How dkpython7/ai-voice-support-agent works
GODIMS Voice Support Agent answers customer calls with a VideoSDK agent backed by Gemini Realtime. The workflow configures a voice model and agent session, waits for a participant, then conducts a bilingual support conversation: greeting and language choice, service/pricing/timeline answers, and conversational collection of project and contact details before closing. The prompt defines these capabilities in main.py; there is no separate lead-storage component. Edges reflect the setup code and documented workflow; framework-internal audio and callback routing is not shown in the excerpts.
Main entry points: main.py Support Agent
Components 9
Agent Runtime
- VideoSDK Worker
main.py - Agent Session
main.py - Voice Pipeline
main.py - Gemini Realtime
- Voice Response
Support Conversation
- Support Agent
main.py - Service Guidance
main.py - Lead Collection
main.py
Other components
- Customer
Connections 9
- VideoSDK Worker → Agent Session: starts
main.py - Agent Session → Voice Pipeline: uses
main.py - Voice Pipeline → Gemini Realtime: configures
main.py - Agent Session → Support Agent: instantiates
main.py - Customer → Agent Session: joins call
Readme.md - Support Agent → Service Guidance: answers from
main.py - Support Agent → Lead Collection: collects details
main.py - Support Agent → Voice Response: speaks
main.py - Voice Response → Customer: responds to