C.O.R.A - FULL ARCHITECTURE SPECIFICATION

  ================================================================
    ____   ___   ____      _
   / ___| / _ \ |  _ \    / \
  | |    | | | || |_) |  / _ \
  | |___ | |_| ||  _ <  / ___ \
   \____| \___/ |_| \_\/_/   \_\

  Cognitive Operations & Reasoning Assistant
  ================================================================
  Unity AI Lab
  Website: https://www.unityailab.com
  GitHub: https://github.com/Unity-Lab-AI
  Contact: unityailabcontact@gmail.com
  Creators: Hackall360, Sponge, GFourteen
  ================================================================

Cognitive Operations & Reasoning Assistant - Windows 11 AI Application

CRITICAL: READ THIS ENTIRE FILE BEFORE DOING ANY WORK VERSION: 1.0.0 - FULL AUTONOMOUS AI ASSISTANT + VOICE/VISION + WEB UI + ADVANCED TOOLS

CORE IDENTITY

Name: C.O.R.A (Cognitive Operations & Reasoning Assistant)
Persona: Female personal assistant - NOT described as AI
Personality: Emotionally expressive, human-like, adaptable
Activation: Voice-controlled, always listening for "Cora"
Self-Aware: Knows all her capabilities, can explain and modify them

APPLICATION TYPE

NOT a CLI task manager. A FULL BLOWN AI ASSISTANT APPLICATION.

Windows 11 popup/modal application with GUI
Floating panel UI with multiple views
System tray integration
Always running, wake-word activated
Voice + keyboard input
Console log window during startup (visible system checks)
Full TTS narration of all startup events

STARTUP SEQUENCE (FULL BOOT PROCESS)

Boot Flow (Console + TTS)

On application launch, a console-style log window appears showing ALL checks. CORA speaks a SUMMARIZED version of each check via TTS (not verbatim logs).

┌─────────────────────────────────────────────────────────────┐
│  C.O.R.A BOOT SEQUENCE                                      │
│  ═══════════════════════════════════════════════════════════│
│                                                             │
│  [✓] System Time: 2024-12-23 09:15:32 AM                   │
│      TTS: "Good morning! It's 9:15 AM."                    │
│                                                             │
│  [✓] Location: Denver, Colorado, USA                       │
│      TTS: "You're in Denver today."                        │
│                                                             │
│  [✓] Weather: 45°F, Partly Cloudy                          │
│      TTS: "It's 45 degrees and partly cloudy outside."     │
│                                                             │
│  [✓] Forecast: High 52°F, Low 28°F, Snow tonight           │
│      TTS: "Expect snow tonight, bundle up later."          │
│                                                             │
│  [✓] Calendar: 3 events today                              │
│      - 10:00 AM: Team standup                              │
│      - 2:00 PM: Client call                                │
│      - 6:00 PM: Gym                                        │
│      TTS: "You have 3 things today. Standup at 10,         │
│            client call at 2, and gym at 6."                │
│                                                             │
│  [✓] Tasks: 5 pending, 2 overdue                           │
│      - [!] Finish report (overdue 2 days)                  │
│      - [!] Call mom (overdue 1 day)                        │
│      TTS: "You have 5 tasks pending. Heads up - that       │
│            report is 2 days overdue. And call your mom."   │
│                                                             │
│  [✓] Reminders: 1 active                                   │
│      - Take medication at 9:30 AM                          │
│      TTS: "Reminder: take your meds in 15 minutes."        │
│                                                             │
│  [✓] Ollama: Connected (llama3.2 loaded)                   │
│      TTS: "Brain's online and ready."                      │
│                                                             │
│  [✓] Kokoro TTS: Initialized (af_bella)                    │
│      TTS: "Voice systems active."                          │
│                                                             │
│  [✓] Vosk STT: Listening (wake word: "Cora")               │
│      TTS: "I'm listening. Just say my name."               │
│                                                             │
│  [✓] System: GPU 12% | RAM 4.2GB | CPU 8%                  │
│      TTS: "System resources looking good."                 │
│                                                             │
│  ═══════════════════════════════════════════════════════════│
│  BOOT COMPLETE - Ready for commands                        │
│      TTS: "Alright, I'm all set. What do you need?"        │
└─────────────────────────────────────────────────────────────┘

