title	Personal Website Backend
emoji	🌐
colorFrom	blue
colorTo	indigo
sdk	docker
pinned	false

RAG LLM Agent Backend

This is the backend service for my personal website, implementing a Retrieval-Augmented Generation (RAG) agent that processes queries about my professional background and experience.

Features

• RAG system for retrieving relevant information from a knowledge base
• Integration with GitHub and LinkedIn for external data
• Groq LLM integration for generating responses
• FastAPI server for handling requests

API Usage

Send a POST request to /query with a JSON body containing your query:

curl -X POST "https://huggingface.co/spaces/ereinha3/personal-website/query" \
     -H "Content-Type: application/json" \
     -d '{"text": "Your query here"}'

The response will include:

• The original query
• LLM-generated response
• Relevant context from the knowledge base
• A list of relevant documents

Environment Variables

The following environment variables need to be set in your Hugging Face Space:

• GITHUB_TOKEN: Your GitHub personal access token
• GITHUB_USERNAME: Your GitHub username
• LINKEDIN_EMAIL: Your LinkedIn email
• LINKEDIN_PASSWORD: Your LinkedIn password
• GROQ_API_KEY: Your Groq API key

Local Development

1. Install dependencies:

pip install -r requirements.txt

2. Set up environment variables in a .env file

3. Run the server:

python main.py

The server will start on http://localhost:8000

Local Setup

1. Install dependencies:

pip install -r requirements.txt

2. Create a .env file with your API keys:

# GitHub credentials
GITHUB_TOKEN=your_github_personal_access_token
GITHUB_USERNAME=your_github_username

# LinkedIn credentials
LINKEDIN_EMAIL=your_linkedin_email
LINKEDIN_PASSWORD=your_linkedin_password

# Groq API credentials
GROQ_API_KEY=your_groq_api_key

3. Create a knowledge directory and add your documents:

knowledge/
├── resume/
│   ├── resume.pdf
│   └── resume.md
├── projects/
│   ├── project1/
│   │   ├── README.md
│   │   └── code/
│   └── project2/
│       ├── README.md
│       └── code/
└── other/
    └── additional_info.md

4. Run the server:

python main.py

The server will start on http://localhost:8000.

Deployment to Hugging Face Spaces

1. Create a new Space on Hugging Face:

   • Go to https://huggingface.co/spaces
   • Click "Create new Space"
   • Choose a name for your space
   • Select "Docker" as the SDK
   • Choose "Public" or "Private" visibility

2. Add your environment variables:

   • Go to your Space settings
   • Navigate to the "Repository Secrets" section
   • Add the following secrets:
     • GITHUB_TOKEN
     • GITHUB_USERNAME
     • LINKEDIN_EMAIL
     • LINKEDIN_PASSWORD
     • GROQ_API_KEY

3. Push your code to the Space:

   • The Space will automatically build and deploy your application
   • You can monitor the build process in the "Factory" tab

4. Access your deployed application:

   • Once deployed, your application will be available at https://huggingface.co/spaces/YOUR_USERNAME/YOUR_SPACE_NAME

Troubleshooting Deployment Issues

If you encounter issues when pushing to Hugging Face Spaces, try these solutions:

1. Check Git Configuration:

   git config --global user.name "Your Name"
   git config --global user.email "your.email@example.com"

2. Verify Remote URL:

   git remote -v

   Make sure the URL is correct: https://huggingface.co/spaces/YOUR_USERNAME/YOUR_SPACE_NAME

3. Try Using Hugging Face CLI:

   pip install huggingface_hub
   huggingface-cli login
   git push space main

4. Check File Size Limits:

   • Hugging Face has file size limits
   • If your knowledge base is large, consider hosting it elsewhere

5. Check Docker Build Logs:

   • Go to your Space's "Factory" tab
   • Look for any error messages in the build logs

6. Simplify Your Dockerfile:

   • If the build is failing, try a simpler Dockerfile first
   • Once that works, gradually add back the complexity

7. Use Hugging Face's Web Interface:

   • Instead of pushing via Git, try uploading files directly through the web interface
   • This can help identify if the issue is with Git or with the files themselves

Notes

• The agent uses the sentence-transformers/all-MiniLM-L6-v2 model for embeddings
• Documents are chunked into 1000-character segments with 200-character overlap
• The agent returns the top 3 most relevant documents for each query
• External data (GitHub and LinkedIn) is updated every 24 hours
• The Groq Mixtral model is used for generating responses