A simple RAG-powered assistant for querying Ellie.ai's technical documentation using natural language.
- Natural language queries over documentation
- Source citation with document references
- Conversation history tracking
- Hallucination safeguards
- Python 3.10+
- Groq/Google Gemini API Key (set as
GROQ_API_KEYorGOOGLE_API_KEYin environment) - Milvus DB
git clone https://github.com/Senthi1Kumar/chat-with-docs.git
cd chat-with-docs
python -m venv .venv
# source .venv/bin/activate or .venv\Scripts\activate
pip install -r requirements.txt- Copy example config and update environment variable (.env) with your Groq API:
cp config/.env.example config/.env- Update with your:
- Documentation URLs
- Milvus connection details - path to store the DB
- Device preferences (CPU/GPU)
- Ingest documentation:
python index_docs.py- Start chat interface:
streamlit run app.py- To Execute evaluation, please set the Gemini API key in
.envfile:
python evaluate.pyThe evaluation framework provided in evaluate.py is designed to access the performance of our RAG system which leverages Google Generative AI (via ChatGoogleGenerativeAI class). It prepares an evaluation dataset, runs the RAG system on a set of pre-defined test questions, and computes multiple evaluation metrics such as Faithfulness, Answer Correctness, Context Recall, LLM Context Recall, and Factual Correctness using the RAGAS library.
MIT License