Gaurav970/jarvius

🤖 J.A.R.V.I.S. — Production Autonomous Personal AI Agent

A production-quality personal AI agent system inspired by Iron Man's JARVIS, featuring autonomous multi-step execution, dynamic replanning, 9 specialist agents, 21 operating tools, 5-tier persistent memory, proactive workflow automation, real-time voice mode with continuous listening & barge-in interruption, multi-model AI providers, and a cybernetic 3D WebGL HUD command center with a global emergency killswitch.

---

⚡ Quickstart

🚀 Option 1: GitHub Codespaces (1-Click Cloud Launch)

1. Fork or push this repository to GitHub.
2. Click Code -> Codespaces -> Create codespace on main.
3. Codespaces automatically builds the Python 3.11 environment, installs dependencies, copies .env.example to .env, and forwards port 8000.
4. Run in the Codespaces terminal:

   ./start_jarvis.sh
   # or: python -m jarvis.main
   

5. Click Open in Browser when port 8000 prompts, or visit the forwarded port URL!

---

💻 Option 2: Local Launch

On Windows:

• Using PowerShell:

  .\start_jarvis.ps1
  

• Using Batch:

  start_jarvis.bat
  

On Linux / macOS:

chmod +x start_jarvis.sh
./start_jarvis.sh

Then open your browser and navigate to: 👉 http://localhost:8000 (or http://127.0.0.1:8000)

---

🎙️ Real-Time Voice Runtime & Hologram

• Continuous 24/7 Voice Listening: Hands-free interaction using browser Web Audio and Voice Activity Detection (VAD).
• Instant Barge-in Interruption: Speak while JARVIS is responding, and JARVIS stops talking within 100ms and listens to your new command.
• Stop Command Classification: "Stop talking" silences speech while preserving background tasks; "Cancel task" halts active tasks.
• 3D WebGL Solar Core: Dynamic particle hologram reacting smoothly to IDLE, LISTENING, THINKING, SPEAKING, and ERROR states.

---

🏗️ System Architecture

jarvis/
├── .devcontainer/              # GitHub Codespaces container specification
│   └── devcontainer.json       # Auto Python 3.11 environment, port forwarding
├── agents/                     # Autonomous Agents & Planner
│   ├── base.py                 # Specialist agent base class & state
│   ├── core_agent.py           # Master JARVIS orchestrator & dialogue manager
│   ├── planner.py              # Multi-step task planner & replanner
│   └── specialists/            # 9 Autonomous specialist agents
│       ├── research_agent.py   # Web & citations research
│       ├── coding_agent.py     # Software engineering & test loops
│       ├── browser_agent.py    # Web navigation & DOM extractor
│       ├── computer_agent.py   # Desktop telemetry & app controller
│       ├── file_agent.py       # Workspace file manager
│       ├── data_agent.py       # JSON/CSV/SQL data analyst
│       ├── writing_agent.py    # Executive documentation writer
│       ├── scheduler_agent.py  # Temporal workflow scheduler
│       └── monitor_agent.py    # Health & sentinel monitor
├── api/                        # REST & WebSockets Layer
│   ├── routes.py               # REST API endpoints (Chat, Tasks, Tools, Memory, Voice)
│   └── websockets.py           # Live telemetry & activity streaming
├── database/                   # Persistent Database Layer
│   ├── db.py                   # Async SQLite connection pool
│   └── models.py               # Tasks, Steps, Messages, Memories, Workflows, Audit
├── memory/                     # 5-Tier Memory Engine
│   └── manager.py              # Short-term, Long-term, Episodic, Semantic, Project
├── providers/                  # Multi-Model AI Layer
│   ├── base.py                 # Abstract AIProvider
│   ├── gemini_provider.py      # Google Gemini 2.5/1.5 REST API
│   ├── openai_provider.py      # OpenAI GPT-4o / Ollama / LocalLLM
│   ├── anthropic_provider.py   # Claude 3.5 Sonnet REST API
│   ├── offline_provider.py     # Autonomous zero-dependency offline engine
│   └── factory.py              # Resilient fallback factory
├── runtime/                    # Task Runner & Swarm
│   ├── task_runner.py          # Asynchronous task execution & status lifecycle
│   └── multi_agent.py          # Parallel multi-agent swarm runner
├── security/                   # Security & Permissions
│   ├── permissions.py          # LOW, MEDIUM, HIGH, CRITICAL & Emergency Stop
│   └── audit.py                # Structured audit logging (audit.jsonl)
├── tools/                      # 21 Plugin Tools
│   ├── terminal_tools.py       # Shell execution & process management
│   ├── filesystem_tools.py     # Read, write, safe edit, search, delete
│   ├── web_tools.py            # Web search, page reader, REST caller
│   ├── browser_tools.py        # Web navigation & extraction
│   ├── computer_tools.py       # Screenshots, telemetry, window lister, app launcher
│   ├── coding_tools.py         # Pytest runner, AST parser, syntax checker
│   ├── memory_tools.py         # Memory store and semantic search
│   └── scheduler_tools.py      # Task scheduler
├── ui/static/                  # Futuristic Obsidian HUD
│   ├── index.html              # Cybernetic Command Center UI
│   ├── styles.css              # Glassmorphic HUD styles & animations
│   ├── app.js                  # WebSocket & Web Speech API client
│   └── ai_blob.js              # 3D WebGL Holographic Orb & Solar Flare
├── voice/                      # Real-time Voice Runtime
│   ├── runtime.py              # AttentionManager, VoiceRuntimeState, barge-in logic
│   └── voice_engine.py         # TTS engine with Web Speech synthesis fallback
├── workflows/                  # Automation Engine
│   └── engine.py               # Cron triggers, proactive morning briefings
├── tests/                      # Automated Test Suite
├── verify_acceptance.py        # 20-Criteria Final Acceptance Verification Script
├── verify_voice_runtime.py     # 12-Criteria Voice Runtime Verification Script
└── main.py                     # FastAPI entrypoint

---

🛡️ Safety & The Emergency Killswitch

A prominent glowing red STOP ALL AGENTS button is embedded in the top HUD header and bound to POST /api/system/emergency-stop. Activating it instantly terminates all active tasks, kills running subtasks, and locks tool execution until manually reset via Resume Agents.

---

🧪 Verification & Tests

All tests are 100% automated and passing:

• PyTest Suite:

  python run_tests.py
  

• 20-Criteria Acceptance Verification:

  python verify_acceptance.py
  

• 12-Criteria Real-Time Voice Verification:

  python verify_voice_runtime.py