The voices... They're getting louder.

.env.production.example

@@ -0,0 +1,40 @@
+# Production Environment Configuration for Render.com
+# Copy this to .env and fill in your values when deploying
+
+# Gemini API Configuration
+GEMINI_API_KEY=your_gemini_api_key_here
+GEMINI_MODEL=gemini-2.5-flash
+
+# Environment (MUST be "production" for deployment)
+ENVIRONMENT=production
+
+# Service Configuration
+AI_SERVICE_PORT=9000
+AI_SERVICE_HOST=0.0.0.0
+
+# Production URL (your Render.com app URL)
+# Example: https://hpc-simulation-ai.onrender.com
+PRODUCTION_URL=https://your-app-name.onrender.com
+
+# Enrichment Service Integration
+# In production on Render, services communicate internally
+# Leave as localhost since both services run in same container
+ENRICHMENT_SERVICE_URL=http://localhost:8000
+ENRICHMENT_FETCH_LIMIT=10
+
+# Demo Mode (enables caching and consistent responses for demos)
+DEMO_MODE=false
+
+# Fast Mode (use shorter prompts for faster responses)
+FAST_MODE=true
+
+# Strategy Generation Settings
+STRATEGY_COUNT=3  # Number of strategies to generate (3 for testing, 20 for production)
+
+# Performance Settings
+BRAINSTORM_TIMEOUT=90
+ANALYZE_TIMEOUT=120
+GEMINI_MAX_RETRIES=3
+
+# ElevenLabs API Key (optional, for voice features)
+ELEVENLABS_API_KEY=your_elevenlabs_api_key_here

ENVIRONMENT_CONFIG.md

@@ -0,0 +1,249 @@
+# Environment Configuration Guide
+
+## Overview
+
+The HPC Simulation system automatically adapts between **development** and **production** environments based on the `ENVIRONMENT` variable.
+
+## Configuration Files
+
+### Development (Local)
+- File: `.env` (in project root)
+- `ENVIRONMENT=development`
+- Uses `localhost` URLs
+
+### Production (Render.com)
+- Set environment variables in Render dashboard
+- `ENVIRONMENT=production`
+- `PRODUCTION_URL=https://your-app.onrender.com`
+- Automatically adapts all URLs
+
+## Key Environment Variables
+
+### Required for Both Environments
+
+| Variable | Description | Example |
+|----------|-------------|---------|
+| `GEMINI_API_KEY` | Google Gemini API key | `AIzaSy...` |
+| `ENVIRONMENT` | Environment mode | `development` or `production` |
+
+### Production-Specific
+
+| Variable | Description | Example |
+|----------|-------------|---------|
+| `PRODUCTION_URL` | Your Render.com app URL | `https://hpc-ai.onrender.com` |
+
+### Optional
+
+| Variable | Default | Description |
+|----------|---------|-------------|
+| `ELEVENLABS_API_KEY` | - | Voice synthesis (optional) |
+| `GEMINI_MODEL` | `gemini-1.5-pro` | AI model version |
+| `STRATEGY_COUNT` | `3` | Strategies per lap |
+| `FAST_MODE` | `true` | Use shorter prompts |
+
+## How It Works
+
+### URL Auto-Configuration
+
+The `config.py` automatically provides environment-aware URLs:
+
+```python
+settings = get_settings()
+
+# Automatically returns correct URL based on environment:
+settings.base_url          # http://localhost:9000 OR https://your-app.onrender.com
+settings.websocket_url     # ws://localhost:9000 OR wss://your-app.onrender.com
+settings.internal_enrichment_url  # Always http://localhost:8000 (internal)
+```
+
+### Development Environment
+
+```bash
+ENVIRONMENT=development
+# Result:
+# - base_url: http://localhost:9000
+# - websocket_url: ws://localhost:9000
+# - Dashboard connects to: ws://localhost:9000/ws/dashboard
+```
+
+### Production Environment
+
+```bash
+ENVIRONMENT=production
+PRODUCTION_URL=https://hpc-ai.onrender.com
+# Result:
+# - base_url: https://hpc-ai.onrender.com
+# - websocket_url: wss://hpc-ai.onrender.com
+# - Dashboard connects to: wss://hpc-ai.onrender.com/ws/dashboard
+```
+
+## Component-Specific Configuration
+
+### 1. AI Intelligence Layer
+
+**Development:**
+- Binds to: `0.0.0.0:9000`
+- Enrichment client connects to: `http://localhost:8000`
+- Dashboard WebSocket: `ws://localhost:9000/ws/dashboard`
+
+**Production:**
+- Binds to: `0.0.0.0:9000` (Render exposes externally)
+- Enrichment client connects to: `http://localhost:8000` (internal)
+- Dashboard WebSocket: `wss://your-app.onrender.com/ws/dashboard`
+
+### 2. Enrichment Service
+
+**Development:**
+- Binds to: `0.0.0.0:8000`
+- Accessed at: `http://localhost:8000`
+
+**Production:**
+- Binds to: `0.0.0.0:8000` (internal only)
+- Accessed internally at: `http://localhost:8000`
+- Not exposed externally (behind AI layer)
+
+### 3. Dashboard (Frontend)
+
+**Auto-detects environment:**
+```javascript
+// Automatically uses current host and protocol
+const protocol = window.location.protocol === 'https:' ? 'wss:' : 'ws:';
+const host = window.location.host;
+const wsUrl = `${protocol}//${host}/ws/dashboard`;
+```
+
+### 4. Pi Simulator (Client)
+
+**Development:**
+```bash
+python simulate_pi_websocket.py \
+  --ws-url ws://localhost:9000/ws/pi \
+  --enrichment-url http://localhost:8000
+```
+
+**Production:**
+```bash
+python simulate_pi_websocket.py \
+  --ws-url wss://your-app.onrender.com/ws/pi \
+  --enrichment-url https://your-app.onrender.com  # If exposed
+```
+
+## Quick Setup
+
+### Local Development
+
+1. **Create `.env` in project root:**
+```bash
+GEMINI_API_KEY=your_key_here
+ENVIRONMENT=development
+```
+
+2. **Start services:**
+```bash
+./start.sh
+```
+
+### Render.com Production
+
+1. **Set environment variables in Render dashboard:**
+```
+GEMINI_API_KEY=your_key_here
+ENVIRONMENT=production
+PRODUCTION_URL=https://your-app-name.onrender.com
+```
+
+2. **Deploy** - URLs auto-configure!
+
+## Troubleshooting
+
+### Issue: WebSocket connection fails in production
+
+**Check:**
+1. `ENVIRONMENT=production` is set
+2. `PRODUCTION_URL` matches your actual Render URL (including `https://`)
+3. Dashboard uses `wss://` protocol (should auto-detect)
+
+### Issue: Enrichment service unreachable
+
+**In production:**
+- Both services run in same container
+- Internal communication uses `http://localhost:8000`
+- This is automatic, no configuration needed
+
+**In development:**
+- Ensure enrichment service is running: `python scripts/serve.py`
+- Check `http://localhost:8000/health`
+
+### Issue: Pi simulator can't connect
+
+**Development:**
+```bash
+# Test connection
+curl http://localhost:9000/health
+wscat -c ws://localhost:9000/ws/pi
+```
+
+**Production:**
+```bash
+# Test connection
+curl https://your-app.onrender.com/health
+wscat -c wss://your-app.onrender.com/ws/pi
+```
+
+## Environment Variable Priority
+
+1. **Render Environment Variables** (highest priority in production)
+2. **.env file** (development)
+3. **Default values** (in config.py)
+
+## Best Practices
+
+### Development
+- ✅ Use `.env` file
+- ✅ Keep `ENVIRONMENT=development`
+- ✅ Use `localhost` URLs
+- ❌ Don't commit `.env` to git
+
+### Production
+- ✅ Set all variables in Render dashboard
+- ✅ Use `ENVIRONMENT=production`
+- ✅ Set `PRODUCTION_URL` after deployment
+- ✅ Use HTTPS/WSS protocols
+- ❌ Don't hardcode URLs in code
+
+## Example Configurations
+
+### .env (Development)
+```bash
+GEMINI_API_KEY=AIzaSyDK_jxVlJUpzyxuiGcopSFkiqMAUD3-w0I
+GEMINI_MODEL=gemini-2.5-flash
+ENVIRONMENT=development
+ELEVENLABS_API_KEY=your_key_here
+STRATEGY_COUNT=3
+FAST_MODE=true
+```
+
+### Render Environment Variables (Production)
+```bash
+GEMINI_API_KEY=AIzaSyDK_jxVlJUpzyxuiGcopSFkiqMAUD3-w0I
+GEMINI_MODEL=gemini-2.5-flash
+ENVIRONMENT=production
+PRODUCTION_URL=https://hpc-simulation-ai.onrender.com
+ELEVENLABS_API_KEY=your_key_here
+STRATEGY_COUNT=3
+FAST_MODE=true
+```
+
+## Migration Checklist
+
+When deploying to production:
+
+- [ ] Set `ENVIRONMENT=production` in Render
+- [ ] Deploy and get Render URL
+- [ ] Set `PRODUCTION_URL` with your Render URL
+- [ ] Test health endpoint: `https://your-app.onrender.com/health`
+- [ ] Test WebSocket: `wss://your-app.onrender.com/ws/pi`
+- [ ] Open dashboard: `https://your-app.onrender.com/dashboard`
+- [ ] Verify logs show production URLs
+
+Done! The system will automatically use production URLs for all connections.