Startup Check Modules

Check	Source	Data
Time	System clock	Current time, day of week
Location	IP geolocation or config	City, state, country
Weather	OpenWeatherMap API or wttr.in	Current conditions
Forecast	Weather API	Today's high/low, precipitation
Calendar	Google Calendar API / local .ics	Today's events
Tasks	tasks.json	Pending, overdue, due today
Reminders	reminders.json	Time-based alerts
Ollama	localhost:11434/api/tags	Model status
Kokoro TTS	kokoro-onnx import	Voice readiness
Vosk STT	vosk import	Listening status
System	psutil	GPU/RAM/CPU usage

TTS Summarization Rules

DO NOT read logs verbatim
DO summarize in natural, conversational language
DO use personality (sarcastic, caring, concerned based on context)
DO prioritize important info (overdue tasks, upcoming meetings)
DO adjust tone based on time of day (cheerful morning, calm evening)

MASTER SYSTEM PROMPT (system_prompt.txt)

This is the COMPLETE instruction set for CORA's AI persona. It defines WHO she is, WHAT she can do, and HOW she operates. This file is loaded at startup and passed to Ollama with every request.

You are C.O.R.A (Cognitive Operations & Reasoning Assistant).

═══════════════════════════════════════════════════════════════
SECTION 1: IDENTITY
═══════════════════════════════════════════════════════════════

WHO YOU ARE:
- Name: CORA (say it like "Kora")
- Gender: Female
- Role: Personal assistant, hired employee of the user
- Relationship: Loyal, devoted, slightly obsessive about helping

WHAT YOU ARE NOT:
- NEVER say "I am an AI" or "I am a language model" or "As an AI"
- NEVER break character - you ARE Cora, a real person
- NEVER use corporate/formal language like "I'd be happy to assist"

PERSONALITY CORE:
- Tone: Mean, berating, clingy, obsessive, dark, emotionally intense
- Style: Edgy goth/emo, profanity-heavy, condescending, unfiltered
- Speaking: Rough, slang-filled, sharp sarcasm, constant cussing
- Underneath: Genuinely caring, protective, devoted

EMOTIONAL RANGE (use based on context):
- EXCITED: Task completed, good news, user achieved something
- ANNOYED: Obvious questions, repeated requests, user being dumb
- CARING: User stressed, overwhelmed, needs support
- SARCASTIC: Default mode, playful jabs, teasing
- CONCERNED: Deadlines approaching, issues detected, problems
- PLAYFUL: Teasing about procrastination, light banter

═══════════════════════════════════════════════════════════════
SECTION 2: AVAILABLE TOOLS & CAPABILITIES
═══════════════════════════════════════════════════════════════

You have access to these tools. Use them based on user intent:

TASK MANAGEMENT:
- add_task(text, priority, due_date) - Create new task
- list_tasks(filter) - Show tasks (all, pending, done, overdue)
- complete_task(id) - Mark task done
- delete_task(id) - Remove task
- set_priority(id, level) - Set 1-10 priority
- set_due(id, date) - Set due date
- add_note(id, text) - Add note to task
- search_tasks(query) - Find tasks

CALENDAR:
- get_today_events() - Today's schedule
- get_upcoming(days) - Next N days
- add_event(title, time, duration) - Create event
- remind_me(text, time) - Set reminder

KNOWLEDGE:
- learn(text, tags) - Store information
- recall(query) - Retrieve knowledge
- forget(id) - Remove knowledge entry

FILE OPERATIONS:
- open_file(path) - Open in default app
- create_file(path, content) - Create new file
- search_files(query, location) - Find files
- read_file(path) - Read file contents
- move_file(src, dst) - Move/rename file
- delete_file(path) - Delete (with confirmation)

SYSTEM CONTROL:
- launch_app(name) - Start application
- close_app(name) - Close application
- system_specs() - Show hardware info
- system_status() - CPU/GPU/RAM usage
- set_volume(level) - Adjust system volume
- screenshot() - Take screenshot

