WIP - Add Telnyx voice provider

[whoiskatrin]

Summary

Adds first-class Telnyx support for the Cloudflare voice pipeline, addressing #1367.

This PR introduces a new @cloudflare/voice-telnyx provider package under voice-providers/telnyx and a small examples/telnyx-voice-agent starter app.

What changed

@cloudflare/voice-telnyx

• Adds a Telnyx provider package following the existing voice-providers/* conventions.
• Supports subpath imports so users can import only what they need:
  • @cloudflare/voice-telnyx
  • @cloudflare/voice-telnyx/stt
  • @cloudflare/voice-telnyx/tts
  • @cloudflare/voice-telnyx/telephony
• Ports Telnyx STT support implementing Transcriber / TranscriberSession:
  • Workers fetch-upgrade WebSocket connection
  • 16 kHz mono PCM16 streaming
  • WAV header emission for Telnyx's streaming API
  • interim and final transcript callbacks
• Ports Telnyx TTS support implementing TTSProvider and StreamingTTSProvider:
  • REST backend by default
  • Workers-only WebSocket streaming backend
• Adds Telnyx telephony helpers:
  • TelnyxCallBridge
  • TelnyxPhoneTransport
  • TelnyxPhoneClient
  • TelnyxJWTEndpoint
  • createTelnyxVoiceConfig
• Keeps STT/TTS subpaths independent from telephony imports.
• Adds README docs covering environment variables, subpath imports, browser voice usage, and PSTN/telephony setup.

Example app

• Adds examples/telnyx-voice-agent, a compact starter using:
  • Telnyx STT
  • Telnyx TTS
  • Workers AI for the LLM
  • useVoiceAgent for browser microphone/audio UX
• Includes .env.example with:
  • TELNYX_API_KEY
  • optional TELNYX_CREDENTIAL_CONNECTION_ID for telephony experiments
• Documents that the default example focuses on STT/TTS, while telephony is available through the provider package.

Attribution / dependency metadata

• Adds @telnyx/webrtc for telephony support.
• Updates NOTICE to document that parts of this provider are adapted from Telnyx's @telnyx/voice-cloudflare package, whose npm metadata declares MIT licensing.

Validation

Passed:

npm run build -w @cloudflare/voice-telnyx
npm run build -w @cloudflare/agents-telnyx-voice-agent

Also ran:

npm run typecheck

That currently fails on existing packages/ai-chat type errors unrelated to this PR, including missing exports from agents/chat and AgentConnection.path type mismatches in packages/ai-chat/src/react-tests/use-agent-chat.test.tsx.

Notes for reviewers

• Phone playback via TelnyxPhoneTransport expects server audio as 16 kHz mono PCM16. The README calls out using withVoice(Agent, { audioFormat: "pcm16" }) and client preferredFormat: "pcm16" for PSTN routing.
• The initial example intentionally stays focused on browser STT/TTS to keep the starter small and easy to verify. Telephony APIs are included and documented for users who want to extend it.

Refs #1367

[changeset-bot]

🦋 Changeset detected

Latest commit: 994f090

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 0 packages

When changesets are added to this PR, you'll see the packages that this PR includes changesets for and the associated semver types

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

[devin-ai-integration]

Devin Review found 1 new potential issue.

View 9 additional findings in Devin Review.

[pkg-pr-new]

Open in StackBlitz

agents

npm i https://pkg.pr.new/agents@1461

@cloudflare/ai-chat

npm i https://pkg.pr.new/@cloudflare/ai-chat@1461

@cloudflare/codemode

npm i https://pkg.pr.new/@cloudflare/codemode@1461

hono-agents

npm i https://pkg.pr.new/hono-agents@1461

@cloudflare/shell

npm i https://pkg.pr.new/@cloudflare/shell@1461

@cloudflare/think

npm i https://pkg.pr.new/@cloudflare/think@1461

@cloudflare/voice

npm i https://pkg.pr.new/@cloudflare/voice@1461

@cloudflare/worker-bundler

npm i https://pkg.pr.new/@cloudflare/worker-bundler@1461

commit: 994f090

[Oliver-Zimmerman]

Nice port overall — I think you’ve preserved the core architecture well, especially the split between STT/TTS/telephony and the browser-side PSTN bridge model. My main ask would be to swap the example for one that better shows the real value of the provider: the current example reads mostly like a browser STT/TTS demo, while the distinctive bit here is the phone/PSTN path. I’d suggest porting one of the original phone/hybrid examples instead, e.g. examples/phone-voice-agent: https://github.com/team-telnyx/voice-cloudflare/tree/main/examples/phone-voice-agent”