RAG

RAG lets an agent answer from documents you provide. Indexing is delegated to the RAG service (th2rag), reachable at RAG_BASE_URL.

Upload documents → indexed by th2rag → the agent searches them when answering

Sources

Source	Endpoint	Notes
Files	`POST /api/rag/index-files`	PDF, CSV, TXT, DOCX… max 50 MB per file
URL	`POST /api/rag/index-url`	Crawls and indexes a web page
Database, SQL	`POST /api/rag/index-db`	Indexes the result of a SQL query
Database, natural language	`POST /api/rag/index-db-nl`	Same, from a question
S3	`POST /api/rag/index-s3`	Indexes files from a bucket

Poll GET /api/rag/status/{knowledge_id}
Or stream GET /api/rag/stream/{agent_id}?session_id=xxx — events connected, status, complete

GET /api/rag/knowledge/{agent_id} lists what an agent currently has indexed.

Path traversal is blocked on session and agent identifiers
URL indexing refuses localhost and private address ranges (SSRF protection)
Session ownership is validated on every call
Completion callbacks from th2rag are verified with an HMAC-SHA256 signature (RAG_WEBHOOK_SECRET)