.activity_ended_at_time == "2026-01-01T00:00:03"
def test_track_entity_activity_scalar_kwargs(self):
prov_mgr = ProvenanceManager()
entry = prov_mgr.track_entity(
"e1", source="doc1",
activity_id="parse_step", activity_started_at_time="t0", activity_ended_at_time="t1",
)
assert entry.activity_id == "parse_step"
assert entry.activity_started_at_time == "t0"
assert entry.activity_ended_at_time == "t1"
def test_export_prov_qualified_generation_usage_derivation(self):
prov_mgr = ProvenanceManager()
prov_mgr.track_entity("parent1", source="doc1")
prov_mgr.track_entity(
"child1", source="doc1",
parent_entity_id="parent1", used_entities=["parent1"],
activity_id="transform",
)
ttl = prov_mgr.export_prov(format="turtle")
assert "qualifiedGeneration" in ttl
assert "Generation" in ttl
assert "qualifiedUsage" in ttl
assert "Usage" in ttl
assert "qualifiedDerivation" in ttl
assert "Derivation" in ttl
def test_export_prov_activity_timing(self):
from semantica.provenance import ActivityRecord
prov_mgr = ProvenanceManager()
prov_mgr.track_entity(
"e1", source="doc1",
activity=ActivityRecord(id="act1", started_at_time="2026-01-01T00:00:00",
ended_at_time="2026-01-01T00:00:05"),
)
ttl = prov_mgr.export_prov(format="turtle")
assert "startedAtTime" in ttl
assert "endedAtTime" in ttl
class TestAssociationDelegationChaining:
"""Issue #825, Part B Tier 2 — wasAssociatedWith, actedOnBehalfOf, wasInformedBy."""
def test_acted_on_behalf_of(self):
prov_mgr = ProvenanceManager()
entry = prov_mgr.track_entity(
"e1", source="doc1", agent_id="bot1", acted_on_behalf_of="org1"
)
assert entry.acted_on_behalf_of == "org1"
ttl = prov_mgr.export_prov(format="turtle")
assert "actedOnBehalfOf" in ttl
def test_informed_by_activities(self):
prov_mgr = ProvenanceManager()
entry = prov_mgr.track_entity(
"e1", source="doc1", activity_id="parse", informed_by=["ingest_activity"]
)
assert entry.informed_by_activities == ["ingest_activity"]
ttl = prov_mgr.export_prov(format="turtle")
assert "wasInformedBy" in ttl
def test_was_associated_with_in_export(self):
prov_mgr = ProvenanceManager()
prov_mgr.track_entity("e1", source="doc1", agent_id="alice", activity_id="act1")
ttl = prov_mgr.export_prov(format="turtle")
assert "wasAssociatedWith" in ttl
def test_track_entities_batch_forwards_tier2_kwargs(self):
"""Regression test for the batch-kwargs-vs-metadata bug (issue #825,
Part A) extended to the new Tier 2/3 keys."""
prov_mgr = ProvenanceManager()
count = prov_mgr.track_entities_batch(
[{"id": "b1"}],
source="doc_batch",
activity_id="bureau_parsing",
acted_on_behalf_of="org1",
informed_by=["ingest_activity"],
bundle_id="run_1",
)
assert count == 1
entry = prov_mgr.storage.retrieve("b1")
assert entry.activity_id == "bureau_parsing"
assert entry.acted_on_behalf_of == "org1"
assert entry.informed_by_activities == ["ingest_activity"]
assert entry.bundle_id == "run_1"
assert "acted_on_behalf_of" not in entry.metadata
class TestBitemporalMerge:
"""Issue #825, Part B Tier 3 — revision_history()/query_recorded_between()
close kg.ProvenanceTracker's documented 'no direct equivalent yet' gaps."""
