Voice AI on infrastructure you own

Until now, AI agents answered calls on somebody else’s cloud. Two modules change that: connect_dograh and connect_pipecat bridge open voice-agent frameworks into the self-hosted FreeSWITCH backend, streaming audio out over mod_audio_fork and writing the summary and transcript back into the call when the conversation ends. The telephony, the model choice and the recordings stay on your own hardware.

connect_livekit lands the same day with a different shape: video rooms with guest links, a SIP bridge, a browser phone, and native LiveKit Agents on the same media stack. The core also pulls the active-calls panel and Calls widgets up out of the provider modules, so every backend gets them.