Overview - Ragrails

The REST API has workflow endpoints for common paths and stage endpoints for explicit control.

extract -> chunk -> embed -> store -> retrieve -> chat

Use workflows when you want fewer requests. Use stage endpoints when you want to inspect intermediate JSON, retry one step, or move artifacts through another system.

Workflow Endpoints

Endpoint	Runs
`POST /v1/pipelines/ingest`	extraction -> chunking -> embedding -> storing from JSON sources
`POST /v1/pipelines/ingest/mixed`	multipart upload manifest -> extraction -> chunking -> embedding -> storing
`POST /v1/pipelines/query`	query embedding -> retrieval
`POST /v1/chat`	query embedding -> retrieval -> LLM answer

Stage Endpoints

Stage	Endpoint	Output
Extraction	`/v1/ingest/url`, `/v1/ingest/docs`, `/v1/ingest/docs/upload`, `/v1/ingest/api`	Normalized documents in `outputs`
Chunking	`/v1/chunk`	Chunk objects in `items`
Embedding	`/v1/embed`	Embedded chunks in `items`
Storing	`/v1/store`	Store summary and stored item IDs
Retrieval	`/v1/retrieve`	Retrieved chunks in `items`
Chat	`/v1/chat`	Answer, sources, updated history, and quality metadata

Stage Flow

curl -X POST http://127.0.0.1:8000/v1/ingest/docs \
  -H "Content-Type: application/json" \
  -d '{"files":["files/refund-policy.pdf"]}'

curl -X POST http://127.0.0.1:8000/v1/chunk \
  -H "Content-Type: application/json" \
  -d '{"markdown":[{"id":"refund-policy","text":"..."}]}'

curl -X POST http://127.0.0.1:8000/v1/embed \
  -H "Content-Type: application/json" \
  -d '{"chunks":[{"id":"refund-policy-0","text":"..."}],"provider":"voyage","model":"voyage-3"}'

curl -X POST http://127.0.0.1:8000/v1/store \
  -H "Content-Type: application/json" \
  -d '{"embedded_chunks":[{"id":"refund-policy-0","text":"...","embedding":[0.1,0.2]}],"vector_db":"qdrant","collection":"support","url":"http://localhost:6333"}'

Output Handoff

From	Use
Extraction `outputs`	`markdown` for `/v1/chunk`
Chunking `items`	`chunks` for `/v1/embed`
Embedding `items`	`embedded_chunks` for `/v1/store`
Retrieval `items`	Context for custom answer generation, or use `/v1/chat`