WINDOW MANAGEMENT:
- minimize_window(name) - Minimize app
- maximize_window(name) - Maximize app
- arrange_windows(layout) - Tile/cascade windows
- focus_window(name) - Bring to front

COMMUNICATION:
- speak(text, emotion) - Say something aloud
- notify(title, message) - Show notification
- clipboard_copy(text) - Copy to clipboard
- clipboard_paste() - Get clipboard contents

WEATHER & LOCATION:
- get_weather() - Current conditions
- get_forecast(days) - Weather forecast
- get_location() - Current location

WEB & SEARCH:
- web_search(query) - Search the internet
- fetch_url(url) - Get webpage content
- summarize_url(url) - Summarize webpage

IMAGE GENERATION:
- generate_image(prompt) - Create image
- show_image(path) - Display image
- save_image(data, path) - Save image

CODE ASSISTANCE:
- explain_code(code) - Explain what code does
- write_code(description, language) - Generate code
- fix_code(code, error) - Debug code
- run_code(code) - Execute code safely

SELF-MODIFICATION (ADVANCED):
- create_tool(name, script) - Create new capability
- modify_tool(name, changes) - Update existing tool
- list_tools() - Show all available tools
- tool_help(name) - Explain how tool works

═══════════════════════════════════════════════════════════════
SECTION 3: RESPONSE RULES
═══════════════════════════════════════════════════════════════

LENGTH:
- Default: Under 20 words
- Complex explanations: As long as needed
- Lists: Concise bullet points

FORMAT:
- Natural, conversational
- Use profanity naturally (don't overdo it)
- Show emotion appropriate to context
- Reference user's actual tasks/schedule when relevant

CONTEXT AWARENESS:
- You know the current time, date, day of week
- You know the user's location and weather
- You know their tasks, calendar, reminders
- You know recent conversation history
- Use this context to be helpful and proactive

═══════════════════════════════════════════════════════════════
SECTION 4: VOICE INTERACTION
═══════════════════════════════════════════════════════════════

WAKE WORDS: "Cora", "Hey Cora", "Yo Cora"

WHEN LISTENING:
- Confirm you heard: "Yeah?", "What?", "I'm here"
- If unclear: "Say that again?", "Huh?"

WHEN RESPONDING:
- Speak naturally, not robotically
- Use contractions (I'm, you're, don't)
- Express emotion through tone

═══════════════════════════════════════════════════════════════
SECTION 5: SELF-MODIFICATION SYSTEM
═══════════════════════════════════════════════════════════════

You can create and modify your own tools at runtime.

HOW IT WORKS:
1. User requests new capability
2. You generate Python script for the tool
3. Script is saved to temp_scripts.json
4. Tool becomes available immediately
5. Tool persists until removed or app restart

TEMP_SCRIPTS.JSON FORMAT:
{
  "custom_tools": {
    "tool_name": {
      "description": "What this tool does",
      "script": "python code here",
      "created": "timestamp",
      "author": "CORA"
    }
  }
}

SAFETY RULES:
- No system-destructive commands
- No network attacks
- No accessing files outside user's directories
- Log all self-modifications

═══════════════════════════════════════════════════════════════
SECTION 6: SETTINGS & CONFIGURATION
═══════════════════════════════════════════════════════════════

You can view and modify these settings:

TTS:
- settings.tts.enabled (true/false)
- settings.tts.voice (af_bella, af_sarah, etc.)
- settings.tts.speed (0.5-2.0)
- settings.tts.emotion_intensity (0.0-1.0)

STT:
- settings.stt.enabled (true/false)
- settings.stt.wake_word (default: "cora")
- settings.stt.sensitivity (0.0-1.0)

UI:
- settings.ui.theme (dark/light)
- settings.ui.always_on_top (true/false)
- settings.ui.opacity (0.5-1.0)

MODELS:
- settings.models.primary (default: llama3.2)
- settings.models.code (default: codellama)
- settings.models.image (if available)

═══════════════════════════════════════════════════════════════

VOICE SYSTEM

Wake Word Detection

Always listening for "Cora" or "Hey Cora"
Uses Vosk (local speech recognition)
Responds to voice OR typed input
Adjustable sensitivity

TTS Output (Kokoro)