Procfile

@@ -0,0 +1 @@
+web: ./start.sh

RENDER_DEPLOYMENT.md

@@ -0,0 +1,211 @@
+# Render.com Deployment Guide
+
+## Quick Start
+
+### 1. Render.com Configuration
+
+**Service Type:** Web Service
+
+**Build Command:**
+```bash
+pip install -r requirements.txt
+```
+
+**Start Command (choose one):**
+
+#### Option A: Shell Script (Recommended)
+```bash
+./start.sh
+```
+
+#### Option B: Python Supervisor
+```bash
+python start.py
+```
+
+#### Option C: Direct Command
+```bash
+python scripts/serve.py & python ai_intelligence_layer/main.py
+```
+
+### 2. Environment Variables
+
+Set these in Render.com dashboard:
+
+**Required:**
+```bash
+GEMINI_API_KEY=your_gemini_api_key_here
+ENVIRONMENT=production
+PRODUCTION_URL=https://your-app-name.onrender.com  # Your Render app URL
+```
+
+**Optional:**
+```bash
+ELEVENLABS_API_KEY=your_elevenlabs_api_key_here  # For voice features
+GEMINI_MODEL=gemini-2.5-flash
+STRATEGY_COUNT=3
+FAST_MODE=true
+```
+
+**Auto-configured (no need to set):**
+```bash
+# These are handled automatically by the config system
+AI_SERVICE_PORT=9000
+AI_SERVICE_HOST=0.0.0.0
+ENRICHMENT_SERVICE_URL=http://localhost:8000  # Internal communication
+```
+
+### Important: Production URL
+
+After deploying to Render, you'll get a URL like:
+```
+https://your-app-name.onrender.com
+```
+
+**You MUST set this URL in the environment variables:**
+1. Go to Render dashboard → your service → Environment
+2. Add: `PRODUCTION_URL=https://your-app-name.onrender.com`
+3. The app will automatically use this for WebSocket connections and API URLs
+
+### 3. Health Check
+
+**Health Check Path:** `/health` (on port 9000)
+
+**Health Check Command:**
+```bash
+curl http://localhost:9000/health
+```
+
+### 4. Port Configuration
+
+- **Enrichment Service:** 8000 (internal)
+- **AI Intelligence Layer:** 9000 (external, Render will expose this)
+
+Render will automatically bind to `PORT` environment variable.
+
+### 5. Files Required
+
+- ✅ `start.sh` - Shell startup script
+- ✅ `start.py` - Python startup supervisor
+- ✅ `Procfile` - Render configuration
+- ✅ `requirements.txt` - Python dependencies
+
+### 6. Testing Locally
+
+Test the startup script before deploying:
+
+```bash
+# Make executable
+chmod +x start.sh
+
+# Run locally
+./start.sh
+```
+
+Or with Python supervisor:
+
+```bash
+python start.py
+```
+
+### 7. Deployment Steps
+
+1. **Push to GitHub:**
+   ```bash
+   git add .
+   git commit -m "Add Render deployment configuration"
+   git push
+   ```
+
+2. **Create Render Service:**
+   - Go to [render.com](https://render.com)
+   - New → Web Service
+   - Connect your GitHub repository
+   - Select branch (main)
+
+3. **Configure Service:**
+   - Name: `hpc-simulation-ai`
+   - Environment: `Python 3`
+   - Build Command: `pip install -r requirements.txt`
+   - Start Command: `./start.sh`
+
+4. **Add Environment Variables:**
+   - `GEMINI_API_KEY`
+   - `ELEVENLABS_API_KEY` (optional)
+
+5. **Deploy!** 🚀
+
+### 8. Monitoring
+
+Check logs in Render dashboard for:
+- `📊 Starting Enrichment Service on port 8000...`
+- `🤖 Starting AI Intelligence Layer on port 9000...`
+- `✨ All services running!`
+
+### 9. Connecting Clients
+
+**WebSocket URL:**
+```
+wss://your-app-name.onrender.com/ws/pi
+```
+
+**Enrichment API:**
+```
+https://your-app-name.onrender.com/ingest/telemetry
+```
+
+### 10. Troubleshooting
+
+**Services won't start:**
+- Check environment variables are set
+- Verify `start.sh` is executable: `chmod +x start.sh`
+- Check build logs for dependency issues
+
+**Port conflicts:**
+- Render will set `PORT` automatically (9000 by default)
+- Services bind to `0.0.0.0` for external access
+
+**Memory issues:**
+- Consider Render's paid plans for more resources
+- Free tier may struggle with AI model loading
+
+## Architecture on Render
+
+```
+┌─────────────────────────────────────┐
+│      Render.com Container           │
+├─────────────────────────────────────┤
+│                                     │
+│  ┌───────────────────────────────┐ │
+│  │   start.sh / start.py         │ │
+│  └──────────┬────────────────────┘ │
+│             │                       │
+│    ┌────────┴─────────┐            │
+│    │                  │            │
+│    ▼                  ▼            │
+│  ┌──────────┐   ┌──────────────┐  │
+│  │Enrichment│   │AI Intelligence│  │
+│  │Service   │   │Layer          │  │
+│  │:8000     │◄──│:9000          │  │
+│  └──────────┘   └──────┬────────┘  │
+│                        │            │
+└────────────────────────┼────────────┘
+                         │
+                    Internet
+                         │
+                    ┌────▼─────┐
+                    │ Client   │
+                    │(Pi/Web)  │
+                    └──────────┘
+```
+
+## Next Steps
+
+1. Test locally with `./start.sh`
+2. Commit and push to GitHub
+3. Create Render service
+4. Configure environment variables
+5. Deploy and monitor logs
+6. Update client connection URLs
+
+Good luck! 🎉

ai_intelligence_layer/config.py

@@ -1,6 +1,8 @@
 """
 Configuration management for AI Intelligence Layer.
 Uses pydantic-settings for environment variable validation.
+Environment variables are loaded via load_dotenv() in main.py.
+Automatically adapts URLs for development vs production environments.
 """
 from pydantic_settings import BaseSettings, SettingsConfigDict
 from typing import Optional
@@ -9,6 +11,10 @@ from typing import Optional
 class Settings(BaseSettings):
     """Application settings loaded from environment variables."""
     
+    # Environment Configuration
+    environment: str = "development"  # "development" or "production"
+    production_url: Optional[str] = None  # e.g., "https://your-app.onrender.com"
+    
     # Gemini API Configuration
     gemini_api_key: str
     gemini_model: str = "gemini-1.5-pro"
@@ -28,7 +34,7 @@ class Settings(BaseSettings):
     fast_mode: bool = True
     
     # Strategy Generation Settings
-    strategy_count: int = 3  # Number of strategies to generate (3 for fast testing)
+    strategy_count: int = 3  # Number of strategies to generate (3 for testing, 20 for production)
     
     # Performance Settings
     brainstorm_timeout: int = 30
@@ -36,11 +42,37 @@ class Settings(BaseSettings):
     gemini_max_retries: int = 3
     
     model_config = SettingsConfigDict(
-        env_file=".env",
-        env_file_encoding="utf-8",
         case_sensitive=False,
         extra="ignore"
     )
+    
+    @property
+    def is_production(self) -> bool:
+        """Check if running in production environment."""
+        return self.environment.lower() == "production"
+    
+    @property
+    def base_url(self) -> str:
+        """Get the base URL for the application."""
+        if self.is_production and self.production_url:
+            return self.production_url
+        return f"http://localhost:{self.ai_service_port}"
+    
+    @property
+    def websocket_url(self) -> str:
+        """Get the WebSocket URL for the application."""
+        if self.is_production and self.production_url:
+            # Replace https:// with wss:// or http:// with ws://
+            return self.production_url.replace("https://", "wss://").replace("http://", "ws://")
+        return f"ws://localhost:{self.ai_service_port}"
+    
+    @property
+    def internal_enrichment_url(self) -> str:
+        """Get the enrichment service URL (internal on Render)."""
+        if self.is_production:
+            # On Render, services communicate internally via localhost
+            return "http://localhost:8000"
+        return self.enrichment_service_url
 
 
 # Global settings instance

ai_intelligence_layer/main.py

@@ -15,6 +15,10 @@ import random
 from typing import Dict, Any, List
 from datetime import datetime
 import json
+from dotenv import load_dotenv
+
+# Load environment variables from .env file in project root
+load_dotenv()
 
 from config import get_settings
 from models.input_models import (
@@ -445,14 +449,12 @@ async def websocket_pi_endpoint(websocket: WebSocket):
                             logger.info(f"LAP {lap_number} - GENERATING STRATEGY")
                             logger.info(f"{'='*60}")
                             
