How to query the Eleven labs Knowledge base through API | ElevenLabs | Page 1

AI-generated Answer for kishore_vb

Short answer: you don’t “query the Knowledge Base” directly. You attach KB documents to an Agent and send the Agent a message (text or voice). The Agent retrieves relevant snippets from the KB (optionally via RAG) and returns an answer.

Implement it via API:

Add documents to the Knowledge Base

File/URL upload: POST /v1/convai/knowledge-base (multipart form; name, file, or url). After creation, you’ll update the Agent to use this doc. (elevenlabs.io)
Text document: POST /v1/convai/knowledge-base/text (JSON: { text, name? }). (elevenlabs.io)
List/inspect docs: GET /v1/convai/knowledge-base and GET /v1/convai/knowledge-base/{documentation_id}. (elevenlabs.io)

(Optional but recommended for large KB) Enable RAG indexing on the doc

Start or check indexing: POST /v1/convai/knowledge-base/{documentation_id}/rag-index (model: e5_mistral_7b_instruct or multilingual_e5_large_instruct). (elevenlabs.io)
View a doc’s indexes: GET /v1/convai/knowledge-base/{documentation_id}/rag-index; workspace usage overview: GET /v1/convai/knowledge-base/rag-index. (elevenlabs.io)
RAG behavior, limits, and API implementation guidance (including enabling RAG per agent) are documented here. (elevenlabs.io)

Attach documents and (optionally) RAG to your Agent

Update the Agent configuration to include the KB document(s) and enable RAG in the prompt/knowledge base settings: PATCH /v1/convai/agents/{agent_id}. (elevenlabs.io)

Query the KB by talking to the Agent (Chat Mode or Voice)

Establish a WebSocket conversation: wss://api.elevenlabs.io/v1/convai/conversation?agent_id=... (use a server-signed URL for private agents). Send a “user_message” with your question; receive “agent_response”. (elevenlabs.io)
For text-only Chat Mode (no audio), use the SDK or the same WebSocket but send text messages (conversation.send_user_message in SDKs). (elevenlabs.io)

Notes and limits

The Knowledge Base feature equips Agents; there is no separate “KB question/answer” REST endpoint in the public docs. Use an Agent conversation to get grounded answers. Overview and limits are described here. (elevenlabs.io)
After uploading, remember to update the Agent to actually use the doc; otherwise it won’t be consulted. (elevenlabs.io)

Relevant docs:

KB overview and limits. (elevenlabs.io)
Add file/URL, add text, list/get docs. (elevenlabs.io)
RAG indexing endpoints and guide. (elevenlabs.io)

#How to query the Eleven labs Knowledge base through API