finds.dev← search

// the find

PatterAI/Patter

★ 874 · Python · MIT · updated Jun 2026

Open-source voice-AI SDK. The Vapi/Retell alternative for builders who want to own the stack. Give your AI agent a phone number in 4 lines — Python and TypeScript, MIT licensed, Twilio, Telnyx, and Plivo.

Patter is a voice-AI SDK that handles the full telephony stack — carrier integration, STT, TTS, LLM, VAD — so you can wire up an AI phone agent without touching WebSockets or carrier APIs directly. It targets developers who want to own their infrastructure rather than pay Vapi or Retell markup. Python and TypeScript, both MIT licensed.

Provider abstraction is genuinely useful: swapping Twilio for Telnyx or Deepgram for AssemblyAI is one line, which matters when you're optimizing latency or cost. The three voice modes (Realtime, Pipeline, Hybrid) cover the spectrum from low-latency all-in-one engines to full STT+LLM+TTS pipelines, rather than forcing one architecture on everyone. Built-in LLM fallback chain with mid-call failover is a real production concern that most hobby projects ignore. The bundled Cloudflare quick tunnel for local dev removes the usual ngrok ceremony for webhook testing.

874 stars and 89 forks is thin for a project marketing itself as a production-ready Vapi alternative — the community support and battle-testing that matters for telephony edge cases (DTMF, carrier-specific audio quirks, bad STT transcripts mid-sentence) is unproven at scale. Python 3.11+ hard requirement and the sheer number of optional provider dependencies means your environment can get messy fast; no mention of a lockfile strategy or optional dependency groups in the README. The dashboard is a separate Vite app with its own sync script, which feels bolted on rather than integrated — unclear how it handles concurrent users or auth in production. Telemetry is on by default and requires an explicit opt-out, which will annoy security-conscious teams before they even run their first call.

View on GitHub → Homepage ↗

// want more like this?

We dig through GitHub every week and send a few repos picked for what you actually care about — each with an honest take like this one.

Get finds in your inbox → Search again →