Distributable packages so any codebase can embed Telenow voice AI. Isolated from
the app (voice_ai_frontend) and backend (voice_ai_rust) — these are the
shippable artifacts, built from the same primitives.
Two buckets:
- Client / audio (needs real per-platform audio I/O): web, React, React
Native, Swift/iOS, Kotlin/Android, Flutter. Heavy parts share one Rust core.
All expose the same
TelenowCall-style controller: session init (public slug, client token, or a backend-minted session — recommended), reconnect, mic capture (μ-law 8 kHz default, matching the live server), jitter-buffered playback, barge-in flush, transcripts, latency ping echo. - Backend / control (REST/WS + token mint + webhook verify + AI and manual/softphone call placement + text chat (Chat API) + Custom API SSE helpers for bring-your-own-LLM endpoints): Node, Python/Django, and an OpenAPI spec to generate the long tail (Kotlin, Go, Ruby, PHP, C#…).
Embed phone calls in your CRM. Place AI agent calls, or manual/softphone
telephony calls — your backend sends a to (customer) and from (your
caller-ID) number, Telenow bridges the carrier leg to a human on a browser/app
softphone (the client SDK), and webhooks deliver the recording
and full call data when it ends. See server-node /
server-python → "Manual / softphone calls".
| Package | Dir | Registry | Bucket | Status |
|---|---|---|---|---|
@telenow/client |
client-web/ |
npm | client | ✅ builds (14 tests incl. wire protocol) |
@telenow/react |
react/ |
npm | client | ✅ builds |
@telenow/server |
server-node/ |
npm | backend | ✅ builds (14 tests) |
telenow (py) |
server-python/ |
PyPI | backend | ✅ 13 tests |
| OpenAPI + generated clients | openapi/ |
(multi) | backend | ✅ spec |
telenow-audio-core |
audio-core/ |
crates.io / native libs | client core | ✅ 11 tests |
TelenowSDK (swift) |
swift/ |
SwiftPM / CocoaPods | client | ✅ builds + DSP verified |
ai.telenow:sdk |
android/ |
Maven | client | 🟢 complete source |
telenow (flutter) |
flutter/ |
pub.dev | client | 🟢 complete source |
@telenow/react-native |
react-native/ |
npm | client | ✅ packaged (dist + podspec + gradle autolink; device smoke-test pending) |
✅ = builds/tests in this repo today (Node, Rust, Python, and Swift toolchains were available). 🟢 = complete, idiomatic source — the DSP + jitter buffer are ported from the verified reference and the audio I/O is fully written — but the toolchain to compile it (Android SDK/NDK, Flutter, an RN app) isn't present here, so it compiles on its own platform rather than in this repo's CI.
Naming is final: npm scope @telenow, PyPI telenow, crates.io
telenow-audio-core, Swift TelenowSDK (public repo TelenowAI/telenow-swift),
Maven ai.telenow — register exactly these (fallbacks in RELEASING.md §1 if a
name is taken). Public source home: TelenowAI/VOICE_AI_SDKs.
See RELEASING.md for how to publish each one, and
INTEGRATION_CHECKLIST.md for wiring the backend
hand-offs (client tokens, init-call, HD audio, keep-session-warm, scale) into the
live voice_ai_rust.
Telenow is a voice AI platform for building production-grade phone and web agents. Pick a brain from the built-in LLM/STT/TTS providers (or bring your own model and carrier), give the agent a prompt, tools, and knowledge, and put it on a phone number, your website, or your app. Every call comes with recordings, transcripts, analytics, warm transfer to humans, outbound campaigns, and webhooks.
- Website: telenow.ai
- Documentation: telenow.ai/docs
- This SDK's guide: telenow.ai/docs/sdk-overview