Skip to content

PLAN.md

Latest commit

 

History

History
92 lines (76 loc) · 3.63 KB

PLAN.md

File metadata and controls

92 lines (76 loc) · 3.63 KB
Raw

AI Agent Integration for AcreetionOS

Goal

Integrate an AI agent into AcreetionOS to provide intelligent system assistance, automation, and natural language interaction for users.

Options Investigated

1. Aider AI (aider.chat)

  • Type: CLI pair-programming AI
  • Pros: Excellent code generation, git-aware, multi-model support
  • Cons: Focused on development tasks, not general system administration
  • Verdict: Good for development tool, not ideal for end-user system agent

2. Open Interpreter (openinterpreter.com)

  • Type: Natural language computer interface
  • Pros: Runs code locally, shell access, file management, web research
  • Cons: Requires API key for LLM, can be dangerous if unrestricted
  • Verdict: Strong candidate — can be bundled as optional package

3. Custom AcreetionOS Assistant (Recommended Path)

  • Type: Lightweight shell-based assistant with optional LLM backend
  • Architecture:
    • Shell script interface for local operations (package management, system info, logs)
    • Optional LLM backend via Ollama (local) or API (remote)
    • TUI/CLI mode with plans for GUI integration
  • Pros: Fully controllable, privacy-respecting, offline-capable
  • Cons: Requires development effort

4. Ollama + Custom Frontend

  • Type: Local LLM runtime
  • Pros: Fully offline, many models available, lightweight runtime
  • Cons: Requires significant RAM for larger models, GPU recommended
  • Verdict: Best foundation — use as default LLM backend

5. GitHub Copilot / Claude Code

  • Type: Developer AI tools
  • Pros: Excellent for development
  • Cons: Not designed for system-level tasks
  • Verdict: Keep as optional developer tools only

Recommended Architecture

┌─────────────────────────────────────┐
│         User Interface              │
│  (CLI / TUI / GUI Plugin)           │
├─────────────────────────────────────┤
│         acreetionos-ai              │
│  (Core agent orchestrator)          │
├──────────────────┬──────────────────┤
│  Local Actions   │  LLM Backend     │
│  (pacman, logs,  │  (Ollama / API)  │
│   system, files) │                  │
└──────────────────┴──────────────────┘

Implementation Plan

Phase 1 — CLI Agent (Minimum Viable)

  • Create acreetionos-ai CLI script
  • Integrate with pacman for package queries/installs
  • System info, log analysis, configuration helpers
  • Optional: shell pipeline execution with confirmation

Phase 2 — LLM Integration

  • Support Ollama for local inference
  • Support remote API (OpenAI-compatible) as alternative
  • Context-aware responses based on system state
  • Conversation history

Phase 3 — TUI/GUI

  • Terminal UI using gum or fzf
  • GUI plugin for Cinnamon desktop
  • Status panel integration

Files to Create

  • ai/acreetionos-ai — Main agent script
  • ai/actions/ — Pluggable action modules
  • ai/llm/ — LLM backend connectors
  • ai/config — User configuration

Package Dependencies (Optional)

  • ollama — Local LLM runtime (optional)
  • python or bash — Agent runtime
  • gum — TUI components (optional)
  • fzf — Fuzzy finding (optional)

Security Considerations

  • All local actions require user confirmation
  • Remote API keys stored in user config
  • Sandboxed execution for LLM-generated commands
  • Audit log of all AI-initiated changes