T.I.T.A.N.

Tactical Interface for Task Automation & Navigation

An autonomous, hybrid AI assistant that lives locally on your Windows PC (optimized for AMD Ryzen AI NPU + Radeon Graphics), with persistent multi-tier memory, Windows OS automation, web browsing, self-learning & skill synthesis, and voice wake-word capabilities.

---

⚡ Key Features

1. Local-First Brain with Smart Cloud Escalation

   • Tier 1 (Local NPU): FastFlowLM running qwen2.5-it:3b on your AMD Ryzen AI NPU handles commands, OS automation, and daily queries with 0ms internet latency and complete privacy.
   • Tier 2 (Cloud Escalation): Tasks requiring heavy multi-step coding, extensive document analysis, or complex reasoning automatically escalate to Google Gemini (gemini-2.5-flash).
   • Visual badges in terminal show whether a response was generated by [LOCAL NPU] or [CLOUD GEMINI].

2. Persistent Memory Subsystem

   • Tier 1 (Working): Active conversation and planning buffer.
   • Tier 2 (Episodic): SQLite database (titan_memory.db) recording full interaction logs, tool runs, and timestamps.
   • Tier 3 (Semantic): JSON long-term knowledge base (knowledge_store.json) for user preferences, facts, and habits.

3. Self-Reflection, Code Evolution & Git Version Control

   • Autonomous Self-Improvement: Can inspect, refactor, and rewrite its own codebase (core/evolution.py) to fix bugs, optimize algorithms, or add new capabilities.
   • Pre-Flight Test Verification: Automatically executes the test suite (test_titan.py) before committing any code changes.
   • Automatic GitHub Checkpoints: Every verified code evolution and synthesized skill is automatically committed and pushed to GitHub with descriptive changelogs.
   • Fail-Safe Rollback: If an experimental change fails self-tests or degrades behavior, TITAN automatically restores the last healthy Git commit.
   • Procedural Skills: Generates executable Python modules into tools/dynamic_skills/ on the fly, hot-reloading new skills without restarting.

4. Windows & Web Automation

   • Launch apps (calc, notepad, chrome, spotify, steam, discord, explorer, settings).
   • Open websites, wikis, and direct search topics in the default browser.
   • Run PowerShell commands and scripts.
   • Control audio volume and mute state.
   • Live web search and webpage content extraction.
   • System telemetry (CPU, RAM, Battery, Storage).

5. Deep Media, Window & Workstation Controls

   • System Media Transport: Play/pause, next track, previous track, stop media across Spotify, YouTube, and media players.
   • Window Management: Minimize all / show desktop, restore, snap left/right, maximize, minimize, switch tasks.
   • Workstation Power & Security: Lock workstation (Win+L), sleep system, turn off display screens, or schedule auto-shutdowns.

6. Screen Perception & Multimodal Vision

   • Real-Time Desktop Capture: Captures active displays and windows at native High-DPI resolutions.
   • Visual Error & Code Analysis: Can inspect IDE errors, charts, documents, or UI layouts on your screen and provide instant tactical solutions.
   • Command Trigger: "Look at my screen and tell me what is open", "Explain this error on my display".

7. Voice Wake Word & Audio

   • Voice wake-word detection for "TITAN".
   • Offline Speech-To-Text via faster-whisper.
   • Offline Text-To-Speech via Windows SAPI5.

---

🚀 Quick Start

1. Configure .env

Copy .env.example to .env and insert your Google AI Studio API key:

GEMINI_API_KEY=your_google_ai_studio_api_key_here
GEMINI_MODEL=gemini-2.5-flash

# Local NPU / FLM Server Settings (when running FLM / local server)
LOCAL_MODEL_ENABLED=true
LOCAL_MODEL_ENDPOINT=http://127.0.0.1:52625/v1
LOCAL_MODEL_NAME=qwen2.5-it:3b

TITAN_MODE=hybrid
WAKE_WORD=titan
VOICE_ENABLED=false

2. Launch T.I.T.A.N.

# Interactive Terminal Mode
python titan.py

# Floating Cyberpunk Desktop HUD Overlay
python titan.py --hud

# HUD with Voice Wake-Word Activated
python titan.py --hud --voice

Global Hotkey: Press Ctrl + Space or Ctrl + Shift + T anywhere in Windows to summon or hide the HUD instantly!

---

⌨️ Tactical Slash Commands

• /help - Show command manual
• /stats - Live CPU, RAM, and Battery telemetry
• /memory - Inspect saved facts and preferences
• /skills - View dynamic learned skills
• /history - View autonomous code evolution & Git version history
• /rollback - Revert codebase to previous Git commit (e.g. /rollback or /rollback 2)
• /voice - Toggle wake-word listening & voice synthesizer
• /clear - Reset current conversation context
• /exit - Power down T.I.T.A.N.