VoicEra
Open-source, self-hosted infrastructure for real-time voice AI telephony agents in Indian languages, with swappable speech-to-text, text-to-speech, LLM and telephony providers.
VoicEra is open-source infrastructure for building self-hosted, real-time voice AI telephony agents, aimed particularly at Indian languages, so an operator keeps its own infrastructure, data and choice of models and carriers rather than depending on a single vendor.
It separates a control plane, an API handling agents, configuration and campaigns behind a dashboard, from real-time execution, a Pipecat-based voice runtime that connects telephony providers such as Vobiz or Plivo to speech-to-text, LLM and text-to-speech providers. Call media and artifacts are stored in MinIO, structured data in FerretDB over PostgreSQL, and background jobs run through Redis-backed workers.
It is provider-neutral, so speech-to-text, text-to-speech, LLM and telephony providers can be swapped without rewriting the platform, and deploys as a Docker stack that an operator runs on their own infrastructure.