tillg/langchain_playground

langchain_playground

Playing around with langchain.

Dev Setup & Running

I use a couple of windows with a standard command (this all assumes you are in the project folder):

# Follow the application log 
less +F data/logs/app.log

# Follow the 
less +F ~/.ollama/logs/server.log

To run a script I use the module syntax: python -m rag.simple_chat.

History

2024-04-20 Configurable vectorestores

• Vectorestores are now in config file
• Config of a configuration contains also the model and the directory with the data to ingest
• Use a

2024-04-20 Debugging LangChain chains

• Direct logging into a file so it doesn't interfere with my dialog in the terminal.

2024-04-06 Doing RAG tutorial

• Folllowing this tutorial from langchain

2024-04-08 Ingesting

• Trying to ingest the A12 documentation, ~ 3'800 markdown docs.
• Trying to ingest with llama2:latest. Takes long:
  • 3456/19666 [30:02<2:23:58, 1.88it/s]
  • Expected time > 3 hours
• With default embedding model ()
  • 1224/19666 [10:28<2:46:50, 1.84it/s]
  • Expected time > 3 hours
• With embedding model nomic-embed-text
  • 15 minutes! 🥰

Tech reading

• Which model to use for local embeddings with Ollama? In this blog entry (April 8, 2024) they give an overview.
• For having iTerm layouts with commands so I get back my screen setup: iTomate
• For a later version: Building a Confluence Q&A App with LangChain and ChatGPT. Hint: There is a ConfluenceLoader in LangChain!!
• While looking for a fast embedding model: Local Embedding done right - Medium