Pi Deployment

Pi Deployment & Remote Testing Skill

Comprehensive workflow for deploying, testing, and debugging Bob The Skull on Raspberry Pi.

When to Use This Skill

• "Deploy to the Pi" - Full deployment workflow
• "Push updates to Raspberry Pi" - Code synchronization
• "Test on the Pi" - Remote testing procedures
• "Pi isn't working" - Troubleshooting remote issues
• "Setup new Raspberry Pi" - Initial Pi configuration
• "Check Pi logs" - Remote debugging

Quick Reference

Pre-Deployment Checklist

# 1. Check git status - ensure clean or intentional changes
git status

# 2. Verify critical files exist
- .env.bob (API keys configured)
- requirements-body.txt (Pi dependencies)
- setup_raspberry_pi_no_git.sh (setup script)
- BobTheSkull.py or BobSkullOnly.py (main entry point)

# 3. Test locally if possible (optional but recommended)
./venv/Scripts/python test_system_config.py

Deployment Command

# Windows: Use deploy_to_pi.bat
cmd /c deploy_to_pi.bat

# The script will:
# 1. Copy Python files (*.py)
# 2. Copy config files (.env.bob, requirements*.txt)
# 3. Copy all component directories (wake_word/, stt/, llm/, tts/, etc.)
# 4. Copy setup scripts

Post-Deployment: SSH Connection

# Connect to Pi
plink -pw peacock7 knarl@192.168.1.44

# Navigate to deployment directory
cd /home/knarl/BobTheSkull5

# Verify files copied
ls -la

Post-Deployment: Setup on Pi

# Make setup script executable
chmod +x setup_raspberry_pi_no_git.sh

# Run setup (installs dependencies, creates venv)
./setup_raspberry_pi_no_git.sh

# Setup installs:
# - System packages (python3, portaudio, mosquitto-clients, avahi-daemon)
# - Python venv
# - requirements-body.txt dependencies (excludes heavy vision/GPU packages)

Post-Deployment: Configuration

# Copy environment file
cp .env.bob .env

# Verify API keys are present
grep -E "OPENAI_API_KEY|PICOVOICE_ACCESS_KEY|ELEVEN_LABS_API_KEY" .env

# Check audio device configuration
python list_audio_devices.py

# Update .env if audio device indices changed
nano .env
# BOBTHESKULL_AUDIO_INPUT_DEVICE_INDEX=X
# BOBTHESKULL_AUDIO_OUTPUT_DEVICE_INDEX=Y

Running on Pi

# Activate virtual environment
source venv/bin/activate

# Run Bob (body mode - no vision)
python BobTheSkull.py

# Or run in background with logging
nohup python BobTheSkull.py > bob.log 2>&1 &

# Monitor logs in real-time
tail -f bob.log

# Or use screen for persistent session
screen -S bob
python BobTheSkull.py
# Detach: Ctrl+A then D
# Reattach: screen -r bob

Deployment Workflow (Step-by-Step)

Step 1: Pre-Deployment Preparation

Check what will be deployed:

# Review git status
git status

# Check for uncommitted changes to critical files
git diff BobConfig.py
git diff BobTheSkull.py
git diff .env.bob

Verify configuration files:

# Ensure .env.bob has valid API keys
cat .env.bob | grep -E "API_KEY|ACCESS_KEY"

# Ensure deployment mode is correct
grep "DEPLOYMENT_MODE" .env.bob
# Should show: BOBTHESKULL_DEPLOYMENT_MODE=bob

Step 2: Deploy Files

Run deployment script:

# Windows
cmd /c deploy_to_pi.bat

# Expected output:
# [1/4] Creating directory on Pi...
# [2/4] Copying Python files...
# [3/4] Copying configuration files...
# [4/4] Copying component directories...
# Deployment Complete!

