🧭 General Knowledge Assistant

A modern, Streamlit-powered chatbot that combines Groq’s Llama-3-70B with Google Gemini 1.5 and on-demand web search to answer anything from trivia to deep technical questions.
It features multi-chat sessions, elegant UI/UX, and secure API-key handling – all in a single Python file.

🌐 Live demo: Try it instantly on Hugging Face → https://huggingface.co/spaces/Ashkchamp/General_Knowledge_Assistant

---

✨ Features

Capability	Details
Dual-LLM pipeline	• Groq Llama-3-70B (via langchain_groq) for core reasoning and responses. • Gemini 1.5-pro for meta-reasoning (decides when to web-search) and safe-content filtering.
Smart Web Search	Uses DuckDuckGo through langchain_community.tools.DuckDuckGoSearchRun only when Gemini signals <SEARCH>.
Persistent Chats	Auto-saves each conversation (name, messages, timestamp) in Streamlit Session State; switch, rename, or delete with one click.
Polished UI	Custom CSS for clean, mobile-friendly chat bubbles, avatars, and a subtle typing indicator.
Zero-backend setup	Runs locally – no database or server-side code required.
Secure Keys	API keys loaded from a local .env file (never stored in code or state).

---

🚀 Quick Start (Local)

# 1. Clone
git clone https://github.com/<your-handle>/general-knowledge-assistant.git
cd general-knowledge-assistant

# 2. Create & activate a virtual environment (recommended)
python -m venv .venv
source .venv/bin/activate         # On Windows: .venv\Scripts\activate

# 3. Install dependencies
pip install -r requirements.txt

# 4. Add your keys
cp .env.example .env              # then edit .env
# ────────────────────────────────
# GROQ_API_KEY=your_groq_key_here
# GEMINI_API_KEY=your_gemini_key_here
# ────────────────────────────────

# 5. Run
streamlit run app.py

Open http://localhost:8501 in your browser and start chatting!

(Or just use the hosted version on Hugging Face if you don’t have keys handy.)

---

🗂️ Project Structure

├── app.py              # Streamlit application
├── requirements.txt    # Python dependencies
├── .env.example        # Sample environment file
└── README.md

---

🔧 Configuration

Variable	Description
GROQ_API_KEY	Obtain from https://console.groq.com/
GEMINI_API_KEY	Obtain from https://aistudio.google.com/
(Optional) PORT	Override Streamlit default port via streamlit run app.py --server.port <PORT>

Tweaks

• Model choice – change model="llama-3.3-70b-versatile" in setup_models.
• UI theme – edit local_css() for colours, fonts, layouts.
• Search provider – swap DuckDuckGo for any other langchain tool.

---

🏗️ How It Works

1. User prompt ➜ stored in session history.
2. Gemini decides if the query needs a web search (<SEARCH> keywords) or not (NO_SEARCH).
3. If search required, DuckDuckGo fetches top results → passed to Groq with a “Search Results” prompt.
4. Groq Llama-3 generates the final answer.
5. UI shows a typing indicator while processing, then streams the response.
6. All messages & session metadata persist until the browser tab is closed (or deleted via sidebar).

---

📄 License

MIT – see LICENSE.

---

🙏 Acknowledgements

• Streamlit
• Groq
• Google Gemini
• LangChain
• DuckDuckGo