The assistant answers questions from this knowledge base. Documents are split into chunks, each embedded into a 768-dimensional vector, and retrieved by cosine similarity. Add your own knowledge below by typing, uploading a Markdown file, or speaking it.
Sources
0docs
Chunks
0embedded
Added by you
0
Dimensions
768per vector
Add knowledge
Type it, upload a Markdown file, or speak it
Embedding map
Every chunk projected to 2D by meaning
Each dot is a chunk, placed by meaning (PCA to 2D). Lines link the nearest neighbours. Hover to inspect.
Test retrieval
See which chunks answer a question, with similarity scores
Embedded chunks
0 total