What gets copied:

• Python files: All *.py in root
• Config files: .env.bob, requirements*.txt, setup scripts
• Component directories: events/, wake_word/, stt/, llm/, tts/, state_machine/, vision/, hardware/, web/

What does NOT get copied:

• venv/ - Virtual environment (recreated on Pi)
• .git/ - Git repository
• __pycache__/ - Python cache
• logs/ - Old log files
• EmbeddingsDB/, LongTermDB/ - Databases

Step 3: Connect to Pi

SSH Connection:

# Connect via plink (PuTTY)
plink -pw peacock7 knarl@192.168.1.44

# Or use WSL/Git Bash
ssh knarl@192.168.1.44
# Password: peacock7

Network Info:

• Pi IP: 192.168.1.44
• User: knarl
• Password: peacock7
• Hostname: bob-pi.local (mDNS)

Step 4: Setup Environment on Pi

First-time setup (or after major changes):

cd /home/knarl/BobTheSkull5

# Make setup script executable
chmod +x setup_raspberry_pi_no_git.sh

# Run setup (5-10 minutes)
./setup_raspberry_pi_no_git.sh

What setup does:

1. Updates system packages (apt update && apt upgrade)
2. Installs system dependencies (Python, PortAudio, MQTT, mDNS)
3. Creates Python virtual environment
4. Installs Python packages from requirements-body.txt
5. Enables Avahi daemon for mDNS discovery

Setup complete when you see:

=========================================
Setup Complete!
=========================================

Next steps:
1. Copy .env.body to .env and update with your PC's IP address
2. Add your Picovoice API key to .env
3. Run: source venv/bin/activate
4. Run: python BobSkullOnly.py --broker <YOUR_PC_IP>

Step 5: Configure Audio Devices

List available audio devices:

python list_audio_devices.py

Example output:

Input Devices:
  [0] Built-in Microphone
  [4] PS3 Eye Camera (USB Mic)

Output Devices:
  [2] Built-in Speaker
  [3] USB Audio Device

Update .env with correct indices:

nano .env

# Set to your device indices
BOBTHESKULL_AUDIO_INPUT_DEVICE_INDEX=4
BOBTHESKULL_AUDIO_OUTPUT_DEVICE_INDEX=3

Step 6: Test Components

Test audio output:

# Test speaker/audio output
python test_audio_output.py
# Should hear test tones

Test wake word detection:

# Test microphone and wake word
python test_wake_word_live.py
# Say "Wake up Bob" or "Hey Bob"
# Should see detection events

Test MQTT connection (if using distributed mode):

# Test MQTT broker connection
python test_mqtt.py
# Should connect to localhost:1883

Step 7: Run Bob

Standard run:

source venv/bin/activate
python BobTheSkull.py

Background run with logging:

nohup python BobTheSkull.py > bob.log 2>&1 &

# Monitor logs
tail -f bob.log

# Stop background process
pkill -f BobTheSkull.py

Using screen (recommended for long-term):

# Start screen session
screen -S bob
python BobTheSkull.py

# Detach from screen: Ctrl+A then D
# Reattach to screen: screen -r bob
# Kill screen: Ctrl+C then exit

Testing Strategies

Component-by-Component Testing

Test in this order to isolate issues:

1. Audio Output

   python test_audio_output.py
   

   Verifies: Speaker/audio device working

2. Audio Input

   python test_wake_word_live.py
   

   Verifies: Microphone working, wake word detection

3. MQTT/EventBus

   python test_mqtt.py
   

   Verifies: Event bus communication

4. Web Monitor

   python test_web_monitor.py
   # Open browser: http://192.168.1.44:5001
   

   Verifies: Web interface accessible

5. Full System

   python BobTheSkull.py
   

   Verifies: Complete integration

Remote Debugging via Web Monitor

Access from Windows PC:

http://192.168.1.44:5001

Monitor dashboard shows:

• Current state machine state
• Recent events
• Component status
• System metrics
• Real-time logs

Common Errors & Troubleshooting

Error: "pscp not found"

Problem: PuTTY not installed or not in PATH

Solution:

