Skip to main content

Ingest a single item

POST 
/ingest/items

Submit a transcript payload for asynchronous ingestion into the Fora pipeline.

The endpoint accepts transcript text along with metadata and enqueues a background job to:

  1. Persist a document record (with metadata, but NOT the full transcript)
  2. Generate embeddings if configured
  3. Extract tasks from the transcript content
  4. Record an ingestion event for observability

Important notes:

  • The transcript text is ephemeral and will NOT be stored after processing
  • Only a small snippet is kept for debugging purposes
  • Duplicate submissions (same source_type + external_id) will update existing records
  • The endpoint returns immediately with a job ID; processing is asynchronous

Guardrails:

  • Maximum request body: 1 MiB
  • Maximum transcript: 200,000 characters
  • Maximum metadata: 64 KiB per field
  • Rate limit: 120 requests/minute per token, burst up to 300

Responses​

Successful Response