SmartRAG is a terminal-based Retrieval-Augmented Generation (RAG) system built using LangGraph. It routes user queries through a custom flow that includes message history, query transformation, and document retrieval from a vector store.

🔗 GitHub: https://github.com/aimaster-dev/SmartRAG

• LangGraph-powered RAG pipeline
• Smart routing of user queries
• PDF and Markdown ingestion support
• Optional webpage-to-PDF and PDF-to-Markdown conversion
• OpenAI GPT integration for natural language responses

SmartRAG/
├── architecture/         # LangGraph RAG workflow logic
├── data/                 # Processed markdown or PDF content
├── modules/              # Core logic for query handling & doc processing
├── main.py               # Entry point
└── processDocs.py        # Document preprocessing script

Follow the steps below to get SmartRAG up and running:

git clone https://github.com/aimaster-dev/SmartRAG.git
cd SmartRAG

python3.12 -m venv venv
source venv/bin/activate  # Windows: venv\Scripts\activate

pip install -r requirements.txt
choco install wkhtmltopdf  # for HTML to PDF conversion (Windows only)

Copy and edit the .env file:

cp .env.example .env

Edit .env to include:

OPENAI_API_KEY=your_openai_key
URLS=url1,url2         # Optional: URLs to fetch as PDF
GET_WEB_PAGES_TO_PDF=True
CONVERT_PDF_TO_MD=True
INTERMEDIATE_PDF_DIR=./pdfs
DATA_DIR=./data

python modules/processDocs.py

⚠️ Make sure to update .env parameters based on your use case.

python main.py

1. User query is passed into a LangGraph workflow.
2. Message history is cached and contextually enriched.
3. If needed, input is transformed for better retrieval.
4. Documents are pulled from a vector store using similarity search.
5. GPT model generates a context-aware answer.

We welcome contributions!

• Fork the repo
• Create a feature branch
• Submit a pull request

Got a big idea? Open an issue to discuss it first.

For questions, feedback, or collaboration ideas — feel free to open an issue or reach out through GitHub!