# Option 1: Install PuTTY
# Download from: https://www.putty.org/

# Option 2: Use WSL/Git Bash instead
scp -r *.py knarl@192.168.1.44:/home/knarl/BobTheSkull5/

Error: "Connection refused" when deploying

Problem: Pi is offline or IP address changed

Solution:

# Ping to verify Pi is online
ping 192.168.1.44

# Try mDNS hostname
ping bob-pi.local

# If IP changed, update deploy_to_pi.bat:
# set PI_HOST=<NEW_IP>

Error: "PortAudio not found" on Pi

Problem: System dependencies not installed

Solution:

# SSH to Pi
plink -pw peacock7 knarl@192.168.1.44

# Install PortAudio manually
sudo apt update
sudo apt install -y portaudio19-dev python3-pyaudio

# Recreate venv
rm -rf venv
python3 -m venv venv
source venv/bin/activate
pip install -r requirements-body.txt

Error: "No module named 'pvporcupine'"

Problem: Wrong requirements file used or venv not activated

Solution:

# Ensure using requirements-body.txt (not requirements.txt)
source venv/bin/activate
pip install -r requirements-body.txt

# Verify pvporcupine installed
pip list | grep pvporcupine

Error: Audio device index invalid

Problem: Device indices changed or incorrect

Solution:

# List current audio devices
python list_audio_devices.py

# Update .env with correct indices
nano .env
# Change BOBTHESKULL_AUDIO_INPUT_DEVICE_INDEX
# Change BOBTHESKULL_AUDIO_OUTPUT_DEVICE_INDEX

# Test audio
python test_audio_output.py

Error: "Permission denied" for serial port

Problem: User not in dialout group

Solution:

# Add user to dialout group
sudo usermod -a -G dialout knarl

# Logout and login again (or reboot)
sudo reboot

Error: Wake word not detecting

Problem: Microphone not working or sensitivity too low

Solution:

# 1. Verify microphone
python list_audio_devices.py
# Ensure correct input device index

# 2. Test microphone recording
arecord -d 5 -f cd test.wav
aplay test.wav
# Should hear your recording

# 3. Adjust wake word sensitivity
nano .env
# BOBTHESKULL_WAKE_WORD_SENSITIVITY=0.3  # Lower = more sensitive (0.0-1.0)

# 4. Test wake word
python test_wake_word_live.py

Error: MQTT connection failed

Problem: MQTT broker not running or wrong config

Solution:

# Check if mosquitto is running
sudo systemctl status mosquitto

# Start mosquitto if not running
sudo systemctl start mosquitto
sudo systemctl enable mosquitto

# Verify MQTT config in .env
grep MQTT .env
# Should show:
# BOBTHESKULL_MQTT_BROKER_HOST=localhost
# BOBTHESKULL_MQTT_BROKER_PORT=1883

Error: Eyes controller not found

Problem: Eyes controller offline or discovery failed

Solution:

# Test mDNS discovery
python test_mdns_discovery.py

# Try serial discovery
ls /dev/ttyUSB*
# If serial device exists: /dev/ttyUSB0

# Update .env to force serial mode
nano .env
# BOBTHESKULL_EYES_DISCOVERY_MODE=serial
# BOBTHESKULL_EYES_SERIAL_PORT=/dev/ttyUSB0

Slow performance on Pi

Problem: Pi CPU limited or wrong Python version

Solution:

# Check CPU usage
top
# Look for high CPU processes

# Verify Python version (should be 3.9-3.11)
python3 --version

# Ensure using body requirements (no vision overhead)
pip list | grep -E "opencv|onnx|dlib"
# Should NOT be installed in body mode

# Consider disabling wake word processing on Pi
# Use distributed architecture with vision PC doing wake word

Remote Log Monitoring

View logs from Windows

Option 1: SSH and tail

plink -pw peacock7 knarl@192.168.1.44 "tail -f /home/knarl/BobTheSkull5/bob.log"

