ibrahimm7004/AI-Voice-Agents
Production-ready voice agents and speech pipelines: STT → LLM/Agent → TTS, voice receptionists, telephony, call recording, tool/function calling. Built with Twilio, OpenAI Whisper, ElevenLabs, Vapi/Retell, FastAPI, WebSockets, ffmpeg; designed for deployment, monitoring, and real-world reliability
AI Voice Agents
Production-minded AI agent projects focused on real-time interaction, telephony integration, and practical evaluation workflows.
This repository contains two independent builds you can run and extend.
Projects
Projects Overview
| Project | What it is | Primary stack | Folder |
|---|---|---|---|
| Restaurant AI Phone Receptionist (Twilio) | Inbound call agent for restaurant operations with live transfer and SMS tooling. | Twilio Voice, Media Streams, OpenAI Realtime, Fastify, Firebase | restaurant-voice-receptionist-twilio/ |
| AI Interviewer Voice Agent (OpenAI) | Adaptive interview agent that changes questions based on responses and parameters. | OpenAI Assistants API, FastAPI, Python, Jinja2 | ai-interview-agent-openai/ |
Restaurant AI Phone Receptionist (Twilio)
Quick Breakdown
- Handles real inbound restaurant calls with low-latency AI responses.
- Streams call audio through Twilio Media Streams and WebSockets.
- Supports interruption (barge-in), FAQ handling, and menu/location/hours queries.
- Sends SMS follow-ups (links and details) during calls when requested.
- Escalates to live staff and stores call logs for QA and analytics.
Tech stack
- Twilio Voice, Twilio Media Streams, Twilio SMS
- OpenAI Realtime API (
gpt-4o-realtime-preview) - Node.js, Fastify,
@fastify/websocket,ws - Firebase Auth, Firestore, React/Vite admin tooling
- Docker, Google Cloud Run, Firebase Hosting
Where to look next: restaurant-voice-receptionist-twilio/README.md
AI Interviewer Voice Agent (OpenAI)
Quick Breakdown
- Runs dynamic interviews that adapt follow-up questions in real time.
- Uses user-defined parameters to shape role focus and interview direction.
- Delivers an interactive text-based interview flow (no video required).
- Adjusts question difficulty based on candidate responses.
- Continues evaluation-oriented questioning as answers evolve.
Tech stack
- OpenAI Assistants API
- Python, FastAPI, Uvicorn
- Jinja2 templates, HTML/CSS frontend
python-dotenv,requests- Heroku deployment target (plus Docker/Procfile support in repo)
Live app: AI Interviewer Web App
Where to look next: ai-interview-agent-openai/README.md
Repo Notes
Who this repo is for
- Teams building production-oriented AI reception or interviewing agents.
- Developers evaluating telephony workflows, real-time interaction, and adaptive interview logic.
How to run
Each project is self-contained and has its own setup instructions.
Use the project READMEs in restaurant-voice-receptionist-twilio/ and ai-interview-agent-openai/.