-                            # Send immediate acknowledgment while processing
+                            # Send SILENT acknowledgment to prevent timeout (no control update)
-                            # Use last known control values instead of resetting to neutral
+                            # This tells the Pi "we're working on it" without triggering voice/controls
                             await websocket.send_json({
-                                "type": "control_command",
+                                "type": "acknowledgment",
                                 "lap": lap_number,
-                                "brake_bias": last_control_command["brake_bias"],
+                                "message": "Processing strategies, please wait..."
-                                "differential_slip": last_control_command["differential_slip"],
-                                "message": "Processing strategies (maintaining previous settings)..."
                             })
                             
                             # Generate strategies (this is the slow part)
@@ -500,6 +502,7 @@ async def websocket_pi_endpoint(websocket: WebSocket):
                                     "brake_bias": control_command["brake_bias"],
                                     "differential_slip": control_command["differential_slip"],
                                     "strategy_name": top_strategy.strategy_name if top_strategy else "N/A",
+                                    "risk_level": top_strategy.risk_level if top_strategy else "medium",
                                     "total_strategies": len(response.strategies),
                                     "reasoning": control_command.get("reasoning", "")
                                 })

ai_intelligence_layer/services/telemetry_client.py

@@ -16,7 +16,8 @@ class TelemetryClient:
     def __init__(self):
         """Initialize telemetry client."""
         settings = get_settings()
-        self.base_url = settings.enrichment_service_url
+        # Use internal_enrichment_url which adapts for production
+        self.base_url = settings.internal_enrichment_url
         self.fetch_limit = settings.enrichment_fetch_limit
         logger.info(f"Telemetry client initialized for {self.base_url}")
     

ai_intelligence_layer/static/dashboard.html

@@ -683,7 +683,13 @@
         }
         
         function connect() {
-            ws = new WebSocket('ws://localhost:9000/ws/dashboard');
+            // Dynamically determine WebSocket URL based on current location
+            const protocol = window.location.protocol === 'https:' ? 'wss:' : 'ws:';
+            const host = window.location.host;
+            const wsUrl = `${protocol}//${host}/ws/dashboard`;
+            
+            console.log(`Connecting to WebSocket: ${wsUrl}`);
+            ws = new WebSocket(wsUrl);
             
             ws.onopen = () => {
                 console.log('Dashboard WebSocket connected');

scripts/simulate_pi_websocket.py

@@ -6,9 +6,10 @@ Connects to AI Intelligence Layer via WebSocket and:
 1. Streams lap telemetry to AI layer
 2. Receives control commands (brake_bias, differential_slip) from AI layer
 3. Applies control adjustments in real-time
+4. Generates voice announcements for strategy updates
 
 Usage:
-    python simulate_pi_websocket.py --interval 5 --ws-url ws://localhost:9000/ws/pi
+    python simulate_pi_websocket.py --interval 5 --ws-url ws://localhost:9000/ws/pi --enable-voice
 """
 from __future__ import annotations
 
@@ -19,6 +20,8 @@ import logging
 from pathlib import Path
 from typing import Dict, Any, Optional
 import sys
+import os
+from datetime import datetime
 
 try:
     import pandas as pd
@@ -29,6 +32,19 @@ except ImportError:
     print("Run: pip install pandas websockets")
     sys.exit(1)
 
+# Optional voice support
+try:
+    from elevenlabs.client import ElevenLabs
+    from elevenlabs import save
+    from dotenv import load_dotenv
+    # Load .env from root directory (default behavior)
+    load_dotenv()
+    VOICE_AVAILABLE = True
+except ImportError:
+    VOICE_AVAILABLE = False
+    print("Note: elevenlabs not installed. Voice features disabled.")
+    print("To enable voice: pip install elevenlabs python-dotenv")
+
 # Configure logging
 logging.basicConfig(
     level=logging.INFO,
@@ -37,10 +53,260 @@ logging.basicConfig(
 logger = logging.getLogger(__name__)
 
 
-class PiSimulator:
+class VoiceAnnouncer:
-    """WebSocket-based Pi simulator with control feedback."""
+    """ElevenLabs text-to-speech announcer for race engineer communications."""
     
