Skip to content

CLAUDE.md

Latest commit

 

History

History
111 lines (83 loc) · 3.97 KB

CLAUDE.md

File metadata and controls

111 lines (83 loc) · 3.97 KB

CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

Repository Overview

This is a curated collection of LLM applications, tutorials, and resources focused on large language models (LLMs), particularly ChatGPT, Claude, DeepSeek, and Llama series. The repository is primarily a documentation and tutorial collection rather than a single unified codebase.

Repository Structure

The repository is organized into distinct categories under the src/ directory:

  • src/agents/ - AI agent implementations using frameworks like CrewAI, AutoGen, and multi-agent systems
  • src/chatbots/ - Custom chatbot implementations (chat with PDF, GitHub, Gmail, YouTube, etc.)
  • src/frameworks/ - Advanced framework examples (Gemini multimodal, LLM routers, web scrapers)
  • src/rag/ - Retrieval-Augmented Generation (RAG) systems with various vector databases
  • src/finetuning/ - Fine-tuning guides for models like Llama 3.2
  • src/memory/ - LLM applications with long-term memory and context management

Each subdirectory is a standalone project with its own:

  • README.md - Setup instructions and feature descriptions
  • requirements.txt - Python dependencies
  • Python scripts (typically named descriptively, e.g., customer_support_agent.py)

Working with Individual Projects

Running a Project

  1. Navigate to the specific project directory:

    cd src/agents/ai_customer_support_agent

  2. Install dependencies:

    pip install -r requirements.txt

  3. Set up required API keys (varies by project):

    export OPENAI_API_KEY='your-key-here'

  4. Run the application (check the project's README for specific commands):

    # For Streamlit apps
streamlit run customer_support_agent.py

# For Python scripts
python3 rag_agent.py

Common Dependencies

Most projects use one or more of:

  • OpenAI API - GPT-4o, GPT-4o-mini models
  • Streamlit - Web UI framework for many applications
  • Vector Databases - Qdrant, LanceDB, ChromaDB for RAG systems
  • Memory Libraries - mem0ai for persistent memory
  • Agent Frameworks - phi (phidata), CrewAI, LangChain/LangGraph

External Services

Some projects require running external services:

Qdrant (Vector Database):

docker pull qdrant/qdrant
docker run -p 6333:6333 -p 6334:6334 \
    -v $(pwd)/qdrant_storage:/qdrant/storage:z \
    qdrant/qdrant

Key Architectural Patterns

RAG Applications

  • Use vector databases (LanceDB, Qdrant) for similarity search
  • Knowledge bases typically loaded from PDFs or web URLs
  • Often include web search tools (DuckDuckGo) for real-time information

Agent Applications

  • Built with frameworks like phi.agent, CrewAI
  • Use tool calling for web search, calculations, API access
  • Often include playground interfaces for testing

Memory-Enabled Applications

  • Use mem0ai library with vector stores for persistent memory
  • Store user interactions and preferences
  • Retrieve relevant context for personalized responses

Important Notes

  • No unified build system - Each project is independent
  • API keys required - Most projects need OpenAI or other LLM provider keys
  • Check individual READMEs - Setup steps vary significantly between projects
  • Python 3.x required - Most projects use modern Python features
  • Docker may be needed - For vector databases and other services

Repository Purpose

This is primarily a learning and reference repository containing:

  • Tutorial code demonstrating LLM application patterns
  • Example implementations of common LLM use cases
  • Documentation of LLM tools, frameworks, and models
  • Curated lists of papers, models, and resources in the main README.md

When modifying this repository, focus on:

  • Keeping individual projects self-contained
  • Maintaining clear README documentation for each project
  • Ensuring requirements.txt files are accurate
  • Following the established directory structure