Repository navigation

Releases

v0.10.0 Latest

Latest

github-actions released this 26 Oct 21:55

Immutable

v0.10.0

efa6e2d

Minor Changes

#85 1840a07 Thanks @polvallverdu! - TTS audio cache (pipedog only): TTSService takes an optional audioCache (TTSAudioCache, an in-memory LRU bounded by size, or any get/set store), and warmAudioCache(text) synthesizes fixed lines ahead of time so they play without a request to the provider. Only warmed lines are stored unless cacheSynthesizedAudio allows live synthesis. Works with services that yield audio from runTts (HTTP and streaming-RPC).
#85 1840a07 Thanks @polvallverdu! - MinWordsUserTurnStartStrategy takes two opt-in options: startOnVadWhenBotQuiet starts the turn
on VAD while the bot is silent instead of waiting for a transcript, and keepBotSpeakingOnTurnStart
keeps the bot-speaking state across turn starts. New user mute strategies:
OpeningUserMuteStrategy (mute until the bot's first speech ends or has played for maxSeconds)
and EndingUserMuteStrategy (mute for good once the conversation is ending).

Patch Changes

Updated dependencies []:
- @pipedog/asyncio@0.10.0
- @pipedog/resampler-wasm@0.10.0
- @pipedog/sentencex-wasm@0.10.0

All @pipedog/* packages are released at 0.10.0:

Assets 3

v0.9.0

github-actions released this 23 Oct 21:25

Immutable

v0.9.0

d57aad8

Minor Changes

#76 72c00c8 Thanks @polvallverdu! - Frames move through a pipeline about twice as fast (per-frame latency through 10 processors: 83 → 40 µs on Node, 370 → 180 µs on Cloudflare Workers), with the same behavior:
- The pipeline worker no longer wakes an observer for events it doesn't handle: onProcessFrame events only go to observers that override onProcessFrame (the default observers only watch pushes).
- New BaseObserver.acceptsPush(data): a synchronous filter the worker calls before queueing a push for an observer. RTVIObserver uses it to skip the repeated pushes it ignores, so it is no longer woken at every hop of every frame.
- FrameProcessor skips the per-frame on_before/after_process_frame and on_before/after_push_frame events when nothing listens to them, and builds its trace log line only when trace logging is on.
#72 f7dc044 Thanks @polvallverdu! - Add defineTool (@pipedog/core/adapters/schemas/standard_tool): typed function-calling tools from any Standard Schema (Zod, Valibot, ArkType). The JSON schema comes from Standard JSON Schema (~standard.jsonSchema) or an explicit jsonSchema, the handler's args are inferred, and arguments are validated before the handler runs; invalid arguments are returned to the LLM as the result. It returns a regular FunctionSchema, so existing APIs are unchanged. Also adds registerTools and the copied Standard Schema interfaces in @pipedog/core/utils/standard_schema.

Patch Changes

Updated dependencies []:
- @pipedog/asyncio@0.9.0
- @pipedog/resampler-wasm@0.9.0
- @pipedog/sentencex-wasm@0.9.0

All @pipedog/* packages are released at 0.9.0:

Assets 3

v0.8.0

github-actions released this 22 Oct 12:46

Immutable

v0.8.0

cc598b7

Patch Changes

#69 d6e6f45 Thanks @polvallverdu! - Fixes for apps bundled with bun build (and bun build --compile) and run without node_modules:
- @pipedog/transport-smallwebrtc loads reflect-metadata first itself (a new dependency, pinned). werift's X.509 code needs it before tsyringe, and a bundle that loads the transport lazily, as @pipedog/runner does, could put the import after it, which disabled the runner's WebRTC routes. Hosts don't import or externalize reflect-metadata; it does not belong in core.
- @pipedog/runner counts its own transports (@pipedog/transport-websocket, @pipedog/transport-smallwebrtc) as available instead of resolving them by name, which fails in a bundle: every telephony and WebSocket route answered 404.
- Silero, Smart Turn and RNNoise read their default model or .wasm file next to the package's source, which a bundle does not copy. The error now names the model option (or initRnnoise()) and the with { type: "file" } import that fixes it, instead of a bare file-not-found for a path nobody recognizes.
- Core's Smart Turn fallback warning says that a bundled app has to pass createLocalSmartTurnAnalyzerV3() to the turn strategy, since a bundler can't see core's by-name import.
Docs: a Bundling page under Deployment, with the tested bun build recipe, a snippet per asset, and what has to stay outside the bundle and why. examples/bundle-check bundles and runs a sample bot per configuration in CI.
Updated dependencies []:
- @pipedog/asyncio@0.8.0
- @pipedog/resampler-wasm@0.8.0
- @pipedog/sentencex-wasm@0.8.0

All @pipedog/* packages are released at 0.8.0:

Assets 3

v0.7.0

github-actions released this 22 Oct 09:42

Immutable

v0.7.0

a61f594

Minor Changes

#68 20a84e7 Thanks @polvallverdu! - Traces say pipedog now. The library's tracers are called pipedog and pipedog.turn (they were
pipecat and pipecat.turn, as upstream), and setupTracing() defaults service.name to
"pipedog" (was "pipecat"). Span names and attribute keys are unchanged.

