Election Program Chatbot

Overview

This project provides an interactive chatbot that helps users easily access election program information from major German parties. It uses Retriever-Augmented Generation (RAG) to deliver direct and clear answers from available sources.

Features

• Easy-to-use interface: Initially built as a web application (Gradio MVP), with future expansion to mobile versions.
• Direct answers, not just search results: Uses RAG to extract relevant information from election programs.
• Supports major parties: Includes CDU, SPD, Greens, Die Linke, AfD, BSW, Die Piraten, Volt, and FDP.
• Multiple data sources:
  • PDF documents (e.g., official election programs)
  • YouTube transcripts (for speeches and campaign materials)
  • Future expansion: Podcasts
• Efficient vector search:
  • Pinecone (default) for scalable cloud-based retrieval
• LLM-powered responses:
  • Uses GPT-4o-mini API for intelligent and multilingual answers (German/English).
  • Optimized prompt engineering for precise output.

Setup Instructions

1. Install Dependencies

Run the following command to install the required libraries:

pip install -r requirements.txt

2. Set Up API Keys

The application requires API keys for accessing external services. Store them in a .env file:

YOUTUBE_API_KEY=your_youtube_api_key
OPENAI_API_KEY=your_openai_api_key

For Pinecone vector storage, set up:

PINECONE_API_KEY=your_pinecone_api_key
PINECONE_ENV=your_pinecone_environment

3. Run the Application

Launch the chatbot with:

python main.py

File Structure

1. PDF Loader (pdf_loader.py)

• Extracts text from PDF documents.
• Used to process official election programs.

2. YouTube Loader (YT_loader.py)

• Fetches transcripts from political videos.
• Uses YouTube API for metadata extraction.

3. Retriever (retriever.py)

• Performs vector-based search on stored election data.
• Uses Pinecone for efficient retrieval.

4. Main File (main.py)

• Integrates all components into a functional chatbot.
• Calls retrievers and loaders as needed.

Future Improvements

• Expand to mobile and voice-based interaction.
• Include Twitter and blog content creation for real-time political insights.
• agent improvement asking for personal request, political characters for each party

Contribution

Contributions are welcome! Feel free to submit issues or pull requests.

---

Author: Paul Müller