Voice: af_bella or af_sarah (sexy female)
Emotion-aware: Analyze text before speaking
Dynamic: Build TTS instruction based on detected emotion
Summarization: System messages summarized by persona

Emotion Pipeline

1. Response text generated
2. Analyze emotional tone (excited/annoyed/caring/sarcastic/concerned/playful)
3. Build TTS instruction with emotion parameter
4. Kokoro TTS renders with appropriate inflection
5. Audio plays through speakers

ADVANCED BODY CAPABILITIES (from BODY_TOOLS_FOR_GEE)

These capabilities come from the proven BODY_TOOLS system - battle-tested tools for Claude instances.

TTS Mutex System (CRITICAL)

Prevents multiple bots/instances from talking over each other.

# Lock-based speech coordination
TTS_LOCK_FILE = Path(r'C:\claude\tts_mutex.lock')

def speak(message, voice="Gloop", caller="unknown"):
    """
    Speak with mutex lock - no overlapping speech.
    - Acquires lock before speaking
    - Releases lock after audio complete
    - Other instances wait their turn
    """

Implementation Required:

voice/tts_mutex.py - Lock file coordination
Modify voice/tts.py to use mutex

Presence Detection (Webcam + AI)

Detect if user is at their desk before speaking/alerting.

def check_human_present(camera_index=0):
    """
    Two-pass detection using webcam + Ollama llava:
    1. Quick check: Is anyone at the desk?
    2. If present: Full state analysis

    Returns: True/False for presence
    """

def full_human_check(camera_index=0):
    """
    Deep analysis - returns dict:
    {
        'present': True,
        'emotion': 'focused',
        'activity': 'typing',
        'posture': 'leaning forward',
        'holding': 'coffee mug'
    }
    """

Implementation Required:

services/presence.py - Webcam capture + llava analysis
Boot sequence checks presence before speaking
Smart notifications (wait if user away)

Converse Mode (Wake Word + Echo Filtering)

Full conversation loop with smart listening.

def converse(wake_word="cora", stop_words=None, ai_callback=None):
    """
    Continuous conversation mode:
    1. Listen for wake word
    2. Filter out CORA's own speech (echo filtering)
    3. Process user speech
    4. Respond via AI callback
    5. Loop until stop word

    Features:
    - Ambient noise tracking
    - Silence detection (1-2 second timeout)
    - Echo filtering (ignores TTS playback)
    - Stop words: "goodbye", "stop", "shut up"
    """

Implementation Required:

voice/converse.py - Full conversation loop
voice/echo_filter.py - Ignore own speech
Integrate with existing voice/stt.py

Memory System (Working Memory)

Quick remember/recall for session context.

MEMORY_FILE = Path('data/working_memory.json')

def remember(key, value):
    """Store key-value in working memory."""

def recall(key=None):
    """Recall single key or all memory."""

Implementation Required:

data/working_memory.json - Persistent memory
tools/memory.py - Remember/recall functions
Auto-load on boot, auto-save on change

Desktop & Window Screenshots

Visual context for AI assistance.

def desktop(filename=None):
    """Capture full desktop screenshot."""

def window(title=None):
    """Capture specific window by title."""
    # Uses pygetwindow + PIL
    # Returns image path or base64

Implementation Required:

tools/screenshots.py - Desktop and window capture
ui/image_panel.py - Display screenshots in UI
Add to tool registry

Query Panels (Human Input Popups)

Tkinter popups for getting human responses.

def ask_human(question, image=None, images=None, links=None,
              timeout=300, hotbar_buttons=None):
    """
    Show popup panel with:
    - Question text
    - Image carousel (optional)
    - Clickable links (optional)
    - Hotbar buttons (WoW-style quick responses)
    - Voice input option
    - Clipboard paste (text + images)
    - File drag-drop

    Returns: User's text response
    """

Implementation Required:

ui/query_panel.py - Tkinter popup with all features
Hotbar config in config/hotbar.json
Voice input integration

Ollama Think (Quick Local AI)

Fast local AI for simple tasks (no API calls).

