Adan.ai

Inspiration
We wanted to create a futuristic local AI assistant that works offline, protects user privacy, and feels alive. Inspired by science fiction and modern AI assistants, we decided to merge multiple technologies into one project.

What it does
Adan.AI can:
Chat locally using DeepSeek R1 (7B) integrated with GPT-OSS generated code. Extract text from PDF, Word, Excel, HTML, and more.
Summarize long documents quickly.
Perform sentiment and emotion analysis.
Speak naturally using ElevenLabs TTS integration.

How we built it
Backend: FastAPI with Python, managing chat, file parsing, and TTS endpoints.
Frontend: Custom HTML/CSS/JS futuristic interface inspired by modern Apple-style design.
AI Core: DeepSeek R1 (7B) running fully offline, with prompts generated using GPT-OSS during development.
Extra: ElevenLabs for voice replies, multi-format text extraction libraries.

Challenges we ran into
Running large models locally with limited hardware.
Integrating multiple file formats and making extraction reliable.
Balancing design and usability under strict time pressure.
Submitting under age restrictions (resolved with parental representation).

Accomplishments that we're proud of
A fully working local AI system with chat, extraction, summarization, emotion detection, and speech.
Seamless frontend-backend integration with a polished UI.
Staying true to our vision: offline, private, futuristic AI assistant.

What we learned How to use GPT-OSS effectively in real-world coding tasks.
How to integrate large language models locally without relying on the cloud.
Teamwork: backend, frontend, research all came together.

What's next
Expanding hardware integration (robotics, IoT).
Adding multimodal inputs (camera, gesture recognition).
Releasing Adan.AI as an open-source project for the community.

Built With

• beautifulsoup4
• css
• deepseek
• elevenlabs
• fastapi
• gpt-oss
• html
• javascript
• openpyxl
• pdfminer.six
• python
• python-docx
• python-pptx