Upload documents
Add files to your knowledge base. Supported: PDF, TXT, Markdown, HTML, and images.
What happens after upload
- 1
Images described by Gemini Vision
Other files are read directly
- 2
Split into overlapping chunks
~512 tokens each, with context overlap
- 3
Embedded with all-MiniLM-L6-v2
Runs locally — no API key needed
- 4
Stored in Qdrant
Searchable via BM25 + vector retrieval