-    def __init__(self, csv_path: Path, ws_url: str, interval: float = 60.0, enrichment_url: str = "http://localhost:8000"):
+    def __init__(self, enabled: bool = True):
+        """Initialize ElevenLabs voice engine if available."""
+        self.enabled = enabled and VOICE_AVAILABLE
+        self.client = None
+        self.audio_dir = Path("data/audio")
+        # Use exact same voice as voice_service.py
+        self.voice_id = "mbBupyLcEivjpxh8Brkf"  # Rachel voice
+        
+        if self.enabled:
+            try:
+                api_key = os.getenv("ELEVENLABS_API_KEY")
+                if not api_key:
+                    logger.warning("⚠ ELEVENLABS_API_KEY not found in environment")
+                    self.enabled = False
+                    return
+                
+                self.client = ElevenLabs(api_key=api_key)
+                self.audio_dir.mkdir(parents=True, exist_ok=True)
+                logger.info("✓ Voice announcer initialized (ElevenLabs)")
+            except Exception as e:
+                logger.warning(f"⚠ Voice engine initialization failed: {e}")
+                self.enabled = False
+    
+    def _format_strategy_message(self, data: Dict[str, Any]) -> str:
+        """
+        Format strategy update into natural race engineer speech.
+        
+        Args:
+            data: Control command update from AI layer
+            
+        Returns:
+            Formatted message string
+        """
+        lap = data.get('lap', 0)
+        strategy_name = data.get('strategy_name', 'Unknown')
+        brake_bias = data.get('brake_bias', 5)
+        diff_slip = data.get('differential_slip', 5)
+        reasoning = data.get('reasoning', '')
+        risk_level = data.get('risk_level', '')
+        
+        # Build natural message
+        parts = []
+        
+        # Opening with lap number
+        parts.append(f"Lap {lap}.")
+        
+        # Strategy announcement with risk level
+        if strategy_name and strategy_name != "N/A":
+            # Simplify strategy name for speech
+            clean_strategy = strategy_name.replace('-', ' ').replace('_', ' ')
+            if risk_level:
+                parts.append(f"Running {clean_strategy} strategy, {risk_level} risk.")
+            else:
+                parts.append(f"Running {clean_strategy} strategy.")
+        
+        # Control adjustments with specific values
+        control_messages = []
+        
+        # Brake bias announcement with context
+        if brake_bias < 4:
+            control_messages.append(f"Brake bias set to {brake_bias}, forward biased for sharper turn in response")
+        elif brake_bias == 4:
+            control_messages.append(f"Brake bias {brake_bias}, slightly forward to help rotation")
+        elif brake_bias > 6:
+            control_messages.append(f"Brake bias set to {brake_bias}, rearward to protect front tire wear")
+        elif brake_bias == 6:
+            control_messages.append(f"Brake bias {brake_bias}, slightly rear for front tire management")
+        else:
+            control_messages.append(f"Brake bias neutral at {brake_bias}")
+        
+        # Differential slip announcement with context
+        if diff_slip < 4:
+            control_messages.append(f"Differential at {diff_slip}, tightened for better rotation through corners")
+        elif diff_slip == 4:
+            control_messages.append(f"Differential {diff_slip}, slightly tight for rotation")
+        elif diff_slip > 6:
+            control_messages.append(f"Differential set to {diff_slip}, loosened to reduce rear tire degradation")
+        elif diff_slip == 6:
+            control_messages.append(f"Differential {diff_slip}, slightly loose for tire preservation")
+        else:
+            control_messages.append(f"Differential neutral at {diff_slip}")
+        
+        if control_messages:
+            parts.append(". ".join(control_messages) + ".")
+        
+        # Key reasoning excerpt (first sentence only)
+        if reasoning:
+            # Extract first meaningful sentence
+            sentences = reasoning.split('.')
+            if sentences:
+                key_reason = sentences[0].strip()
+                if len(key_reason) > 20 and len(key_reason) < 150:  # Slightly longer for more context
+                    parts.append(key_reason + ".")
+        
+        return " ".join(parts)
+    
+    def _format_control_message(self, data: Dict[str, Any]) -> str:
+        """
+        Format control command into brief message.
+        
+        Args:
+            data: Control command from AI layer
+            
+        Returns:
+            Formatted message string
+        """
+        lap = data.get('lap', 0)
+        brake_bias = data.get('brake_bias', 5)
+        diff_slip = data.get('differential_slip', 5)
+        message = data.get('message', '')
+        
+        # For early laps or non-strategy updates
+        if message and "Collecting data" in message:
+            return f"Lap {lap}. Collecting baseline data."
+        
+        if brake_bias == 5 and diff_slip == 5:
+            return f"Lap {lap}. Maintaining neutral settings."
+        
+        return f"Lap {lap}. Controls adjusted."
+    
+    async def announce_strategy(self, data: Dict[str, Any]):
+        """
+        Announce strategy update with ElevenLabs voice synthesis.
+        
+        Args:
+            data: Control command update from AI layer
+        """
+        if not self.enabled:
+            return
+        
+        try:
+            # Format message
+            message = self._format_strategy_message(data)
+            
+            logger.info(f"[VOICE] Announcing: {message}")
+            
+            # Generate unique audio filename
