Knowledge Base (RAG)
Tool set: ragDefault: Disabled
Upload documents and let your agent search them using vector similarity. Agents answer questions grounded in your actual content.
How It Works
Upload → Extract text → Chunk (2000 chars) → Embed via Gemini → Store vectors → Search by cosine similarity. Returns top 8 matching chunks with relevance scores.
Tools
| Tool | Description |
|---|---|
| search_documents | Vector search across uploaded documents |
Supported Files
PDF, Word (.docx), Plain text (.txt), Markdown (.md)
Limits
| Plan | Documents per account | Knowledge storage |
|---|---|---|
| Free | 10 | 100 MB |
| Pro | 100 | 1 GB |
| Max | 1,000 | 10 GB |
Each uploaded document can be up to 10 MiB.