def test_revision_history_ascending_with_valid_until(self):
prov_mgr = ProvenanceManager()
prov_mgr.track_entity("X", source="doc1", agent_id="alice")
prov_mgr.track_entity("X", source="doc1", agent_id="bob", revision_type="correction")
prov_mgr.track_entity("X", source="doc1", agent_id="carol", supersedes="X_old_claim")
history = prov_mgr.revision_history("X")
assert len(history) == 3
assert [h["version"] for h in history] == [1, 2, 3]
assert history[0]["author"] == "alice"
assert history[1]["author"] == "bob"
assert history[1]["revision_type"] == "correction"
assert history[2]["author"] == "carol"
assert history[2]["supersedes"] == "X_old_claim"
# Every version except the last has a valid_until set to the next version's timestamp
assert history[0]["valid_until"] == history[1]["valid_from"]
assert history[1]["valid_until"] == history[2]["valid_from"]
assert history[2]["valid_until"] is None
def test_revision_history_empty_for_untracked_entity(self):
prov_mgr = ProvenanceManager()
assert prov_mgr.revision_history("never_tracked") == []
def test_revision_history_single_version(self):
prov_mgr = ProvenanceManager()
prov_mgr.track_entity("Y", source="doc1")
history = prov_mgr.revision_history("Y")
assert len(history) == 1
assert history[0]["version"] == 1
assert history[0]["valid_until"] is None
def test_query_recorded_between(self):
prov_mgr = ProvenanceManager()
prov_mgr.track_entity("e1", source="doc1")
prov_mgr.track_entity("e2", source="doc1")
results = prov_mgr.query_recorded_between("2000-01-01T00:00:00", "2100-01-01T00:00:00")
entity_ids = {r["entity_id"] for r in results}
assert {"e1", "e2"}.issubset(entity_ids)
no_results = prov_mgr.query_recorded_between("1990-01-01T00:00:00", "1990-01-02T00:00:00")
assert no_results == []
def test_check_flags_missing_informed_by_activity(self):
prov_mgr = ProvenanceManager()
prov_mgr.track_entity("e1", source="doc1", informed_by=["never_tracked_activity"])
result = prov_mgr.check()
assert result["valid"] is False
assert any("never_tracked_activity" in ref for ref in result["missing_references"])
class TestBundleAndBaseUri:
"""Issue #825, Part B Tier 3 — prov:Bundle membership and configurable base_uri."""
def test_bundle_id_produces_bundle_triples(self):
prov_mgr = ProvenanceManager()
prov_mgr.track_entity("e1", source="doc1", bundle_id="ingestion_run_1")
ttl = prov_mgr.export_prov(format="turtle")
assert "Bundle" in ttl
assert "hadMember" in ttl
def test_default_base_uri_matches_rdf_exporter_namespace(self):
from semantica.provenance.manager import DEFAULT_BASE_URI
from semantica.export.rdf_exporter import NamespaceManager
assert NamespaceManager().namespaces["semantica"] == DEFAULT_BASE_URI
def test_export_prov_base_uri_override(self):
prov_mgr = ProvenanceManager()
prov_mgr.track_entity("e1", source="doc1")
default_ttl = prov_mgr.export_prov(format="turtle")
assert "https://semantica.dev/ns#" in default_ttl
custom_ttl = prov_mgr.export_prov(format="turtle", base_uri="https://example.org/kg#")
assert "https://example.org/kg#" in custom_ttl
assert "https://semantica.dev/ns#" not in custom_ttl
def test_owl_exporter_default_ontology_uri_matches_shared_namespace(self):
"""Issue #825 follow-up — OWLExporter was the one exporter left out of
the Part B Tier 3 namespace interlinking; its default ontology_uri
must match the same shared DEFAULT_BASE_URI as RDFExporter and
export_prov()."""
from semantica.provenance.manager import DEFAULT_BASE_URI
from semantica.export.owl_exporter import OWLExporter
assert OWLExporter().ontology_uri == DEFAULT_BASE_URI
class TestExplicitBitemporalFields:
"""Issue #825 follow-up — valid_from/valid_until as explicit,
caller-supplied ProvenanceEntry fields (matching the deprecated
kg.ProvenanceTracker's actual contract: these were always caller-supplied
metadata keys, never auto-computed)."""
