Troubleshooting#
Startup Fails#
Symptom#
start_ata.ps1 exits early or one or more services never become ready.
Checks#
# Verify required files exist
Test-Path .\setup_windows.ps1
Test-Path .\start_ata.ps1
# Verify submodule paths
Test-Path .\edge-ai-libraries\microservices\audio-analyzer\main.py
Test-Path .\edge-ai-libraries\microservices\text-to-speech\main.py
Test-Path .\voice-enabled-interactions\smart-kiosk-assistant\main.py
Test-Path .\voice-enabled-interactions\smart-kiosk-assistant\rag-service\main.py
If missing, rerun:
.\setup_windows.ps1
Port Conflicts#
Symptom#
Launcher reports an existing process on required ports.
Checks#
netstat -ano | findstr :7860
netstat -ano | findstr :8012
netstat -ano | findstr :8020
netstat -ano | findstr :8011
netstat -ano | findstr :8010
netstat -ano | findstr :9000
Stop conflicting PIDs or run:
.\stop_ata.ps1 -Force
Models Download Slowly or First Run Is Long#
First startup can take significant time due to model downloads and OpenVINO compilation.
Useful checks:
ping huggingface.co
If interrupted, rerun start_ata.ps1 after connectivity is stable.
UI Not Loading on 7860#
Symptom#
http://127.0.0.1:7860 is unavailable.
Checks#
curl http://127.0.0.1:7860/healthz
If health fails:
Verify
ata_ui_server.pyis running.Ensure React build assets exist under
assistant-react-ui/dist.
Rebuild UI assets if needed:
cd assistant-react-ui
npm install
npm run build
Microphone Permission Issues#
Symptom#
Mic button does nothing or browser shows permission errors.
Fixes#
Allow microphone for
http://127.0.0.1:7860in browser site settings.Verify Windows microphone privacy settings allow the browser.
Test with another input device in Windows sound settings.
Upload or Ingestion Failures#
Symptom#
File uploads fail or context stats do not increase.
Checks#
Supported formats:
.txt,.md,.docx,.pdfVerify file size is within configured limits.
Clear context and retry upload.
Service checks:
curl http://127.0.0.1:8020/health
curl http://127.0.0.1:8020/api/v1/context/stats
Clear Context Fails#
The UI now surfaces clear-context errors. If clear fails, inspect rag-service
output and verify storage paths are writable.
Slow or Low-Quality Answers#
Speed tuning#
Use a smaller LLM in
rag-service/config.yamlLower retrieval complexity (
top_k,fetch_k)Use GPU for LLM where available
Relevance tuning#
Improve document quality and structure
Re-ingest after changing chunking settings
Adjust reranker usage in
rag-service/config.yaml
No Audio Playback#
Symptom#
Transcript/answer text appears but no spoken audio.
Checks#
Verify
text-to-speechhealth:curl http://127.0.0.1:8011/healthVerify browser/tab volume and output device
Verify session snapshot contains
tts_audio_segments
Metrics Panel Errors#
Symptom#
Metrics cards show errors or blank values.
Checks#
curl http://127.0.0.1:9000/health
curl http://127.0.0.1:8012/api/v1/metrics
curl http://127.0.0.1:8012/api/v1/platform-info
Collect Useful Debug Data#
systeminfo > system_info.txt
python --version > python_version.txt
# Record running processes/ports
netstat -ano > netstat_ports.txt
Then include launcher and service terminal logs when reporting issues.