An unfiltered, strategic personal AI — built to think, not please.
Still in its early stage, still evolving — but even now, it cuts through noise like a blade.
Built for myself. cold, strategic eerie AI inspired by HAL 9000 (2001: A Space Odyssey). Always drawn to that inhuman, logical calm. Voice is there, but not quite HAL yet — the text makes up for it. Built with Python (Flask) and vanilla js powered by Groq’s LLaMA API (needs internet). Short-term memory. Still in phase one
A base model for a ruthless AI assistant, HAL is designed to:
- Be brutally honest and coldly strategic
- Store per-user context (ephemeral for now)
- Serve responses via Groq’s blazing-fast LLaMA models
- Be extended, repurposed, or rebranded by anyone
If you're looking for sugarcoated answers or polite small talk, you're in the wrong repo.
| Language | % |
|---|---|
| Python | 51.9% |
| JavaScript | 25.3% |
| CSS | 16.4% |
| HTML | 6.4% |
- Backend: Flask
- Frontend: Vanilla JS, HTML, CSS (minimal and raw)
- AI: Groq API using
llama3-8b-8192 - Memory: In-memory per user session (stateless reset included)
- System Prompt crafted for maximum edge and focus
- User Memory stored per session, with easy
/resetoption - Retry mechanism for Groq API fails (no weak links)
- Simple frontend that gets out of your way
- CORS enabled for full-stack flexibility
Already hosted and running.
But if you're forking it or customizing:
- Update
.envwith your Groq API key - Edit the
SYSTEM_PROMPTinapp.pyto match your vision - Drop any frontend changes in the
frontend/folder — backend serves it automatically
Make your own version:
- Change the name
- Replace the tone
- Extend memory, auth, storage, even add a voice layer
This is the base model. A skeleton with attitude. Build on it. Break it. Push it further.
Use it. Remix it. Sell it. Credit optional.
Just don’t make it lame.
HAL is not done. This is just the beginning.
More features, memory persistence, auth, maybe even a proper frontend — coming when it’s time. Not rushed. Not bloated. Not for show.