Langchain Introduction
Langchain is a framework for developing applications with Large Language Models.
Why Langchain?
It provides a standard interface for developing applications with any LLM. Langchain does this by using its modules.
Modules
- Model I/O
- Retreival
- Chains
- Memory
- Agents
Model I/O
- Basic LLM Inputs and Outputs.
- More flexible than the existing model specific api formats.
- Learn the input and output formats for two type of models:
- Text Completion models
- Chat models
Retreival
- Helps in connecting the LLM to a data source (like a vector database).
- Works in a standardized structure (that means, LLMs and Vector Databases can easily be swapped).
- Process: VectorDB > Query ways > Results to LLM
Chains
- For linking output of one model to another model.
Memory
- For retaining the historical memory of our previous interactions in our own defined way.
Agents
- Use the LLMs to choose the sequence of actions to take to perform a task using the tools we provide.