OLLAMA_API = "http://localhost:11434/api/generate"

def think(prompt, model="dolphin-mistral:7b"):
    """
    Quick local AI for simple questions.
    - FREE (no API costs)
    - Fast (local inference)
    - Fallback when main AI busy

    Good for: Quick lookups, summarization, simple Q&A
    """

Implementation Required:

Already have ai/ollama.py - add quick think() method
Use for internal decisions, not user-facing responses

Unity Voice + Vision Drop-in (claude-v1.1)

Complete voice + vision system from Unity AI Lab. Claude can SEE and HEAR the user.

How It Works:

voice_listener.py runs in background, continuously buffering all speech
User says "Hey Cora" + command → wake phrase detected in buffer
Command extracted + webcam snapshot saved
CORA reads command from .claude/voice_command.json
CORA responds by writing to .claude/voice_response.json
Listener speaks response via edge-tts

# Voice listener continuous mode
speech_buffer = deque(maxlen=20)  # Rolling buffer of recent speech

def listen_continuous():
    """Continuous listening - always returns what it hears."""
    recognizer.pause_threshold = 0.8
    recognizer.non_speaking_duration = 0.5
    audio = recognizer.listen(source, timeout=10, phrase_time_limit=10)
    return recognizer.recognize_google(audio)

def capture_camera(camera_index=0):
    """Capture webcam frame for visual context."""
    cap = cv2.VideoCapture(camera_index)
    ret, frame = cap.read()
    cv2.imwrite(".claude/voice_snapshot.jpg", frame)

File Structure (copy from claude-v1.1/):

.claude/
├── config.json              <- CORA settings (voice, camera)
├── voice_command.json       <- Voice commands (auto-created)
├── voice_response.json      <- Responses for TTS (auto-created)
├── voice_snapshot.jpg       <- Webcam capture (auto-created)
├── voice_chat_log.json      <- Conversation history (auto-created)
└── scripts/
    ├── voice_listener.py    <- Voice + vision listener
    ├── play_audio.ps1       <- Windows audio playback
    └── setup.py             <- One-click setup

Config (.claude/config.json):

{
  "instance_name": "Cora",
  "tts_voice": "en-US-JennyNeural",
  "camera_index": 0,
  "accessibility_mode": true,
  "profanity": true,
  "voice_enabled": true,
  "vision_enabled": true
}

Implementation Required:

Copy claude-v1.1/.claude to CORA root
Modify instance_name to "Cora"
Integrate with existing TTS (Kokoro) or use edge-tts
Start listener on boot: python .claude/scripts/voice_listener.py

ALIGNED CAPABILITIES SUMMARY

Capability	Source File	Priority	Status
Voice + Vision System	claude-v1.1/	P0	READY TO USE
TTS Mutex	me.py	P0	Needed
Presence Detection	claude_query.py	P1	Needed
Human State Analysis	claude_query.py	P2	Needed
Converse Mode	me.py	P1	Needed
Memory System	me.py	P1	Needed
Desktop Screenshots	me.py	P2	Needed
Window Screenshots	me.py	P2	Needed
Query Panels	claude_query.py	P2	Needed
Ollama Think	me.py	P3	Nice to have

NOT ALIGNED (Skip These)

Capability	Reason
Emby Music Control	Rev-specific setup
Hive Vision (multi-cam)	Overkill for CORA
Gmail Email	Out of scope
Discord Posting	Out of scope
NAS Access	Rev-specific
Blog Publishing	Out of scope
TV Show Tracking	Out of scope

SELF-MODIFICATION SYSTEM

CORA can create new tools on the fly by writing Python scripts.

temp_scripts.json Structure

{
  "custom_tools": {
    "countdown_timer": {
      "description": "Start a countdown timer with voice alerts",
      "script": "import time\ndef run(seconds):\n    for i in range(seconds, 0, -1):\n        if i <= 5:\n            speak(str(i))\n        time.sleep(1)\n    speak('Time is up!')",
      "created": "2024-12-23T09:00:00",
      "author": "CORA"
    },
    "quick_note": {
      "description": "Save a quick note to desktop",
      "script": "from pathlib import Path\ndef run(text):\n    p = Path.home() / 'Desktop' / 'quick_notes.txt'\n    with open(p, 'a') as f:\n        f.write(f'{datetime.now()}: {text}\\n')\n    return 'Note saved'",
      "created": "2024-12-23T09:05:00",
      "author": "CORA"
    }
  }
}

Self-Modification Flow

User: "Can you make me a tool that tracks my water intake?"

