TanGentleman / Augmenta

Automate RAG-powered workflows
MIT License
1 stars 0 forks source link

Bug: MultiVectorRetriever -> inputs not correctly set #39

Closed TanGentleman closed 4 weeks ago

TanGentleman commented 1 month ago
(.venv) tan@Tan Augmenta % python chat.py -m arctic -c expert_multivector_collection
Rag config key collection_name overridden
Rag config key rag_mode overridden
Rag config key rag_llm overridden
Rag mode set to True
Enabling RAG mode. Doing some validation steps
Warning: Chunk size * k exceeds model limit.
Vector DB exists, not validating inputs!
Config INFO: System message set to You are a helpful AI.
Collection found in vector DB
WARNING:classes:Switching to multivector enabled
Config initialized and set in active.json.
Loading existing Vector DB: expert_multivector_collection
Indexing url: https://developer.spotify.com/documentation/web-api/reference/get-playlist
^CKeyboard interrupt

The inputs in the vector-db for expert_multivector_collection should be read from manifest.json. The spotify link (my active input in settings.json) should be overridden.