Documents API

Route	Method	Purpose
/api/doc	GET	Signed document URL
/api/doc-upload-session	POST	Allocate a document row and signed upload token
/api/ingest-doc	POST	Enqueue extraction/embedding for an uploaded document
/api/document-inventory	GET	Admin/lead inventory of indexed documents

Upload flow

1. POST /api/doc-upload-session allocates the documents catalog row and a signed upload token — the browser uploads directly to storage using that token, not through this API.
2. POST /api/ingest-doc { document_id } then enqueues indexing.

{ "document_id": "document-uuid" }

This route is intentionally durable-only: extraction, embedding, and atomic publication all run in Trigger.dev (sable.document.backfill-index-batch), not inline in the request handler — the request only validates access and enqueues the work. Rate-limited via SABLE_INGEST_RATE_LIMIT (default 12/min).

GET /api/doc

Returns a signed URL for one document. Enforced by the same access rules across every document surface:

• A valid Supabase session is required.
• Restricted documents (contracts, SOWs) require admin/lead.
• Non-leads must be a member of the document's project.

GET /api/document-inventory

Admin/lead only. Every indexed document plus contract/SOW extraction health flags — buildDocumentInventory (in _deliverable_authoring.js) is shared with the deliverable-drafting pipeline's evidence-collection step, so this inventory and what a deliverable draft can actually cite are always the same list.

Where the code lives

• sable-agents-demo/web/api/doc.js, doc-upload-session.js, ingest-doc.js, document-inventory.js
• sable-agents-demo/web/api/_doc_upload.js, _document_ingest.js, _ingestion_pipeline.js, _deliverable_authoring.js
• sable-agents-demo/trigger/documents.ts
• sable-agents-demo/docs-site/docs/reference/search-indexing.md — the retrieval side of this pipeline