This repository is for demonstration purposes only and should not be used for production under any circumstances.
The core module that orchestrates voice interactions. It manages audio input/output, session state, and coordinates between the other modules to deliver voice-based features.
Handles telephony integration, connecting the voicebot to Asterisk via ARI. It manages call events and sets up the audio streams.
Provides speech-to-text (STT) and text-to-speech (TTS) capabilities. Audio streams from vproxy are sent to and from deepgram for transcription and recitation, enabling the voicebot to understand spoken input and respond accordingly.
Responsible for building the service by providing tools and instructions for the AI and executing choices based on it's output.
Implements natural language understanding and response generation via OpenAI Realtime. Receives transcribed text from deepgram, processes user intent, and generates appropriate replies for the voicebot to deliver.
Acts as a voice proxy layer, managing RTP traffic between modules and external services.
- ariman receives calls and streams audio to deepgram via vproxy for transcription.
- deepgram returns text to voiceai, which interprets and generates responses.
- voiceai sends responses back to deepgram for delivery to the caller via vproxy.
- voicebot manages the overall session and coordinates with rclocal which queries the IVR-Demo http server for external AI instruction and function parsing.
The application is installed via the follwoing command:
go install .
This application requires an OpenAI and Deepgram API key in order to use those services. These are set with the following environment variables:
- OPENAI_API_TOKEN - Your OpenAI API token
- DG_API_TOKEN - Your Deepgram API token
Once these are set, the application is run via:
go run .
The application contains a folder IVR-Demo which defines a sample IVR. Inside this folder is a script ivr-runner.sh that will launch a simple python http server that hosts all files within the folder. The rclocal package is hard coded to perform a GET on http://localhost:8989/ivr-root.json which contains the base configuration for the IVR. The IVR-Demo also includes the IVR leaves that the AI service may pick from. Each leaf json file transfers the call to a specific point in the included Asterisk Dialplan.
This application contains a folder dockerfiles with two scripts: buildAVai-Deb12Image.sh which will build and package a Debian 12 Docker image containing Asterisk, the mini http server and the voicebridge application and runAVai-Deb12.sh which will run a named instance. See dockerfiles/README.md for more information.
This project is licensed under the GNU Affero General Public License (AGPL).
See LICENSE for details.