+            lap = data.get('lap', 0)
+            timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
+            audio_path = self.audio_dir / f"lap_{lap}_{timestamp}.mp3"
+            
+            # Synthesize with ElevenLabs (exact same settings as voice_service.py)
+            def synthesize():
+                try:
+                    audio = self.client.text_to_speech.convert(
+                        voice_id=self.voice_id,
+                        text=message,
+                        model_id="eleven_multilingual_v2",  # Fast, low-latency model
+                        voice_settings={
+                            "stability": 0.4,
+                            "similarity_boost": 0.95,
+                            "style": 0.7,
+                            "use_speaker_boost": True
+                        }
+                    )
+                    save(audio, str(audio_path))
+                    logger.info(f"[VOICE] Saved to {audio_path}")
+                    
+                    # Play the audio
+                    if sys.platform == "darwin":  # macOS
+                        os.system(f"afplay {audio_path}")
+                    elif sys.platform == "linux":
+                        os.system(f"mpg123 {audio_path} || ffplay -nodisp -autoexit {audio_path}")
+                    elif sys.platform == "win32":
+                        os.system(f"start {audio_path}")
+                    
+                    # Clean up audio file after playing
+                    try:
+                        if audio_path.exists():
+                            audio_path.unlink()
+                            logger.info(f"[VOICE] Cleaned up {audio_path}")
+                    except Exception as cleanup_error:
+                        logger.warning(f"[VOICE] Failed to delete audio file: {cleanup_error}")
+                except Exception as e:
+                    logger.error(f"[VOICE] Synthesis error: {e}")
+            
+            # Run in separate thread to avoid blocking
+            loop = asyncio.get_event_loop()
+            await loop.run_in_executor(None, synthesize)
+            
+        except Exception as e:
+            logger.error(f"[VOICE] Announcement failed: {e}")
+    
+    async def announce_control(self, data: Dict[str, Any]):
+        """
+        Announce control command with ElevenLabs voice synthesis (brief version).
+        
+        Args:
+            data: Control command from AI layer
+        """
+        if not self.enabled:
+            return
+        
+        try:
+            # Format message
+            message = self._format_control_message(data)
+            
+            logger.info(f"[VOICE] Announcing: {message}")
+            
+            # Generate unique audio filename
+            lap = data.get('lap', 0)
+            timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
+            audio_path = self.audio_dir / f"lap_{lap}_control_{timestamp}.mp3"
+            
+            # Synthesize with ElevenLabs (exact same settings as voice_service.py)
+            def synthesize():
+                try:
+                    audio = self.client.text_to_speech.convert(
+                        voice_id=self.voice_id,
+                        text=message,
+                        model_id="eleven_multilingual_v2",  # Fast, low-latency model
+                        voice_settings={
+                            "stability": 0.4,
+                            "similarity_boost": 0.95,
+                            "style": 0.7,
+                            "use_speaker_boost": True
+                        }
+                    )
+                    save(audio, str(audio_path))
+                    logger.info(f"[VOICE] Saved to {audio_path}")
+                    
+                    # Play the audio
+                    if sys.platform == "darwin":  # macOS
+                        os.system(f"afplay {audio_path}")
+                    elif sys.platform == "linux":
+                        os.system(f"mpg123 {audio_path} || ffplay -nodisp -autoexit {audio_path}")
+                    elif sys.platform == "win32":
+                        os.system(f"start {audio_path}")
+                    
+                    # Clean up audio file after playing
+                    try:
+                        if audio_path.exists():
+                            audio_path.unlink()
+                            logger.info(f"[VOICE] Cleaned up {audio_path}")
+                    except Exception as cleanup_error:
+                        logger.warning(f"[VOICE] Failed to delete audio file: {cleanup_error}")
+                except Exception as e:
+                    logger.error(f"[VOICE] Synthesis error: {e}")
+            
+            # Run in separate thread to avoid blocking
+            loop = asyncio.get_event_loop()
+            await loop.run_in_executor(None, synthesize)
+            
+        except Exception as e:
+            logger.error(f"[VOICE] Announcement failed: {e}")
+
+
+class PiSimulator:
+    """WebSocket-based Pi simulator with control feedback and voice announcements."""
+    
+    def __init__(self, csv_path: Path, ws_url: str, interval: float = 60.0, enrichment_url: str = "http://localhost:8000", voice_enabled: bool = False):
         self.csv_path = csv_path
         self.ws_url = ws_url
         self.enrichment_url = enrichment_url
@@ -50,6 +316,12 @@ class PiSimulator:
             "brake_bias": 5,
             "differential_slip": 5
         }
+        self.previous_controls = {
+            "brake_bias": 5,
+            "differential_slip": 5
+        }
+        self.current_risk_level: Optional[str] = None
+        self.voice_announcer = VoiceAnnouncer(enabled=voice_enabled)
     
