Lesson 6: Run & test the bot

Start the service

From the repo root:

cd medical-appointment-bot
npm run start:dev

Swagger is live at http://localhost:8000/api.

Kick off a session

curl -X POST http://localhost:8000/ai/run \
  -H 'Content-Type: application/json' \
  -d '{
    "message": "Hi, I need to see someone about a headache",
    "flowName": "MedicalFlow",
    "config": {}
  }'

The JSON response contains session—save it for the next turn.

Continue the conversation

SESSION=<<paste session id>>
curl -X POST http://localhost:8000/ai/run \
  -H 'Content-Type: application/json' \
  -H "CHAT_SESSION_ID: $SESSION" \
  -d '{
    "message": "It started yesterday and is pretty sharp",
    "flowName": "MedicalFlow"
  }'

The engine stays in SymptomsStep until capture_symptoms fires, then jumps into BookingStep with the proposed doctors and time slots.

End the chat

curl -X POST http://localhost:8000/ai/end \
  -H "CHAT_SESSION_ID: $SESSION"

You should see { "success": true, "session": "..." }. The session document is marked complete and will expire based on SESSION_EXPIRATION.

Troubleshooting tips

Config missing? Ensure flowName is exactly MedicalFlow and GEMINI_KEY is set.
Persistent sessions? Point the .env at Mongo or Cosmos; otherwise state may be in-memory only depending on your Picoflow core config.
Model tweaks? Adjust LLM_TEMPERATURE or swap models in ChatController.registerModel.