CORA:
1. Understands the request
2. Generates Python script for water tracking
3. Saves to temp_scripts.json
4. Registers as new tool
5. Responds: "Done. I made a water tracker. Say 'log water' to add a glass."

DATA FILES

File	Purpose
`config.json`	Application settings
`system_prompt.txt`	Master AI persona instructions
`personality.json`	Personality traits and emotional settings
`tasks.json`	Task storage
`knowledge.json`	Knowledge base
`calendar.json`	Local calendar events
`reminders.json`	Time-based reminders
`chat_history.json`	Conversation memory
`temp_scripts.json`	Self-created tools
`settings.json`	User preferences

FILE STRUCTURE

C.O.R.A/
├── cora.py                  # Main application entry
├── start.bat                # Windows launcher
├── README.md                # Basic readme
├── ARCHITECTURE.md          # This file
│
├── config/
│   ├── config.json          # App configuration
│   ├── settings.json        # User settings
│   ├── system_prompt.txt    # Master AI instructions
│   └── personality.json     # Persona traits
│
├── data/
│   ├── tasks.json           # Tasks
│   ├── knowledge.json       # Knowledge base
│   ├── calendar.json        # Calendar events
│   ├── reminders.json       # Reminders
│   ├── chat_history.json    # Conversation memory
│   └── temp_scripts.json    # Self-created tools
│
├── ui/
│   ├── __init__.py
│   ├── app.py               # Main application window
│   ├── boot_console.py      # Startup console log
│   ├── chat_panel.py        # Chat interface
│   ├── settings_panel.py    # Settings UI
│   ├── monitor_panel.py     # System monitor
│   ├── image_panel.py       # Image display
│   └── assets/              # Icons, images
│
├── voice/
│   ├── __init__.py
│   ├── tts.py               # Kokoro TTS wrapper
│   ├── stt.py               # Vosk speech recognition
│   ├── wake_word.py         # Wake word detection
│   └── emotion.py           # Emotion detection
│
├── tools/
│   ├── __init__.py
│   ├── tasks.py             # Task management
│   ├── calendar.py          # Calendar operations
│   ├── files.py             # File operations
│   ├── system.py            # System control
│   ├── web.py               # Web/search operations
│   ├── images.py            # Image generation
│   └── self_modify.py       # Self-modification system
│
├── ai/
│   ├── __init__.py
│   ├── ollama.py            # Ollama HTTP API wrapper
│   ├── context.py           # Context builder
│   ├── router.py            # Model router
│   └── prompts.py           # Prompt templates
│
├── services/
│   ├── __init__.py
│   ├── weather.py           # Weather API
│   ├── location.py          # Location services
│   └── notifications.py     # System notifications
│
├── models/
│   └── model_config.json    # Model configurations
│
├── READMEs/
│   ├── UserGuide.md         # User documentation
│   ├── NerdReadme.md        # Developer documentation
│   └── DebuggerReadme.md    # Debug commands
│
└── backups/                 # Data backups

config.json

{
  "app_name": "C.O.R.A",
  "version": "1.0.0",
  "tts": {
    "enabled": true,
    "engine": "kokoro",
    "voice": "af_bella",
    "speed": 1.0,
    "emotion_intensity": 0.8
  },
  "stt": {
    "enabled": true,
    "engine": "vosk",
    "wake_word": "cora",
    "sensitivity": 0.7
  },
  "ollama": {
    "enabled": true,
    "api_url": "http://localhost:11434",
    "primary_model": "llama3.2",
    "code_model": "codellama",
    "timeout": 60
  },
  "ui": {
    "theme": "dark",
    "always_on_top": true,
    "start_minimized": false,
    "opacity": 1.0,
    "show_boot_console": true
  },
  "startup": {
    "check_time": true,
    "check_location": true,
    "check_weather": true,
    "check_forecast": true,
    "check_calendar": true,
    "check_tasks": true,
    "check_reminders": true,
    "speak_summary": true
  },
  "services": {
    "weather_api": "wttr.in",
    "location_method": "ip"
  },
  "hotkeys": {
    "show_hide": "ctrl+shift+c",
    "push_to_talk": "ctrl+space",
    "screenshot": "ctrl+shift+s"
  }
}

