A local, privacy-focused assistant for Linux users — powered by a fine-tuned Phi-2 model running offline via Ollama.
Smart Linux Co-Pilot is a personal project aiming to bring a smart, offline AI assistant to the Linux terminal. It's designed to help users — especially beginners — understand and use Linux commands through natural language conversations.
The assistant runs entirely offline using a fine-tuned Phi-2 model, making it fast, lightweight, and private.
- 💬 Interact with the assistant through a text-based CLI interface
- 🧠 Uses a locally fine-tuned Phi-2 model for understanding queries
- 📂 Responds to system-related questions like:
- “List files in the current directory”
- “How do I check system memory?”
- “What does the 'grep' command do?”
- ⚙️ Suggests safe Linux commands based on natural language inputs
- 🔒 Works completely offline — no API calls, no cloud
- Expand understanding of Linux commands and system operations
- Add command execution (currently suggests only)
- Add simple help/documentation lookup from
manpages - Optional: integrate voice input in the future (Vosk/Whisper)
- Improve prompt engineering and fine-tuning dataset
Many users avoid Linux because of its learning curve and cryptic commands. This project aims to:
- Make Linux more user-friendly
- Allow users to learn Linux naturally via conversation
- Build an AI assistant that runs entirely on-device, respecting user privacy
| Component | Tool / Library |
|---|---|
| Language Model | Fine-tuned Phi-2 |
| Runtime | Ollama |
| Backend | Python |
| Interface | CLI (Terminal) |
| OS Support | Linux |
"""bash
What command shows disk usage?
Assistant:
You can use: df -h
It shows disk space in human-readable format.
How to see hidden files?
Assistant:
Use: ls -a """
This project is in early development.
Voice input is not yet implemented.
Output is suggestive; command execution is not enabled yet.
GitHub: 07Codex07
LinkedIn: www.linkedin.com/in/vinayak-sahu-8999a9259
Email: vinayak1672006@gmail.com
Let me know if you want:
- Help updating your
run_assistant.pyfile structure - A matching GitHub description and project tags
- LinkedIn or blog post format for this project
You're building this the right way — one real feature at a time.