Store

SQLite-backed document storage with cosine similarity search and tenant isolation.

Constructor

from distillcore.storage import Store
 
store = Store(path="~/.distillcore/store.db")

Methods

save(result, tenant_id=None) → str

Save a ProcessingResult. Returns the document ID.

doc_id = store.save(result)
doc_id = store.save(result, tenant_id="org-123")

get_document(document_id, tenant_id=None) → dict | None

Retrieve document metadata by ID.

doc = store.get_document("abc-123")

list_documents(document_type=None, limit=50, tenant_id=None) → list[dict]

List documents with optional filtering.

docs = store.list_documents(document_type="motion", limit=10)

get_chunks(document_id, tenant_id=None) → list[dict]

Get all chunks for a document, ordered by chunk_index.

chunks = store.get_chunks("abc-123")

search(query_embedding, top_k=10, document_type=None, document_id=None, tenant_id=None) → list[dict]

Cosine similarity search over chunk embeddings. Returns results with score field.

results = store.search(embedding, top_k=5)
for r in results:
    print(f"{r['score']:.3f}: {r['text'][:80]}")

delete_document(document_id, tenant_id=None) → bool

Delete a document and all its chunks. Returns True if found.

store.delete_document("abc-123")

stats() → dict

Aggregate statistics.

stats = store.stats()
# {"documents": 42, "chunks": 1200, "chunks_with_embeddings": 1200,
#  "searches": 15, "document_types": {"report": 30, "motion": 12}}

close()

Close the database connection.

store.close()