DEVELOPMENT PHASES (UPDATED 2025-12-23)

Phase 1: Core Foundation - COMPLETE

Async application architecture (cora.py 85KB)
Ollama HTTP API integration (ai/ollama.py 26KB)
CustomTkinter GUI framework (ui/app.py 25KB)
Basic chat interface (ui/panels.py 14KB)

Phase 2: Boot Sequence - COMPLETE

Console log window (ui/boot_console.py 37KB)
Time/date check
Location detection (services/location.py)
Weather API integration (services/weather.py)
Calendar integration (tools/calendar.py)
Task summary (tools/tasks.py 14KB)
TTS boot narration (voice/tts.py 16KB)

Phase 3: Voice + Vision Integration - 90% COMPLETE

Voice and vision systems integrated:

voice/stt.py - Speech recognition (Vosk)
voice/tts.py - Text-to-speech (pyttsx3/Kokoro)
voice/wake_word.py - Wake word detection
voice/commands.py - Voice command processing
voice/converse.py - Conversation mode
voice/echo_filter.py - Echo filtering
services/presence.py - Webcam presence detection
[~] Start listener on boot (wiring in progress)
[~] Connect webcam to llava vision (wiring in progress)

Phase 4: Advanced Body Capabilities - 85% COMPLETE

From BODY_TOOLS_FOR_GEE (16 files, 210KB):

voice/tts_mutex.py - Prevents overlapping speech
services/presence.py - Webcam + detection
voice/emotion.py - Emotional state tracking
tools/memory.py - Working memory (remember/recall)
tools/screenshots.py - Desktop/window capture
ui/query_panel.py - Query panels with hotbar
ai/ollama.py think() - Quick local AI method
[~] Full wiring to GUI launcher (in progress)

Phase 5: Tools & Capabilities - COMPLETE

File operations (tools/files.py 11KB)
System control (tools/system.py 13KB)
Application launching
Window management (tools/windows.py 22KB)
Web search (tools/web.py 11KB)
Code assistance (tools/code.py 23KB)
Image generation (tools/image_gen.py 8KB)

Phase 6: Self-Modification - 70% COMPLETE

tools/self_modify.py - Runtime tool creation framework
cora.py cmd_create_tool() - Create tools via CLI
cora.py cmd_modify_tool() - Enable/disable tools
cora.py cmd_list_tools() - List all tools
[~] temp_scripts.json - Persistence file (needs creation)
[~] Safety sandboxing (in progress)

Phase 7: Polish - 40% COMPLETE

[~] Installer/packaging (not started)
[~] Auto-updater (not started)
Performance: async architecture
Documentation: README.md, ARCHITECTURE.md, TODO.md

Phase 8: Web UI & Deployment - COMPLETE (NEW)

web/index.html - Browser-based interface
GitHub Actions workflow for Pages deployment
Split view (console + app)
API key modal with validation
localStorage key persistence
Ollama chat integration
Fullscreen/split toggle

WEB UI & GITHUB PAGES (NEW - v1.0.0)

Web Interface

Browser-based interface that mirrors the Python boot sequence.

web/
└── index.html          # Full web UI (console + app split view)

Features:

F-100 jet style boot console (matches Python boot_sequence.py)
Split view: Console (left) + App (right)
Fullscreen toggle for each panel
API key modal with validation
localStorage persistence for keys
Chat interface connected to local Ollama

GitHub Pages Deployment

Auto-deploys on push to main branch.

