This project is a Voice Assistant that can make phone calls and coordinate tasks on behalf of users using two different approaches:
- Vapi Voice Agent Workflow
- Gemini Vertex AI Voice Agent
To test the full experience, place a call to: +1 (956) 528-4589 (or 0019565284589 internationally)
.
├── vapi_workflow.json # Workflow definition for Vapi voice agent
├── main.py # FastAPI backend for handling Vapi webhooks
├── response.json # Sample post-call summaries
├── gemini_voice_agent/ # Gemini-based voice AI implementation
│ └── ... # Gemini Vertex AI logic for call handling
└── README.md # You're here!
- Handles backend logic with FastAPI.
- Generates post-call summaries stored in
response.json.
vapi_workflow.json: Defines the conversation flow and logic.main.py: FastAPI server for handling webhook callbacks.response.json: Sample summaries of previously tested conversations.
+1 (956) 528-4589 (0019565284589 for international format)
[
{
"intent": "Inquiry",
"entities": {
"service": "Darshan",
"location": "Balaji Temple, Trimal",
"date": "second week of next month"
},
"status": "Inquiry received",
"action_taken": "None",
"notes": "Customer is looking to book a pilgrimage for Darshan at Balaji Temple in Trimal during the second week of next month.",
"natural_language_summary": "The caller is inquiring about pilgrimage booking for Darshan at Balaji Temple in Trimal. They are interested in booking for the second week of next month. No action was taken during the call."
}
]- Uses Gemini from Vertex AI for intent parsing and natural language interaction.
- Ideal for more complex or dynamic voice tasks.
All LLM-related code is contained in:
gemini_voice_agent/
This includes:
- Intent classification
- Entity extraction
- Voice assistant chatbot
- save chat history
- summarise chat history