Option 2: Web Monitor

http://192.168.1.44:5001/logs

Log file locations on Pi

# Main application log
~/BobTheSkull5/bob.log

# System logs (if running as service)
/var/log/syslog | grep bob

# MQTT logs
/var/log/mosquitto/mosquitto.log

Useful log filtering

# Show errors only
tail -f bob.log | grep -i error

# Show state transitions
tail -f bob.log | grep "State transition"

# Show event publications
tail -f bob.log | grep "Publishing event"

# Show LLM interactions
tail -f bob.log | grep "LLM"

Deployment Modes

Body Mode (Raspberry Pi)

Configuration (.env.bob):

BOBTHESKULL_DEPLOYMENT_MODE=bob
BOBTHESKULL_VISION_CAN_SEE=false
BOBTHESKULL_WAKE_WORD_ENABLED=true
BOBTHESKULL_MQTT_BROKER_HOST=localhost

Components enabled:

• ✅ Wake Word Detection
• ✅ Speech-to-Text (STT)
• ✅ Language Processing (LLM)
• ✅ Text-to-Speech (TTS)
• ✅ State Machine
• ✅ Eyes Controller
• ✅ Web Monitor
• ❌ Vision System (disabled)

Use when: Running standalone Bob on Pi without vision PC

Distributed Mode (Future)

Vision PC runs:

• Vision system
• MQTT broker

Raspberry Pi runs:

• Wake word, STT, LLM, TTS, Eyes
• Connects to vision PC's MQTT broker

Configuration (.env on Pi):

BOBTHESKULL_MQTT_BROKER_HOST=192.168.1.XXX  # Vision PC IP

Pro Tips

1. Use screen for persistence - Prevents SSH disconnects from stopping Bob

   screen -S bob
   python BobTheSkull.py
   # Ctrl+A D to detach
   

2. Create deployment aliases - Add to ~/.bashrc on Pi:

   alias bob-run='cd ~/BobTheSkull5 && source venv/bin/activate && python BobTheSkull.py'
   alias bob-log='tail -f ~/BobTheSkull5/bob.log'
   alias bob-stop='pkill -f BobTheSkull.py'
   

3. Monitor from Windows - Keep web monitor open during deployment:

   http://192.168.1.44:5001
   

4. Test incrementally - Don't deploy everything at once:

   • Deploy → Test audio → Deploy → Test wake word → Deploy → Test full system

5. Keep backup .env - Copy working .env before changing:

   cp .env .env.backup
   

6. Use version tags - Tag working versions in git:

   git tag -a pi-working-2024-12-09 -m "Working Pi deployment"
   

7. Automated health checks - Run periodic tests:

   # On Pi, add to crontab
   */5 * * * * /home/knarl/BobTheSkull5/health_check.sh
   

8. Remote file editing - Use nano on Pi, VS Code Remote SSH, or WinSCP

9. Quick config changes - Use web interface instead of editing files:

   http://192.168.1.44:5001/config
   

10. Network discovery - Use mDNS instead of IP addresses:

    ping bob-pi.local
    ssh knarl@bob-pi.local
    

Common Workflows

Workflow: Quick Bug Fix Deployment

# 1. Fix bug locally on Windows
# 2. Test locally (if possible)
./venv/Scripts/python test_system_config.py

# 3. Deploy single file (faster than full deploy)
pscp -pw peacock7 BobConfig.py knarl@192.168.1.44:/home/knarl/BobTheSkull5/

# 4. Restart Bob on Pi
plink -pw peacock7 knarl@192.168.1.44 "pkill -f BobTheSkull.py"
plink -pw peacock7 knarl@192.168.1.44 "cd ~/BobTheSkull5 && nohup python BobTheSkull.py > bob.log 2>&1 &"

# 5. Monitor logs
plink -pw peacock7 knarl@192.168.1.44 "tail -f ~/BobTheSkull5/bob.log"

Workflow: Configuration Change

# Option 1: Use web interface (fastest)
# Open: http://192.168.1.44:5001/config
# Make changes via UI
# Click Save

# Option 2: Edit .env remotely
plink -pw peacock7 knarl@192.168.1.44
nano /home/knarl/BobTheSkull5/.env
# Make changes
# Ctrl+X, Y, Enter
exit