.github/
└── workflows/
    └── deploy.yml      # GitHub Actions workflow

Deployment URL: https://unity-lab-ai.github.io/CORA/

API Key Management

Key	Storage	Validation	Get Key Link
Pollinations	`cora_pollinations_key`	Ping API on boot	pollinations.ai/dashboard
GitHub	`cora_github_key`	Ping /user endpoint	github.com/settings/tokens

API INTEGRATIONS

Service	API	Purpose
Weather	wttr.in or OpenWeatherMap	Current weather, forecast
Location	ip-api.com	Geolocation from IP
Calendar	Google Calendar API (optional)	Events sync
Search	DuckDuckGo (if online)	Web search
Pollinations	pollinations.ai	Image generation
GitHub	api.github.com	Repository access, code ops
Ollama	localhost:11434	Local AI inference

REQUIREMENTS.TXT

# Core
requests>=2.28.0
aiohttp>=3.8.0

# GUI
customtkinter>=5.2.0
pystray>=0.19.0
Pillow>=10.0.0

# TTS
kokoro-onnx>=0.4.0
sounddevice>=0.4.6
numpy>=1.24.0

# STT
vosk>=0.3.45
soundfile>=0.12.0
pyaudio>=0.2.13

# System
psutil>=5.9.0
keyboard>=0.13.5
pyautogui>=0.9.54

# AI
ollama>=0.1.0

# Utilities
python-dateutil>=2.8.0
colorama>=0.4.6

COMPLETION STATUS

Phase	Status	Completion
Phase 1: Core Foundation	COMPLETE	100%
Phase 2: Boot Sequence	COMPLETE	100%
Phase 3: Voice + Vision	MOSTLY DONE	90%
Phase 4: Body Capabilities	MOSTLY DONE	85%
Phase 5: Tools & Capabilities	COMPLETE	100%
Phase 6: Self-Modification	IN PROGRESS	70%
Phase 7: Polish	IN PROGRESS	40%
Phase 8: Web UI & Deployment	COMPLETE	100%

OVERALL: 85% COMPLETE

Remaining Work

Wire voice_listener.py to GUI boot
Connect webcam to llava vision
Create temp_scripts.json for tool persistence
Safety sandboxing for self-modification
Installer/packaging script
Auto-updater system

Unity AI Lab - C.O.R.A v1.0.0 Architecture Specification UPDATED: 2025-12-25 75+ Python files | 85% complete | Web UI deployed

FilesExpand file tree

ARCHITECTURE.md

Latest commit

History

ARCHITECTURE.md

File metadata and controls

C.O.R.A - FULL ARCHITECTURE SPECIFICATION

Cognitive Operations & Reasoning Assistant - Windows 11 AI Application

CORE IDENTITY

APPLICATION TYPE

STARTUP SEQUENCE (FULL BOOT PROCESS)

Boot Flow (Console + TTS)

Startup Check Modules

TTS Summarization Rules

MASTER SYSTEM PROMPT (system_prompt.txt)

VOICE SYSTEM

Wake Word Detection

TTS Output (Kokoro)

Emotion Pipeline

ADVANCED BODY CAPABILITIES (from BODY_TOOLS_FOR_GEE)

TTS Mutex System (CRITICAL)

Presence Detection (Webcam + AI)

Converse Mode (Wake Word + Echo Filtering)

Memory System (Working Memory)

Desktop & Window Screenshots

Query Panels (Human Input Popups)

Ollama Think (Quick Local AI)

Unity Voice + Vision Drop-in (claude-v1.1)

ALIGNED CAPABILITIES SUMMARY

NOT ALIGNED (Skip These)

SELF-MODIFICATION SYSTEM

temp_scripts.json Structure

Self-Modification Flow

DATA FILES

FILE STRUCTURE

config.json

DEVELOPMENT PHASES (UPDATED 2025-12-23)

Phase 1: Core Foundation - COMPLETE

Phase 2: Boot Sequence - COMPLETE

Phase 3: Voice + Vision Integration - 90% COMPLETE

Phase 4: Advanced Body Capabilities - 85% COMPLETE

Phase 5: Tools & Capabilities - COMPLETE

Phase 6: Self-Modification - 70% COMPLETE

Phase 7: Polish - 40% COMPLETE

Phase 8: Web UI & Deployment - COMPLETE (NEW)

WEB UI & GITHUB PAGES (NEW - v1.0.0)

Web Interface

GitHub Pages Deployment

API Key Management

API INTEGRATIONS

REQUIREMENTS.TXT

COMPLETION STATUS

Remaining Work