def test_valid_from_valid_until_are_plain_passthrough_fields(self):
prov_mgr = ProvenanceManager()
entry = prov_mgr.track_entity(
"price1", source="doc1", valid_from="2026-01-01", valid_until="2026-06-01"
)
assert entry.valid_from == "2026-01-01"
assert entry.valid_until == "2026-06-01"
def test_revision_history_prefers_explicit_valid_from_until(self):
prov_mgr = ProvenanceManager()
prov_mgr.track_entity(
"price1", source="doc1", valid_from="2026-01-01", valid_until="2026-06-01"
)
history = prov_mgr.revision_history("price1")
assert history[0]["valid_from"] == "2026-01-01"
assert history[0]["valid_until"] == "2026-06-01"
def test_revision_history_falls_back_to_timestamp_when_unset(self):
"""Backward-compat: entries that don't set valid_from/valid_until
explicitly still get the dynamic timestamp-based derivation."""
prov_mgr = ProvenanceManager()
prov_mgr.track_entity("X", source="doc1")
prov_mgr.track_entity("X", source="doc1")
history = prov_mgr.revision_history("X")
assert history[0]["valid_until"] == history[1]["valid_from"]
assert history[1]["valid_until"] is None
def test_valid_from_until_survive_sqlite_round_trip(self, tmp_path):
db_path = str(tmp_path / "bitemporal.db")
prov_mgr = ProvenanceManager(storage_path=db_path)
prov_mgr.track_entity(
"price1", source="doc1", valid_from="2026-01-01", valid_until="2026-06-01"
)
prov_mgr2 = ProvenanceManager(storage_path=db_path)
entry = prov_mgr2.storage.retrieve("price1")
assert entry.valid_from == "2026-01-01"
assert entry.valid_until == "2026-06-01"
class TestActivityTimingAcrossWrappers:
"""Issue #825 follow-up — activity_started_at_time/ended_at_time wired
into all *_provenance.py wrapper modules that measure real work, not
just the 2 wrappers from the original Part B pass."""
def test_embedding_wrapper_records_activity_timing(self):
from semantica.embeddings.embeddings_provenance import EmbeddingGeneratorWithProvenance
wrapper = EmbeddingGeneratorWithProvenance(provenance=True, agent_id="embed_svc")
wrapper._generator.embed = lambda texts, **kw: [[0.1, 0.2] for _ in texts]
wrapper.embed(["hello", "world"])
entries = wrapper._prov_manager.storage.retrieve_all()
assert len(entries) == 1
assert entries[0].activity_started_at_time is not None
assert entries[0].activity_ended_at_time is not None
def test_kg_algorithm_tracker_accepts_caller_supplied_activity_timing(self):
from semantica.kg.kg_provenance import AlgorithmTrackerWithProvenance
tracker = AlgorithmTrackerWithProvenance(provenance=True, agent_id="algo_svc")
eid = tracker.track_embedding_computation(
graph=object(), algorithm="node2vec", embeddings={"n1": [0.1, 0.2]},
parameters={"d": 2},
activity_started_at_time="t0", activity_ended_at_time="t1",
)
entry = tracker._prov_manager.get_provenance(eid)
assert entry["activity_started_at_time"] == "t0"
assert entry["activity_ended_at_time"] == "t1"
def test_graph_builder_build_operation_has_no_end_time_yet(self):
"""The build-operation marker is recorded before the build runs, so
it legitimately has a start but no end time."""
from semantica.kg.kg_provenance import GraphBuilderWithProvenance
builder = GraphBuilderWithProvenance(provenance=True, agent_id="builder_svc")
builder._builder.build_single_source = lambda kg_data, **kw: {"entities": [], "relationships": []}
builder.build_single_source({"foo": "bar"})
entries = builder._prov_manager.storage.retrieve_all()
