Labbing Vector Search, Langchain and local LLMs

Repo:

• Blamechance/quickstart-AI-labs

Week 1: LangChain Basics

• Complete LangChain Build a semantic search engine tutorial.
  • Link.
  • Use Voyage AI to create embeddings.
  • Store vectors locally/in-memory for now.
  • Related Note: Getting-started-with-langchain

Resources:

• Integrate MongoDB with LangChain
• VoyageAI Quick-Start Guide.
• Markdown loader:
  • UnstructuredMarkdownLoader | 🦜️🔗 LangChain

Week 2: Integrating LangChain with MongoDB Atlas

• Complete tutorial for integrating MongoDB with Langchain.
  • MongoDB Vector Search Quick Start - Atlas - MongoDB Docs
  • Integrate MongoDB with LangChain.

Week 3: Re-write app using VoyageAI Python SDK.

To learn the nuances between leveraging langchain abstractions/components, vs. one without a framework dependency.

• VoyageAI | API Key and Python Client
• Chose to not manually dosdk<>mongodb path, since it would take more time than desired to learn to use mongoclient object, uploading embeddings to Atlas etc.
  • This can be done when building an actual implementation.

Week 4: Develop a tool that solves a problem using these foundations.