TL;DR
Asterisk AI Voice Agent is an open-source, modular platform that integrates speech-to-text, large language models, and text-to-speech into Asterisk/FreePBX. Version 4.5.3 adds call history, barge-in improvements, new model support, and a privacy-focused local hybrid default.
What happened
Asterisk AI Voice Agent, an MIT-licensed project, delivers a configurable AI voice agent that runs natively with Asterisk and FreePBX. The system uses a modular pipeline so administrators can combine different STT, LLM and TTS providers (cloud, local, or hybrid). The project's v4.5.3 release introduced full call logging and per-call debugging, improvements to barge-in/interrupt handling, additional model backends and a privacy-first local_hybrid default pipeline. The project includes an Admin UI that can be launched with Docker Compose (default port 3003) and a preflight script that generates a .env and a JWT secret; the documentation warns to change the default admin credentials and to restrict the dashboard network exposure. Deployers can choose production-ready golden baseline configs for OpenAI Realtime, Deepgram, Google Live, ElevenLabs, or a Local Hybrid; there is also documented support for a self-hosted Ollama LLM and CLI tools for installation, diagnostics and dialplan generation.
Why it matters
- Modular provider selection lets operators balance cost, latency and privacy by mixing local and cloud components.
- Built-in call history and per-call debugging aim to simplify troubleshooting and auditing of voice interactions.
- Privacy-focused options and a fully self-hosted Ollama pipeline allow audio processing to remain on-premises when required.
- Multiple validated baseline configurations and transport support lower the integration effort for enterprise Asterisk/FreePBX deployments.
Key facts
- Project license: MIT (fully open source).
- Latest referenced release in the source: v4.5.3, which added call history, barge-in improvements and model hot-swap.
- Admin UI runs via Docker Compose on port 3003; default login shown as admin / admin and changing it is recommended.
- Preflight script (preflight.sh –apply-fixes) creates .env and generates a secure JWT_SECRET; running it is required per docs.
- Health endpoint for ai-engine: http://localhost:15000/health — expected response: {"status":"healthy"}.
- Two transport methods supported: AudioSocket and ExternalMedia RTP (transport compatibility matrix provided).
- Five production-ready golden baseline configs: OpenAI Realtime, Deepgram Voice Agent, Google Live API, ElevenLabs Agent, and Local Hybrid.
- Self-hosted LLM option via Ollama supports tool calling and runs locally without requiring an API key; recommended hardware notes include 8GB+ RAM (16GB+ for larger models).
- CLI utilities available for setup and operations include agent quickstart, agent doctor, agent dialplan and agent troubleshoot.
What to watch next
- Community contributions and plugin ecosystem growth — not confirmed in the source.
- Long-term enterprise adoption and integration patterns with existing Asterisk deployments — not confirmed in the source.
- Future release roadmap and security patch cadence beyond v4.5.3 — not confirmed in the source.
Quick glossary
- STT: Speech-to-text; technology that converts spoken audio into written transcripts.
- TTS: Text-to-speech; software that generates synthetic spoken audio from text.
- LLM: Large language model; a machine learning model trained on large text corpora to generate or interpret language.
- RTP (ExternalMedia): Real-time Transport Protocol used for delivering audio streams; ExternalMedia refers to an Asterisk transport mode for external audio handling.
- Ollama: A local model-serving option referenced by the project that enables on-premises LLM hosting without cloud API keys.
Reader FAQ
Is Asterisk AI Voice Agent open source?
Yes. The project is released under the MIT license per the source.
Can the system be deployed on-premises without cloud APIs?
Yes. The source documents a Local Hybrid baseline and an Ollama-based self-hosted LLM option that requires no API key.
What are the default Admin UI credentials and recommended action?
The dashboard defaults to admin / admin; the documentation advises changing the password immediately and restricting network access.
Does the agent support multiple transports and providers?
Yes. The project supports AudioSocket and ExternalMedia RTP and allows mixing STT, LLM, and TTS providers, including cloud and local options.
Where can I test a demo call?
The source lists a US demo number with menu options for different baselines, but detailed availability and call behavior should be confirmed by the source documentation.
Asterisk AI Voice Agent The most powerful, flexible open-source AI voice agent for Asterisk/FreePBX. Featuring a modular pipeline architecture that lets you mix and match STT, LLM, and TTS providers,…
Sources
- Asterisk AI Voice Agent
- AI Voice Agent for Asterisk: Seeking a Frontend Co-Builder
- AI Voice Agent Highly Cutomizable – AI
- Building Your Own AI Call Center in 15 Minutes | Asterisk …
Related posts
- TOPS-10 in a Box: Ready-to-run PDP-10 TOPS-10 image with Adventure
- Garmin Venu 3S: sleek, feature-packed smartwatch drops to $350
- Confessions to a Data Lake: Building End-to-End Encrypted AI Chats