# NOTE: entity_type isn't asserted here — kg_provenance.py has a
# pre-existing, out-of-scope bug where entity_type is nested inside
# the metadata dict instead of passed as a track_entity kwarg, so it
# never actually populates the real field for this call site.
build_entries = [e for e in entries if e.entity_id.startswith("graph_build_single_")]
assert len(build_entries) == 1
assert build_entries[0].activity_started_at_time is not None
assert build_entries[0].activity_ended_at_time is None
def test_semantic_extract_wrapper_records_activity_timing(self):
from semantica.semantic_extract.semantic_extract_provenance import ProvenanceMixin
class FakeExtractor(ProvenanceMixin):
pass
wrapper = FakeExtractor(provenance=True, agent_id="extract_svc")
wrapper._track_extraction(
entity_id="e1", source="doc1", entity_type="named_entity",
activity_started_at_time="t0", activity_ended_at_time="t1",
)
entry = wrapper._prov_manager.get_provenance("e1")
assert entry["activity_started_at_time"] == "t0"
assert entry["activity_ended_at_time"] == "t1"
assert "activity_started_at_time" not in entry["metadata"]
def test_conflicts_wrapper_records_activity_timing(self):
from semantica.conflicts.conflicts_provenance import SourceTrackerWithUnifiedBackend
class FakeSource:
document = "doc1"
page = 1
section = None
confidence = 0.9
tracker = SourceTrackerWithUnifiedBackend(agent_id="conflicts_svc")
tracker.track_property_source("e1", "prop", "val", FakeSource())
entry = tracker._unified_manager.get_provenance("e1_prop")
assert entry["activity_started_at_time"] is not None
assert entry["activity_ended_at_time"] is not None
tests/provenance/test_schemas.py
+121 -1
View File
@@ -7,7 +7,14 @@ and PropertySource dataclasses.
import pytest
from datetime import datetime
from semantica.provenance.schemas import ProvenanceEntry, SourceReference, PropertySource
from semantica.provenance.schemas import (
ProvenanceEntry,
SourceReference,
PropertySource,
AgentRecord,
ActivityRecord,
Invalidation,
)
class TestProvenanceEntry:
@@ -162,3 +169,116 @@ class TestPropertySource:
assert isinstance(data, dict)
assert data["property_name"] == "name"
assert len(data["sources"]) == 1
class TestProvenanceEntryPart825Fields:
"""Issue #825, Part A — new additive fields on ProvenanceEntry."""
def test_defaults(self):
entry = ProvenanceEntry(entity_id="e1", entity_type="entity", activity_id="act")
assert entry.agent_type == "software_agent"
assert entry.is_automated is True
assert entry.role is None
assert entry.previous_version_id is None
assert entry.derived_from_id is None
assert entry.sequence_id is None
assert entry.previous_checksum is None
assert entry.invalidated is False
assert entry.invalidated_at_time is None
assert entry.invalidated_by is None
assert entry.invalidation_reason is None
# Part B additive fields
assert entry.activity_started_at_time is None
assert entry.activity_ended_at_time is None
assert entry.acted_on_behalf_of is None
assert entry.informed_by_activities == []
assert entry.valid_from is None
assert entry.valid_until is None
assert entry.revision_type is None
assert entry.supersedes is None
assert entry.bundle_id is None
def test_round_trip_to_dict_from_dict(self):
entry = ProvenanceEntry(
entity_id="e1",
entity_type="entity",
activity_id="act",
agent_id="alice",
agent_type="person",
is_automated=False,
role="approver",
previous_version_id="e1:v:1",
derived_from_id="source_entity",
sequence_id=5,
previous_checksum="abc123",
invalidated=True,
invalidated_at_time="2026-01-01T00:00:00",
invalidated_by="reviewer_jane",
invalidation_reason="retracted",
activity_started_at_time="2026-01-01T00:00:00",
activity_ended_at_time="2026-01-01T00:00:05",
acted_on_behalf_of="org1",
informed_by_activities=["ingest_activity"],
valid_from="2026-01-01",
valid_until="2026-06-01",
revision_type="correction",
supersedes="old_claim",
bundle_id="ingestion_run_1",
)
data = entry.to_dict()
restored = ProvenanceEntry.from_dict(data)
assert restored == entry
class TestActivityRecord:
"""Issue #825, Part B Tier 1 — minimum-viable typed activity."""
