System Atlas.
Project index

Voice systems

Adaptive Voice Interview.

Participants answer by voice or text. A finite-state machine tracks the interview, uses the answer to choose a branch, and asks the next question. Each turn also yields structured fields.

Architecture

5 services
Service map3D model
Drag to rotate

Selected service

Interview client

The Next.js client collects answers and displays the next question.

Connections

  • To Send interview answers over a WebSocket
  • To Request dictation and read-aloud
View all connections
  1. Interview client → FastAPI interview APISend interview answers over a WebSocket
  2. Interview client → Azure AI SpeechRequest dictation and read-aloud
  3. FastAPI interview API → Interview state machineProcess the current interview turn
  4. Interview state machine → Azure OpenAIExtract fields and evaluate the next question

Technology stack

Next projectVoice Agent Practice Rooms