aice-knowledge

Python SDK for the Knowledge Service — multi-tenant document ingestion (PDF/DOCX/OCR) into Postgres + pgvector, served over REST and MCP.

from aice_knowledge import KnowledgeClient

client = KnowledgeClient(base_url="https://kb.example.com", api_key="sk-…")

# Upload a document (queued for parsing + embedding)
doc = client.documents.upload(
    file=open("contract.pdf", "rb").read(),
    filename="contract.pdf",
)

# Vector search
hits = client.search(query="payment terms", limit=5)

# Manage the scope tree (multi-tenant document organisation)
client.scopes.create(name="finance", parent_id=None)
client.document_scopes.link(document_id=doc["id"], scope_ids=["sc_finance"])

Endpoints covered (see knowledge_service/app/api/routes.py):

• POST /documents, POST /documents/batch, GET /documents, GET /documents/{id}, POST /documents/{id}/retry, DELETE /documents/{id}
• POST /documents/{id}/share, GET /documents/{id}/share, DELETE /documents/{id}/share/…
• POST /search
• POST /scope-types, GET /scope-types, DELETE /scope-types/{id}
• POST /scopes, GET /scopes, DELETE /scopes
• POST /document-scopes, GET /document-scopes, DELETE /document-scopes