def test_defaults(self):
activity = ActivityRecord(id="act1")
assert activity.activity_type == "process"
assert activity.started_at_time is None
assert activity.ended_at_time is None
def test_round_trip(self):
activity = ActivityRecord(
id="bureau_parsing_run_42",
activity_type="extraction",
started_at_time="2026-01-01T00:00:00",
ended_at_time="2026-01-01T00:00:03",
)
data = activity.to_dict()
restored = ActivityRecord.from_dict(data)
assert restored == activity
class TestAgentRecord:
"""Issue #825, Part A item 3 — minimum-viable typed agent."""
def test_defaults(self):
agent = AgentRecord(id="bot1")
assert agent.agent_type == "software_agent"
assert agent.is_automated is True
assert agent.name is None
def test_round_trip(self):
agent = AgentRecord(id="reviewer_jane", agent_type="person", is_automated=False, name="Jane")
data = agent.to_dict()
restored = AgentRecord.from_dict(data)
assert restored == agent
class TestInvalidation:
"""Issue #825, Part A item 1 — tombstone record."""
def test_round_trip(self):
inv = Invalidation(
entity_id="e1",
invalidated_at_time="2026-01-01T00:00:00",
invalidated_by="reviewer_jane",
reason="retracted",
)
data = inv.to_dict()
restored = Invalidation.from_dict(data)
assert restored == inv
tests/provenance/test_storage.py
+169
View File
@@ -134,10 +134,116 @@ class TestInMemoryStorage:
assert count == 1
assert len(storage.retrieve_all()) == 0
def test_get_chain_head(self):
"""Issue #825, Part A item 2 — chain head reporting."""
storage = InMemoryStorage()
assert storage.get_chain_head() is None
entry1 = ProvenanceEntry(
entity_id="e1", entity_type="entity", activity_id="act",
sequence_id=1, checksum="checksum_1",
)
storage.store(entry1)
assert storage.get_chain_head() == (1, "checksum_1")
entry2 = ProvenanceEntry(
entity_id="e2", entity_type="entity", activity_id="act",
sequence_id=2, checksum="checksum_2",
)
storage.store(entry2)
assert storage.get_chain_head() == (2, "checksum_2")
def test_trace_descendants(self):
"""Issue #825, Part A item 5 — reverse (downstream) lineage traversal."""
storage = InMemoryStorage()
storage.store(ProvenanceEntry(entity_id="a", entity_type="entity", activity_id="act"))
storage.store(ProvenanceEntry(
entity_id="b", entity_type="entity", activity_id="act", parent_entity_id="a"
))
storage.store(ProvenanceEntry(
entity_id="c", entity_type="entity", activity_id="act", parent_entity_id="b"
))
storage.store(ProvenanceEntry(
entity_id="d", entity_type="entity", activity_id="act", used_entities=["a"]
))
descendants = storage.trace_descendants("a")
entity_ids = {e.entity_id for e in descendants}
assert entity_ids == {"b", "c", "d"}
def test_trace_descendants_respects_max_depth(self):
storage = InMemoryStorage()
storage.store(ProvenanceEntry(entity_id="a", entity_type="entity", activity_id="act"))
storage.store(ProvenanceEntry(
entity_id="b", entity_type="entity", activity_id="act", parent_entity_id="a"
))
storage.store(ProvenanceEntry(
entity_id="c", entity_type="entity", activity_id="act", parent_entity_id="b"
))
descendants = storage.trace_descendants("a", max_depth=1)
entity_ids = {e.entity_id for e in descendants}
assert entity_ids == {"b"}
def test_trace_descendants_empty_for_leaf(self):
storage = InMemoryStorage()
storage.store(ProvenanceEntry(entity_id="leaf", entity_type="entity", activity_id="act"))
assert storage.trace_descendants("leaf") == []
class TestSQLiteStorage:
"""Test SQLiteStorage backend."""
def test_all_fields_round_trip_through_sqlite(self):
"""Regression test: every ProvenanceEntry field, including issue #825
Part A and Part B additions, must survive a SQLite store/retrieve
round trip byte-for-byte. Part B's activity/actedOnBehalfOf/
informedBy/revision/bundle fields were initially added to the
dataclass and to export_prov() without updating SQLiteStorage's DDL/
INSERT/_row_to_entry — InMemoryStorage stores the dataclass directly
so it masked the gap, but SQLite silently dropped every one of those
fields on write."""