# Restart Bob
plink -pw peacock7 knarl@192.168.1.44 "pkill -f BobTheSkull.py && cd ~/BobTheSkull5 && nohup python BobTheSkull.py > bob.log 2>&1 &"

Workflow: Fresh Pi Setup

# 1. Deploy files from Windows
cmd /c deploy_to_pi.bat

# 2. SSH to Pi
plink -pw peacock7 knarl@192.168.1.44

# 3. Run full setup
cd /home/knarl/BobTheSkull5
chmod +x setup_raspberry_pi_no_git.sh
./setup_raspberry_pi_no_git.sh

# 4. Configure environment
cp .env.bob .env
nano .env
# Update audio device indices

# 5. Test audio
python list_audio_devices.py
python test_audio_output.py

# 6. Run Bob
screen -S bob
python BobTheSkull.py
# Ctrl+A D to detach

Workflow: Debugging Remote Issues

# 1. Check if Bob is running
plink -pw peacock7 knarl@192.168.1.44 "ps aux | grep BobTheSkull"

# 2. Check recent logs
plink -pw peacock7 knarl@192.168.1.44 "tail -100 ~/BobTheSkull5/bob.log"

# 3. Check for errors
plink -pw peacock7 knarl@192.168.1.44 "grep -i error ~/BobTheSkull5/bob.log | tail -20"

# 4. Open web monitor in browser
# http://192.168.1.44:5001

# 5. If needed, run tests remotely
plink -pw peacock7 knarl@192.168.1.44 "cd ~/BobTheSkull5 && source venv/bin/activate && python test_mqtt.py"

Time Savings

Without this skill:

• 30-45 minutes per deploy (forgotten steps, troubleshooting, trial-and-error)
• 15-20 minutes debugging common errors
• Frequent deployment failures requiring rework

With this skill:

• 10-15 minutes per deploy (documented workflow)
• 5 minutes resolving common errors (troubleshooting guide)
• Higher success rate on first attempt

Estimated time savings: 2-3x faster deployments

Integration with Other Skills

Works well with:

• cross-repo-sync - When deploying files from BobFast5 (audio files)
• config-pattern - Testing new config parameters on Pi
• event-flow - Verifying event handling in distributed MQTT architecture

Common Mistakes to Avoid

1. ❌ Deploying without testing locally first

   • Always test locally when possible before deploying

2. ❌ Forgetting to update .env on Pi

   • Always verify .env after deployment

3. ❌ Using wrong requirements file

   • Pi should use requirements-body.txt, not requirements.txt

4. ❌ Not checking audio device indices

   • Indices can change, always verify with list_audio_devices.py

5. ❌ Deploying with hardcoded paths

   • Use relative paths or environment variables

6. ❌ Not monitoring logs during first run

   • Always tail logs during initial deployment

7. ❌ Forgetting to enable avahi-daemon

   • mDNS discovery requires avahi running

8. ❌ Running full requirements.txt on Pi

   • Installs unnecessary vision/GPU packages (slower, more disk space)

9. ❌ Not using screen/nohup for background runs

   • SSH disconnect will kill Bob

10. ❌ Skipping component-by-component testing

    • Test audio → wake word → full system incrementally

References

Deployment Scripts:

• deploy_to_pi.bat - Windows deployment script
• setup_raspberry_pi_no_git.sh - Pi setup script
• .env.bob - Body mode configuration template

Test Scripts:

• test_audio_output.py - Test speaker
• test_wake_word_live.py - Test microphone and wake word
• test_mqtt.py - Test MQTT connection
• test_web_monitor.py - Test web interface
• list_audio_devices.py - List audio devices

Configuration Files:

• BobConfig.py - Default configuration
• .env - Environment-specific overrides (on Pi)
• config.yaml - Optional YAML configuration

Related Documentation:

• CLAUDE.md - Project overview
• requirements/DeploymentRequirements.md - Deployment specs