#how to use custom knowledge base with voice ai agent?

1 messages · Page 1 of 1 (latest)

obtuse umbraBOT
#
AI-generated Answer for waterfallbay

Here’s how to use a custom knowledge base with an ElevenLabs Voice AI agent, based only on the official docs.

Dashboard (no code)

  • Create or open an Agent, then go to the Knowledge Base section and add items as Files, URLs, or Text. Supported file types include PDF, TXT, DOCX, HTML, and EPUB.
  • Optionally enable RAG in the Knowledge Base section to let the agent retrieve only the most relevant chunks from large documents; indexing starts automatically and may take a few minutes. RAG typically adds ~500 ms latency. You can set each document’s usage mode to Auto (retrieve when relevant) or Prompt (always in system prompt; use sparingly).
  • You can reuse existing knowledge base documents across multiple agents from the Agent tab’s “Add document” flow; documents can’t be deleted while any agent depends on them.

API/SDK workflow

  • Create knowledge base documents by uploading a file or providing a URL via POST /v1/convai/knowledge-base, then attach them to your agent by calling Update Agent.
  • For RAG via API/SDK: index the document for retrieval and enable RAG in the agent configuration; you can also adjust embedding model and retrieval limits.

Limits and behaviors you should know

  • Non‑enterprise: limit of 5 files/links and up to 20 MB total or 300,000 characters. With RAG enabled, the maximum size per knowledge base item increases from 300 KB to 10 MB (overall limits still depend on plan).
  • The Knowledge Base page reiterates non‑enterprise maximums of 20 MB or 300k characters.
  • RAG enforces workspace-wide index size caps by tier (e.g., Creator 20 MB, Pro 100 MB, Scale 500 MB, Business 1 GB).
  • URL items: full-site crawling and continuous auto-updating are not currently supported; you must add/update specific URLs as needed.

Prompting tip

  • In your system prompt, explicitly tell the agent what sources it can use (your knowledge base, tools, etc.) and when to consult them; this improves grounding and reduces hallucinations.

If you want, tell me whether you’ll manage the KB via dashboard or API, and I can point you to the exact endpoints or UI path for your setup.

pure ledge
#

I mean can i use my own proprietary RAG service and link it to 11 labs AI Agents?