with tempfile.NamedTemporaryFile(delete=False, suffix=".db") as tmp:
db_path = tmp.name
try:
storage = SQLiteStorage(db_path)
entry = ProvenanceEntry(
entity_id="entity_1",
entity_type="entity",
activity_id="extraction",
agent_id="alice",
agent_type="person",
is_automated=False,
role="approver",
sequence_id=1,
previous_checksum="prevchk",
parent_entity_id="parent_1",
used_entities=["u1", "u2"],
previous_version_id="entity_1:v:1",
derived_from_id="source_1",
invalidated=True,
invalidated_at_time="2026-01-01T00:00:00",
invalidated_by="reviewer_jane",
invalidation_reason="retracted",
activity_started_at_time="2026-01-01T00:00:00",
activity_ended_at_time="2026-01-01T00:00:05",
acted_on_behalf_of="org1",
informed_by_activities=["act_a", "act_b"],
valid_from="2026-01-01",
valid_until="2026-06-01",
revision_type="correction",
supersedes="old_claim",
bundle_id="ingestion_run_1",
)
storage.store(entry)
retrieved = storage.retrieve("entity_1")
assert retrieved == entry
finally:
if os.path.exists(db_path):
os.unlink(db_path)
def test_store_and_retrieve(self):
"""Test storing and retrieving entries."""
with tempfile.NamedTemporaryFile(delete=False, suffix=".db") as tmp:
@@ -216,3 +322,66 @@ class TestSQLiteStorage:
finally:
if os.path.exists(db_path):
os.unlink(db_path)
def test_get_chain_head(self):
"""Issue #825, Part A item 2 — chain head reporting."""
with tempfile.NamedTemporaryFile(delete=False, suffix=".db") as tmp:
db_path = tmp.name
try:
storage = SQLiteStorage(db_path)
assert storage.get_chain_head() is None
entry1 = ProvenanceEntry(
entity_id="e1", entity_type="entity", activity_id="act",
sequence_id=1, checksum="checksum_1",
)
storage.store(entry1)
assert storage.get_chain_head() == (1, "checksum_1")
entry2 = ProvenanceEntry(
entity_id="e2", entity_type="entity", activity_id="act",
sequence_id=2, checksum="checksum_2",
)
storage.store(entry2)
assert storage.get_chain_head() == (2, "checksum_2")
finally:
if os.path.exists(db_path):
os.unlink(db_path)
def test_trace_descendants(self):
"""Issue #825, Part A item 5 — reverse (downstream) lineage traversal."""
with tempfile.NamedTemporaryFile(delete=False, suffix=".db") as tmp:
db_path = tmp.name
try:
storage = SQLiteStorage(db_path)
storage.store(ProvenanceEntry(entity_id="a", entity_type="entity", activity_id="act"))
storage.store(ProvenanceEntry(
entity_id="b", entity_type="entity", activity_id="act", parent_entity_id="a"
))
storage.store(ProvenanceEntry(
entity_id="c", entity_type="entity", activity_id="act", parent_entity_id="b"
))
storage.store(ProvenanceEntry(
entity_id="d", entity_type="entity", activity_id="act", used_entities=["a"]
))
descendants = storage.trace_descendants("a")
entity_ids = {e.entity_id for e in descendants}
assert entity_ids == {"b", "c", "d"}
finally:
if os.path.exists(db_path):
os.unlink(db_path)
def test_trace_descendants_empty_for_leaf(self):
with tempfile.NamedTemporaryFile(delete=False, suffix=".db") as tmp:
db_path = tmp.name
try:
storage = SQLiteStorage(db_path)
storage.store(ProvenanceEntry(entity_id="leaf", entity_type="entity", activity_id="act"))
assert storage.trace_descendants("leaf") == []
finally:
if os.path.exists(db_path):
os.unlink(db_path)