README.md

LLM Toolchain based on LangChain

A comprehensive toolchain for Large Language Models (LLMs) built on LangChain, providing a flexible framework for document processing, retrieval-augmented generation (RAG), and model evaluation.

Key Capabilities

• LangChain Integration: Leverages LangChain's powerful components for building sophisticated LLM applications
• RAG (Retrieval Augmented Generation): Multiple implementations for enhanced context-aware responses
• Fine Tuning Models: Support for custom model fine-tuning and specialized chain creation
• Vector Database Support Pinecone and other intergations
• Experiments: Integrated experiment tracking and evaluation using LangSmith
• Evaluations: Comprehensive evaluation framework for assessing model performance

TODO NEXT

• CI/DI: Dockerize
• CI/DI: Deploy to AWS/Azure templates
• Chains: Chat history = Chains: from langchain_core.documents import Document https://python.langchain.com/docs/tutorials/retrievers/
• Chains: over SQL data
• Agents https://python.langchain.com/docs/tutorials/agents/
• RAG: similarity threshold for fallback answer
• UI: Graphical Interface for documents uploading
• 2 more Vector DB supports
• 2 more Document loaders
• CI/DI: HELM + k8 deployments

Components

• rag_gpt_model_pdf_chain.py: PDF document processing chain
• rag_gpt_model_txt_from_dir_chain.py: Text file processing chain
• rag_gpt_model_pdf_pinecone_chain.py: Pinecone-based document processing
• test_chains.py: Test utilities for different chain implementations
• run_experiment.py: Experiment runner with LangSmith integration
• fine_tuned_chain.py: Custom chain implementation
• get_datasets.py: Dataset management utilities

Useful docs

• Dashboard – https://smith.langchain.com/
• Models – https://python.langchain.com/docs/integrations/chat/
• Document loaders – https://python.langchain.com/docs/integrations/document_loaders/
• Vector Stores - https://python.langchain.com/docs/integrations/vectorstores/
• Retrievers – https://python.langchain.com/docs/integrations/retrievers/
• Tutorials - https://python.langchain.com/docs/tutorials/

Usage

• cd LLM-toolchain
• python3 -m venv venv
• source myenv/bin/activate
• pip install -r requirements.txt

Run any script with python ...