Extract-time provenance (#661)

1. Shared Provenance Module - URI generators, namespace constants,
   triple builders, vocabulary bootstrap
2. Librarian - Emits document metadata to graph on processing
   initiation (vocabulary bootstrap + PROV-O triples)
3. PDF Extractor - Saves pages as child documents, emits parent-child
   provenance edges, forwards page IDs
4. Chunker - Saves chunks as child documents, emits provenance edges,
   forwards chunk ID + content
5. Knowledge Extractors (both definitions and relationships):
   - Link entities to chunks via SUBJECT_OF (not top-level document)
   - Removed duplicate metadata emission (now handled by librarian)
   - Get chunk_doc_id and chunk_uri from incoming Chunk message
6. Embedding Provenance:
   - EntityContext schema has chunk_id field
   - EntityEmbeddings schema has chunk_id field
   - Definitions extractor sets chunk_id when creating EntityContext
   - Graph embeddings processor passes chunk_id through to
     EntityEmbeddings

Provenance Flow:
Document → Page (PDF) → Chunk → Extracted Facts/Embeddings
    ↓           ↓          ↓              ↓
  librarian  librarian  librarian    (chunk_id reference)
  + graph    + graph    + graph

Each artifact is stored in librarian with parent-child linking, and PROV-O
edges are emitted to the knowledge graph for full traceability from any
extracted fact back to its source document.

Also, updating tests

trustgraph-flow/trustgraph/extract/kg/definitions/extract.py

@@ -128,10 +128,12 @@ class Processor(FlowProcessor):
             triples = []
             entities = []
 
-            # FIXME: Putting metadata into triples store is duplicated in
-            # relationships extractor too
-            for t in v.metadata.metadata:
-                triples.append(t)
+            # Get chunk document ID for provenance linking
+            chunk_doc_id = v.document_id if v.document_id else v.metadata.id
+            chunk_uri = v.metadata.id  # The URI form for the chunk
+
+            # Note: Document metadata is now emitted once by librarian at processing
+            # initiation, so we don't need to duplicate it here.
 
             for defn in defs:
 
@@ -159,22 +161,27 @@ class Processor(FlowProcessor):
                     s=s_value, p=DEFINITION_VALUE, o=o_value
                 ))
 
+                # Link entity to chunk (not top-level document)
                 triples.append(Triple(
                     s=s_value,
                     p=SUBJECT_OF_VALUE,
-                    o=Term(type=IRI, iri=v.metadata.id)
+                    o=Term(type=IRI, iri=chunk_uri)
                 ))
 
                 # Output entity name as context for direct name matching
+                # Include chunk_id for embedding provenance
                 entities.append(EntityContext(
                     entity=s_value,
                     context=s,
+                    chunk_id=chunk_doc_id,
                 ))
 
                 # Output definition as context for semantic matching
+                # Include chunk_id for embedding provenance
                 entities.append(EntityContext(
                     entity=s_value,
                     context=defn["definition"],
+                    chunk_id=chunk_doc_id,
                 ))
 
             # Send triples in batches

trustgraph-flow/trustgraph/extract/kg/relationships/extract.py

@@ -109,10 +109,12 @@ class Processor(FlowProcessor):
 
             triples = []
 
-            # FIXME: Putting metadata into triples store is duplicated in
-            # relationships extractor too
-            for t in v.metadata.metadata:
-                triples.append(t)
+            # Get chunk document ID for provenance linking
+            chunk_doc_id = v.document_id if v.document_id else v.metadata.id
+            chunk_uri = v.metadata.id  # The URI form for the chunk
+
+            # Note: Document metadata is now emitted once by librarian at processing
+            # initiation, so we don't need to duplicate it here.
 
             for rel in rels:
 
@@ -168,19 +170,19 @@ class Processor(FlowProcessor):
                         o=Term(type=LITERAL, value=str(o))
                     ))
 
-                # 'Subject of' for s
+                # Link entity to chunk (not top-level document)
                 triples.append(Triple(
                     s=s_value,
                     p=SUBJECT_OF_VALUE,
-                    o=Term(type=IRI, iri=v.metadata.id)
+                    o=Term(type=IRI, iri=chunk_uri)
                 ))
 
                 if rel["object-entity"]:
-                    # 'Subject of' for o
+                    # Link object entity to chunk
                     triples.append(Triple(
                         s=o_value,
                         p=SUBJECT_OF_VALUE,
-                        o=Term(type=IRI, iri=v.metadata.id)
+                        o=Term(type=IRI, iri=chunk_uri)
                     ))
 
             # Send triples in batches