Voice Bot Server is an opinionated version of Vocode Telephony Server with FastAPI - Deepgram, OpenAI, and ElevenLabs are used here to achieve a realistic sounding agent. This project enables you to build and deploy conversational AI agents that can handle inbound and outbound phone calls for various use cases such as information collection, appointment scheduling, sales, customer support, and more.

Voice Bot leverages the power of Vocode to support using AI agents with inbound and outbound phone calls. Users can create their own custom agents and integrate them seamlessly into their telephony systems.

• 🐳 Docker (required for running the project)

API keys for:

• 🎙️ Deepgram (for speech transcription)
• 🧠 OpenAI (for the underlying agent)
• 🗣️ ElevenLabs (for speech synthesis)
• 📞 Twilio (for telephony)

The project uses the following components, which are automatically set up through Docker:

• 🎬 ffmpeg (for audio processing)
• 🗄️ Redis (for data storage and caching)
• 📊 Prometheus (for metrics collection)

These dependencies are defined in the docker-compose.yml file and will be pulled in automatically when you run the project.

For local testing and development:

• 🌐 Ngrok or Cloudflare Tunnels (optional, used to expose your local server to the internet for testing purposes)

Note: Ngrok or Cloudflare Tunnels are not required for deployment but can be useful for local testing and development.

The Voice Bot now includes Prometheus metrics for monitoring session activity:

• voicebot_session_count: A counter that tracks the total number of sessions started.
• voicebot_active_sessions: A gauge that shows the current number of active sessions.

These metrics are exposed on port 8000 and can be scraped by a Prometheus server.

1. Copy the .env.template file and fill in the values of your API keys:

   cp .env.template .env
   

2. Set up hosting so that Twilio can hit your server. An easy way to do this is ngrok or cloudflare:

   ngrok http 6000
   

   Copy the URI that is tunneling localhost:3000 to your .env without https://, e.g.:

   BASE_URI=asdf1234.ngrok.app
   

The TelephonyServer is responsible for receiving and making phone calls. The server is built using FastAPI and utilizes Twilio for telephony services.

1. Build and run the application using docker-compose:

   docker-compose up --build
   

1. Create a Twilio account
2. In your Twilio dashboard, go to Phone Numbers -> Manage -> Buy a number to get a phone number
3. Go to Phone Numbers -> Manage -> Active Numbers and select the number you want to set up
4. Update the config to point the Webhook URL to https://<BASE_URI>/inbound_call
5. Hit Save and call the number!

1. Ensure the server is running
2. Install Poetry and dependencies:

   poetry 
   

3. Ensure .env has TO_PHONE set as the number you want to call and FROM_PHONE with your Twilio number.
4. Run the script:

   poetry run python outbound_call.py
   

Both the OutboundCall (in outbound_call.py) and InboundCallConfig (in inbound.py) classes can accept a TranscriberConfig, AgentConfig, or SynthesizerConfig.

The default transcriber is Deepgram

The default synthesizer is ElevenLabs

• Kubernetes cluster (e.g., Minikube, EKS, GKE, AKS)
• Helm 3.0+
• kubectl configured to communicate with your cluster

1. Clone the repository and navigate to the project directory:

   git clone https://github.com/Agentic-Insights/voice-bot
   cd voice-bot
   

2. Create a my-values.yaml file in the project root with your specific configuration:

   env:
     BASE_URI: "application-base-uri"
     DEEPGRAM_API_KEY: "your-deepgram-key"
     OPENAI_API_KEY: "your-openai-key"
     OPENAI_BASE_URL: "your-openai-based-base-url"
     ELEVEN_LABS_API_KEY: "your-eleven-labs-key"
     ELEVEN_LABS_VOICE_ID: "your-eleven-labs-voice-id"
     TWILIO_ACCOUNT_SID: "your-twilio-sid"
     TWILIO_AUTH_TOKEN: "your-twilio-token"
     FROM_PHONE: "your-from-phone"
     TO_PHONE: "your-to-phone"

3. Install the Helm chart:

   helm install voice-bot ./kubernetes/voice-bot-helm/voice-bot -f my-values.yaml
   

4. Check the status of your deployment:

   kubectl get pods
   kubectl get services
   

5. To access your application, set up port-forwarding:

   k8s-port-forward.bat
   (TODO does not work) kubectl port-forward service/voice-bot 6000:3000 (/TODO)
   

   Your application should now be accessible at http://localhost:6000.

To update your deployment after making changes:

helm upgrade voice-bot ./kubernetes/voice-bot-helm/voice-bot -f my-values.yaml

To remove the Voice Bot Server from your cluster:

helm uninstall voice-bot

• If you encounter issues, check the logs of your pods:

  kubectl logs deployment/voice-bot
  

• For more detailed information about your deployment:

  kubectl describe deployment voice-bot
  kubectl describe service voice-bot
  

Remember to update your Twilio webhook URL to point to your new Kubernetes-hosted Voice Bot Server.