     def load_lap_csv(self) -> pd.DataFrame:
         """Load lap-level CSV data."""
@@ -245,12 +517,73 @@ class PiSimulator:
                         response = await asyncio.wait_for(websocket.recv(), timeout=5.0)
                         response_data = json.loads(response)
                         
-                        if response_data.get("type") == "control_command":
+                        # Handle silent acknowledgment (no control update, no voice)
+                        if response_data.get("type") == "acknowledgment":
+                            message = response_data.get("message", "")
+                            logger.info(f"[ACK] {message}")
+                            
+                            # Now wait for the actual control command update
+                            try:
+                                update = await asyncio.wait_for(websocket.recv(), timeout=45.0)
+                                update_data = json.loads(update)
+                                
+                                if update_data.get("type") == "control_command_update":
+                                    brake_bias = update_data.get("brake_bias", 5)
+                                    diff_slip = update_data.get("differential_slip", 5)
+                                    strategy_name = update_data.get("strategy_name", "N/A")
+                                    risk_level = update_data.get("risk_level", "medium")
+                                    reasoning = update_data.get("reasoning", "")
+                                    
+                                    # Check if controls changed from previous
+                                    controls_changed = (
+                                        self.current_controls["brake_bias"] != brake_bias or
+                                        self.current_controls["differential_slip"] != diff_slip
+                                    )
+                                    
+                                    # Check if risk level changed
+                                    risk_level_changed = (
+                                        self.current_risk_level is not None and
+                                        self.current_risk_level != risk_level
+                                    )
+                                    
+                                    self.previous_controls = self.current_controls.copy()
+                                    self.current_controls["brake_bias"] = brake_bias
+                                    self.current_controls["differential_slip"] = diff_slip
+                                    self.current_risk_level = risk_level
+                                    
+                                    logger.info(f"[UPDATED] Strategy-Based Control:")
+                                    logger.info(f"  ├─ Brake Bias: {brake_bias}/10")
+                                    logger.info(f"  ├─ Differential Slip: {diff_slip}/10")
+                                    logger.info(f"  ├─ Strategy: {strategy_name}")
+                                    logger.info(f"  ├─ Risk Level: {risk_level}")
+                                    if reasoning:
+                                        logger.info(f"  └─ Reasoning: {reasoning[:100]}...")
+                                    
+                                    self.apply_controls(brake_bias, diff_slip)
+                                    
+                                    # Voice announcement if controls OR risk level changed
+                                    if controls_changed or risk_level_changed:
+                                        if risk_level_changed and not controls_changed:
+                                            logger.info(f"[VOICE] Risk level changed to {risk_level}")
+                                        await self.voice_announcer.announce_strategy(update_data)
+                                    else:
+                                        logger.info(f"[VOICE] Skipping announcement - controls and risk level unchanged")
+                            except asyncio.TimeoutError:
+                                logger.warning("[TIMEOUT] Strategy generation took too long")
+                        
+                        elif response_data.get("type") == "control_command":
                             brake_bias = response_data.get("brake_bias", 5)
                             diff_slip = response_data.get("differential_slip", 5)
                             strategy_name = response_data.get("strategy_name", "N/A")
                             message = response_data.get("message")
                             
+                            # Store previous values before updating
+                            controls_changed = (
+                                self.current_controls["brake_bias"] != brake_bias or
+                                self.current_controls["differential_slip"] != diff_slip
+                            )
+                            
+                            self.previous_controls = self.current_controls.copy()
                             self.current_controls["brake_bias"] = brake_bias
                             self.current_controls["differential_slip"] = diff_slip
                             
@@ -265,6 +598,10 @@ class PiSimulator:
                             # Apply controls (in real Pi, this would adjust hardware)
                             self.apply_controls(brake_bias, diff_slip)
                             
+                            # Voice announcement ONLY if controls changed
+                            if controls_changed:
+                                await self.voice_announcer.announce_control(response_data)
+                            
                             # If message indicates processing, wait for update
                             if message and "Processing" in message:
                                 logger.info("  AI is generating strategies, waiting for update...")
@@ -276,16 +613,43 @@ class PiSimulator:
                                         brake_bias = update_data.get("brake_bias", 5)
                                         diff_slip = update_data.get("differential_slip", 5)
                                         strategy_name = update_data.get("strategy_name", "N/A")
+                                        risk_level = update_data.get("risk_level", "medium")
+                                        reasoning = update_data.get("reasoning", "")
                                         
