How adm-8/dsbc25 works
Project type and purpose
• This is a full‑stack, containerized web application (“Interviewee Agent”) that exposes a FastAPI backend (Uvicorn server) and serves static HTML pages as its frontend.
• Its primary function is to run conversational AI “agents” (interviewee and evaluation), leverage OpenAI’s API (and Agent SDK), perform speech‑to‑text (STT) and text‑to‑speech (TTS), and return results to the user via a browser interface.
File structure
• Top level:
– docker‑compose.yml, .env.example, setup.sh, dev.sh, _deploy.sh: build/deployment scripts
– config/: Nginx reverse‑proxy config + Prometheus scrape config
– certs/: TLS certificates for Nginx
– dashboards/: Grafana dashboard JSON
– .github/workflows/: CI/CD pipeline for DigitalOcean deploy
• app/ directory (core application):
– app/main.py + app/Dockerfile: application entry point & image recipe
– app/api/: FastAPI router modules (evaluation.py, interview.py)
– app/agents/: two high‑level agents and subfolders:
• prompts/: YAML system prompts + utilities
• tools/: parsers (extract_situation, action, etc.) + lie detector
• evaluation_agent.py & interviewee_agent.py orchestrate prompt/tool workflows
– app/core/: configuration & constants (OpenAI keys, environment)
– app/model/: local modules for STT, TTS, text‑to‑text (TTT) processing
– app/frontend/: static HTML for onboarding, interview, reporting, stats
– app/utils/: miscellaneous helper functions
• Monitoring & observability: Prometheus + Grafana JSON dashboards
• Deployment: orchestrated via Docker Compose behind Nginx, secrets managed via .env
README
• Confirms FastAPI + Uvicorn stack, external OpenAI integration, and Docker‑Compose based development & deploy steps.
• Describes agents’ directory layout and high‑level responsibilities.
• Provides CI/CD secret variables for DigitalOcean deployment.