Assistant Knowledge (/app/assistants/{id}/knowledge) ingests files and websites into vector chunks. Embeddings use a local sentence-transformers model. Knowledge nodes on the flow canvas and runtime fallback path query these chunks.
Sources
- Upload PDF, DOCX, TXT, or website URLs.
- Per-source status — indexing, ready, failed with error message.
- Chunk count and embedding model stored per source.
- Intent bindings — attach sources to specific intents for scoped retrieval.
- Website crawl filtering — include/exclude URL patterns to control what gets indexed.
- Test retrieval — query box with combined vector + keyword scores.
Ingestion pipeline
- Parse document or crawl URL.
- Chunk with configurable overlap.
- Embed and store with keyword indexing.
- Synthesis optional — LLM-powered answer synthesis from retrieved chunks.
AI Agent Knowledge (/app/agents/{slug}/knowledge) is a separate RAG store for the agentic runtime — same embedding approach, different API namespace.