💬 Chain-Field Chatbot (Solana + EVM)

A simple chatbot with two selectable chain fields (Solana and EVM). You pick a field and chat; answers are grounded in that field's knowledge base via RAG.

---

📸 Preview

The app shows a clean UI:

• 🔽 Chain field – Dropdown to select Solana or EVM
• 💬 Chat – Type your question (e.g. "Hello, what is solana"); the bot replies using that chain's docs

---

✅ Requirements

• Python 3.10+ (including 3.14). The app uses FAISS for the vector store (no ChromaDB).

---

🚀 Setup

1. Clone or open the project and go to the project root.

2. Create a virtual environment (recommended):

   python -m venv .venv
   .venv\Scripts\activate   # Windows
   # source .venv/bin/activate  # macOS/Linux

3. Install dependencies:

   pip install -r requirements.txt

4. Configure environment:

   • Copy .env.example to .env
   • Chat (LLM): set OPENAI_API_KEY (required for replies).
   • Ingest (embeddings): default is free Hugging Face API. Set HF_TOKEN (get one at huggingface.co/settings/tokens). If you hit OpenAI quota (429), use EMBEDDING_PROVIDER=huggingface and HF_TOKEN=hf_... so ingest runs without OpenAI.

   Example .env:

   OPENAI_API_KEY=sk-...
   LLM_MODEL=gpt-4o-mini
   EMBEDDING_PROVIDER=huggingface
   HF_TOKEN=hf_...
   

---

📚 Ingest documents (optional but recommended)

• Put Solana-related docs (.md, .txt, .pdf) in data/solana/
• Put EVM-related docs in data/evm/
• See data/README.md for sources (e.g. Solana Cookbook, Ethereum docs)

Then run:

python -m app.rag.ingest

This builds FAISS indexes under vector_store/solana and vector_store/evm. If you skip this, the app will error on first chat until you run ingest.

---

▶️ Run the app

1. Start the backend (from project root):

   uvicorn app.main:app --reload --host 0.0.0.0 --port 8000

2. Start the frontend (in another terminal):

   streamlit run frontend/streamlit_app.py

3. Open the URL shown by Streamlit (e.g. http://localhost:8501). Choose Solana or EVM, then type your message.

---

🔧 Optional: backend URL

If the API runs on another host/port, set:

set CHATBOT_BACKEND_URL=http://localhost:8000
streamlit run frontend/streamlit_app.py

---

📁 Project layout

• app/ – FastAPI app, config, RAG (ingest + chains), Pydantic models
• app/rag/ingest.py – Ingest script: python -m app.rag.ingest
• app/rag/chains.py – RAG retrieval and get_answer(field, message)
• frontend/streamlit_app.py – Streamlit UI
• data/solana/, data/evm/ – Documents to ingest per field
• vector_store/ – FAISS index per field (created at runtime; in .gitignore)