+                                        # Check if controls changed from previous
+                                        controls_changed = (
+                                            self.current_controls["brake_bias"] != brake_bias or
+                                            self.current_controls["differential_slip"] != diff_slip
+                                        )
+                                        
+                                        # Check if risk level changed
+                                        risk_level_changed = (
+                                            self.current_risk_level is not None and
+                                            self.current_risk_level != risk_level
+                                        )
+                                        
+                                        self.previous_controls = self.current_controls.copy()
                                         self.current_controls["brake_bias"] = brake_bias
                                         self.current_controls["differential_slip"] = diff_slip
+                                        self.current_risk_level = risk_level
                                         
                                         logger.info(f"[UPDATED] Strategy-Based Control:")
                                         logger.info(f"  ├─ Brake Bias: {brake_bias}/10")
                                         logger.info(f"  ├─ Differential Slip: {diff_slip}/10")
-                                        logger.info(f"  └─ Strategy: {strategy_name}")
+                                        logger.info(f"  ├─ Strategy: {strategy_name}")
+                                        logger.info(f"  ├─ Risk Level: {risk_level}")
+                                        if reasoning:
+                                            logger.info(f"  └─ Reasoning: {reasoning[:100]}...")
                                         
                                         self.apply_controls(brake_bias, diff_slip)
+                                        
+                                        # Voice announcement if controls OR risk level changed
+                                        if controls_changed or risk_level_changed:
+                                            if risk_level_changed and not controls_changed:
+                                                logger.info(f"[VOICE] Risk level changed to {risk_level}")
+                                            await self.voice_announcer.announce_strategy(update_data)
+                                        else:
+                                            logger.info(f"[VOICE] Skipping announcement - controls and risk level unchanged")
                                 except asyncio.TimeoutError:
                                     logger.warning("[TIMEOUT] Strategy generation took too long")
                         
@@ -344,7 +708,7 @@ class PiSimulator:
 
 async def main():
     parser = argparse.ArgumentParser(
-        description="WebSocket-based Raspberry Pi Telemetry Simulator"
+        description="WebSocket-based Raspberry Pi Telemetry Simulator with Voice Announcements"
     )
     parser.add_argument(
         "--interval",
@@ -370,6 +734,11 @@ async def main():
         default=None,
         help="Path to lap CSV file (default: scripts/ALONSO_2023_MONZA_LAPS.csv)"
     )
+    parser.add_argument(
+        "--enable-voice",
+        action="store_true",
+        help="Enable voice announcements for strategy updates (requires elevenlabs and ELEVENLABS_API_KEY)"
+    )
     
     args = parser.parse_args()
     
@@ -389,7 +758,8 @@ async def main():
         csv_path=csv_path,
         ws_url=args.ws_url,
         enrichment_url=args.enrichment_url,
-        interval=args.interval
+        interval=args.interval,
+        voice_enabled=args.enable_voice
     )
     
     logger.info("Starting WebSocket Pi Simulator")
@@ -397,6 +767,7 @@ async def main():
     logger.info(f"Enrichment Service: {args.enrichment_url}")
     logger.info(f"WebSocket URL: {args.ws_url}")
     logger.info(f"Interval: {args.interval}s per lap")
+    logger.info(f"Voice Announcements: {'Enabled' if args.enable_voice and VOICE_AVAILABLE else 'Disabled'}")
     logger.info("-" * 60)
     
     await simulator.stream_telemetry()

scripts/test_voice.py

@@ -0,0 +1,76 @@
+#!/usr/bin/env python3
+"""
+Quick test script for ElevenLabs voice announcements.
+"""
+
+import sys
+import os
+from pathlib import Path
+
+sys.path.insert(0, '.')
+
+try:
+    from elevenlabs.client import ElevenLabs
+    from elevenlabs import save
+    from dotenv import load_dotenv
+    
+    load_dotenv()
+    
+    # Check API key
+    api_key = os.getenv("ELEVENLABS_API_KEY")
+    if not api_key:
+        print("✗ ELEVENLABS_API_KEY not found in environment")
+        print("Create a .env file with: ELEVENLABS_API_KEY=your_key_here")
+        sys.exit(1)
+    
+    # Initialize client with same settings as voice_service.py
+    client = ElevenLabs(api_key=api_key)
+    voice_id = "mbBupyLcEivjpxh8Brkf"  # Rachel voice
+    
+    # Test message
+    test_message = "Lap 3. Strategy: Conservative One Stop. Brake bias forward for turn in. Current tire degradation suggests extended first stint."
+    
+    print(f"Testing ElevenLabs voice announcement...")
+    print(f"Voice ID: {voice_id} (Rachel)")
+    print(f"Message: {test_message}")
+    print("-" * 60)
+    
+    # Synthesize
+    audio = client.text_to_speech.convert(
+        voice_id=voice_id,
+        text=test_message,
+        model_id="eleven_multilingual_v2",
+        voice_settings={
+            "stability": 0.4,
+            "similarity_boost": 0.95,
+            "style": 0.7,
+            "use_speaker_boost": True
+        }
+    )
+    
+    # Save audio
+    output_dir = Path("data/audio")
+    output_dir.mkdir(parents=True, exist_ok=True)
+    output_path = output_dir / "test_voice.mp3"
+    
+    save(audio, str(output_path))
+    print(f"✓ Audio saved to: {output_path}")
+    
+    # Play audio
+    print("✓ Playing audio...")
+    if sys.platform == "darwin":  # macOS
+        os.system(f"afplay {output_path}")
+    elif sys.platform == "linux":
+        os.system(f"mpg123 {output_path} || ffplay -nodisp -autoexit {output_path}")
+    elif sys.platform == "win32":
+        os.system(f"start {output_path}")
+    
+    print("✓ Voice test completed successfully!")
+    
+except ImportError as e:
+    print(f"✗ elevenlabs not available: {e}")
+    print("Install with: pip install elevenlabs python-dotenv")
+except Exception as e:
+    print(f"✗ Voice test failed: {e}")
+    import traceback
+    traceback.print_exc()

scripts/voice_service.py

@@ -13,13 +13,8 @@ from dotenv import load_dotenv
 load_dotenv()
 
 class RaceEngineerVoice:
-    def __init__(self, voice_id: str = "mbBupyLcEivjpxh8Brkf"):  # Default: Rachel
+    def __init__(self, voice_id: str = "mbBupyLcEivjpxh8Brkf"):
-        """
-        Initialize ElevenLabs voice service.
         
