Developed Auranix, an AI-powered query assistant capable of processing long, complex natural language queries and delivering 100% authenticated, source-backed answers with deep contextual understanding.
Built with a Flutter frontend and FastAPI backend, Auranix leverages WebSockets for real-time streaming responses, integrates Tavily and Trafilatura for intelligent web search and content extraction, and uses Google Gemini API (Generative AI) to generate fast, human-like, and trustworthy conversational search results — boosting productivity through seamless interaction.
- Purpose: Provide a fast, citation-aware AI assistant that searches the web, extracts high-quality content, ranks & summarizes it, and returns streamed answers to the user with source links.
- Platforms: Flutter (mobile/web) frontend, FastAPI backend (Python).
- Key integrations: Gemini API (generative responses), Tavily (web search), Trafilatura (clean HTML extraction),
sentence-transformers(semantic similarity ranking). - Realtime: WebSocket for low-latency streaming responses and updates.
- Real-time conversational interface (WebSocket streaming)
- Web search via Tavily, extract content using Trafilatura
- Semantic rank of candidate sources using Sentence-Transformers (e.g.,
all-MiniLM-L6-v2) - Result generation via Gemini API (Generative AI)
- Source cards with clickable URLs
- Flutter frontend (mobile + web) with responsive UI
Backend
- Python 3.10+
- FastAPI, Uvicorn
- WebSockets
tavily(real-time web search API)trafilatura(HTML content extraction & cleanup)sentence-transformers(semantic similarity ranking on retrieved sources)numpy(cosine similarity)python-dotenv(environment variable management)
AI / External Services
- Google Gemini API (Generative AI)
- Tavily (web search)
Frontend
- Flutter (Dart) — cross-platform mobile & web app
flutter_markdown— render markdown-formatted AI responsesurl_launcher— open external source links in browserskeletonizer— display skeleton loading states while fetching AI results
git clone https://github.com/abbasdx/auranix-ai.git
cd auranix-aicd server
python -m venv venv
# Windows
venv\Scripts\activate.ps1
# macOS / Linux
source venv/bin/activate
pip install -r requirements.txtCreate .env (in server folder):
GEMINI_API_KEY=your_gemini_api_key
TAVILY_API_KEY=your_tavily_api_key
Run backend:
fastapi dev main.py- HTTP endpoints:
http://localhost:8000/ - WebSocket endpoint (example):
ws://localhost:8000/ws/chat
flutter pub get
flutter run -d chrome # or `flutter run` for deviceCreate a .env file in server/:
# Required
GEMINI_API_KEY=your_gemini_api_key
TAVILY_API_KEY=your_tavily_api_keyThanks for wanting to contribute! Please:
- Fork the repo
- Create a branch:
git checkout -b feature/short-description - Commit:
git commit -m "Add: short description" - Push:
git push origin feature/short-description - Open a Pull Request describing the change
Please run tests and update docs as needed.
This project is open source under the MIT License — see the LICENSE file.
Please take permission from me before using this project in commercial or high-scale applications.
- Google Gemini & Google AI Platform — for powering generative and conversational AI.
- Tavily — for accurate and relevant web search capabilities.
- Trafilatura — for robust HTML content extraction.
- FastAPI — for the high-performance backend framework.
- Uvicorn — for the lightning-fast ASGI server.
- Sentence-Transformers — for semantic similarity ranking of search results.
- NumPy — for efficient mathematical operations and cosine similarity calculations.
- python-dotenv — for secure environment variable management.
- Flutter — for building cross-platform web and mobile applications.
- flutter_markdown — for rendering AI responses with rich text formatting.
- url_launcher — for seamlessly opening external source links.
- skeletonizer — for smooth loading state placeholders in UI.
- Starlette — for WebSocket support and async capabilities in FastAPI.
- Open Source Community — for contributing tools, libraries, and knowledge that made this project possible.
If you want help, improvements, or collaboration — open an issue or reach out at: abbasxansari@gmail.com
Enjoy building with Auranix — fast, source-aware, and human-like AI conversations.