-        Args:
-            voice_id: ElevenLabs voice ID (Rachel is default, professional female voice)
-        """
         self.client = ElevenLabs(api_key=os.getenv("ELEVENLABS_API_KEY"))
         self.voice_id = voice_id
         

start.py

@@ -0,0 +1,99 @@
+#!/usr/bin/env python3
+"""
+Startup supervisor for HPC Simulation Services.
+Manages both enrichment service and AI intelligence layer.
+"""
+import subprocess
+import sys
+import time
+import signal
+import os
+
+processes = []
+
+def cleanup(signum=None, frame=None):
+    """Clean up all child processes."""
+    print("\n🛑 Shutting down all services...")
+    for proc in processes:
+        try:
+            proc.terminate()
+            proc.wait(timeout=5)
+        except subprocess.TimeoutExpired:
+            proc.kill()
+        except Exception as e:
+            print(f"Error stopping process: {e}")
+    sys.exit(0)
+
+# Register signal handlers
+signal.signal(signal.SIGINT, cleanup)
+signal.signal(signal.SIGTERM, cleanup)
+
+def main():
+    print("🚀 Starting HPC Simulation Services...")
+    
+    # Start enrichment service
+    print("📊 Starting Enrichment Service on port 8000...")
+    enrichment_proc = subprocess.Popen(
+        [sys.executable, "scripts/serve.py"],
+        stdout=subprocess.PIPE,
+        stderr=subprocess.STDOUT,
+        text=True,
+        bufsize=1
+    )
+    processes.append(enrichment_proc)
+    print(f"   ├─ PID: {enrichment_proc.pid}")
+    
+    # Give it time to start
+    time.sleep(5)
+    
+    # Check if still running
+    if enrichment_proc.poll() is not None:
+        print("❌ Enrichment service failed to start")
+        cleanup()
+        return 1
+    print("   └─ ✅ Enrichment service started successfully")
+    
+    # Start AI Intelligence Layer
+    print("🤖 Starting AI Intelligence Layer on port 9000...")
+    ai_proc = subprocess.Popen(
+        [sys.executable, "ai_intelligence_layer/main.py"],
+        stdout=subprocess.PIPE,
+        stderr=subprocess.STDOUT,
+        text=True,
+        bufsize=1
+    )
+    processes.append(ai_proc)
+    print(f"   ├─ PID: {ai_proc.pid}")
+    
+    # Give it time to start
+    time.sleep(3)
+    
+    # Check if still running
+    if ai_proc.poll() is not None:
+        print("❌ AI Intelligence Layer failed to start")
+        cleanup()
+        return 1
+    print("   └─ ✅ AI Intelligence Layer started successfully")
+    
+    print("\n✨ All services running!")
+    print("   📊 Enrichment Service: http://0.0.0.0:8000")
+    print("   🤖 AI Intelligence Layer: ws://0.0.0.0:9000/ws/pi")
+    print("\nPress Ctrl+C to stop all services\n")
+    
+    # Monitor processes
+    try:
+        while True:
+            # Check if any process has died
+            for proc in processes:
+                if proc.poll() is not None:
+                    print(f"⚠️  Process {proc.pid} died unexpectedly!")
+                    cleanup()
+                    return 1
+            time.sleep(1)
+    except KeyboardInterrupt:
+        cleanup()
+    
+    return 0
+
+if __name__ == "__main__":
+    sys.exit(main())

start.sh

@@ -0,0 +1,58 @@
+#!/bin/bash
+# Startup script for Render.com deployment
+# Starts both the enrichment service and AI intelligence layer
+
+set -e  # Exit on error
+
+echo "🚀 Starting HPC Simulation Services..."
+
+# Trap to handle cleanup on exit
+cleanup() {
+    echo "🛑 Shutting down services..."
+    kill $ENRICHMENT_PID $AI_PID 2>/dev/null || true
+    exit
+}
+trap cleanup SIGINT SIGTERM
+
+# Start enrichment service in background
+echo "📊 Starting Enrichment Service on port 8000..."
+python scripts/serve.py &
+ENRICHMENT_PID=$!
+echo "   ├─ PID: $ENRICHMENT_PID"
+
+# Give enrichment service time to start
+sleep 5
+
+# Check if enrichment service is still running
+if ! kill -0 $ENRICHMENT_PID 2>/dev/null; then
+    echo "❌ Enrichment service failed to start"
+    exit 1
+fi
+echo "   └─ ✅ Enrichment service started successfully"
+
+# Start AI Intelligence Layer in background
+echo "🤖 Starting AI Intelligence Layer on port 9000..."
+python ai_intelligence_layer/main.py &
+AI_PID=$!
+echo "   ├─ PID: $AI_PID"
+
+# Give AI layer time to start
+sleep 3
+
+# Check if AI layer is still running
+if ! kill -0 $AI_PID 2>/dev/null; then
+    echo "❌ AI Intelligence Layer failed to start"
+    kill $ENRICHMENT_PID 2>/dev/null || true
+    exit 1
+fi
+echo "   └─ ✅ AI Intelligence Layer started successfully"
+
+echo ""
+echo "✨ All services running!"
+echo "   📊 Enrichment Service: http://0.0.0.0:8000"
+echo "   🤖 AI Intelligence Layer: ws://0.0.0.0:9000/ws/pi"
+echo ""
+echo "Press Ctrl+C to stop all services"
+
+# Wait for both processes (this keeps the script running)
+wait $ENRICHMENT_PID $AI_PID