If a dashboard, alert or query filters on the Pipecat tracer names, bring them back at startup,
before creating a pipeline, with setTracerName("pipecat") or setupTracing({ tracerName: "pipecat" }). getTracerName() and DEFAULT_TRACER_NAME are exported too. Set your own
serviceName in production either way.

Patch Changes

Updated dependencies []:
- @pipedog/asyncio@0.7.0
- @pipedog/resampler-wasm@0.7.0
- @pipedog/sentencex-wasm@0.7.0

All @pipedog/* packages are released at 0.7.0:

Assets 3

v0.6.1

github-actions released this 21 Oct 20:18

Immutable

v0.6.1

41f5012

Patch Changes

Updated dependencies [9f1ce7a]:
- @pipedog/asyncio@0.6.1
- @pipedog/resampler-wasm@0.6.1
- @pipedog/sentencex-wasm@0.6.1

All @pipedog/* packages are released at 0.6.1:

Assets 3

v0.6.0

github-actions released this 21 Oct 18:37

Immutable

v0.6.0

08c367a

Minor Changes

#32 dd4712b Thanks @polvallverdu! - Speech-to-speech providers: Inworld and Ultravox.
- @pipedog/inworld: InworldRealtimeLLMService (Inworld's Realtime API, with its events and
  adapter), and InworldTTSService (WebSocket) and InworldHttpTTSService.
- @pipedog/ultravox: UltravoxRealtimeLLMService, with agent, one-shot and join-URL calls.
- @pipedog/core: ServiceSettings.fromMapping() builds the subclass it is called on again, so
  TTSSettings.fromMapping({ voice_id: "alice" }) sets voice instead of putting every key in
  extra.
Both run on Bun, Node, Deno and Cloudflare Workers.

Patch Changes

Updated dependencies [b53032d]:
- @pipedog/asyncio@0.6.0
- @pipedog/resampler-wasm@0.6.0
- @pipedog/sentencex-wasm@0.6.0

All @pipedog/* packages are released at 0.6.0:

Assets 3

v0.5.0

github-actions released this 21 Oct 15:29

Immutable

v0.5.0

855397c

Patch Changes

#28 70dce6f Thanks @polvallverdu! - GoogleLLMService sends frequencyPenalty and presencePenalty when they are set.
GeminiFileAPI.uploadFile() takes the file's contents (a Uint8Array, an ArrayBuffer, a
Blob or a ReadableStream, with an optional mimeType) as well as a path, so uploads work on
Cloudflare Workers. A path on a runtime without node:fs throws an error that says what to pass
instead. GeminiLiveLLMService logs a warning when it drops an LLMSetToolsFrame. Comments and
docs now say that WebSocket auth headers are sent on Cloudflare Workers.
#20 566fb45 Thanks @polvallverdu! - Package READMEs on npm. @pipedog/core, @pipedog/asyncio, @pipedog/openai,
@pipedog/anthropic, @pipedog/deepgram, @pipedog/cartesia, @pipedog/elevenlabs,
@pipedog/tracing and @pipedog/whisper now ship a README with install steps, a usage snippet,
runtimes and links to the docs. Every package gains npm keywords, and the @pipedog/elevenlabs
description no longer lists Text-to-Dialogue, which is not ported yet.
Updated dependencies [566fb45]:
- @pipedog/asyncio@0.5.0
- @pipedog/resampler-wasm@0.5.0
- @pipedog/sentencex-wasm@0.5.0

All @pipedog/* packages are released at 0.5.0:

Assets 3

v0.4.0

github-actions released this 21 Oct 13:20

Immutable

v0.4.0

8d000b8

Minor Changes

#18 b88231a Thanks @polvallverdu! - Lighter sessions, working default turn detection, and WebSockets with headers on Workers.
- @pipedog/onnx: sharedSession(backend, model, options) shares one inference session per
  backend, model and options; runs are serialized on onnxruntime-web. disposeSharedSessions()
  releases them.
- @pipedog/audio-silero, @pipedog/audio-smart-turn: analyzers share their session by default
  (shareSession: false opts out). Each further call now costs about 1.3 MB instead of about 52 MB.
- @pipedog/core: LocalSmartTurnAnalyzerV3 without createSession loads
  @pipedog/audio-smart-turn when it's installed, on first use. Without it, turns end on the VAD's
  silence, with one warning, instead of failing. On Cloudflare Workers, service websockets connect
  with fetch(), so their auth headers are sent. Warning and error logs mask credentials in their
  message and attached error, for every sink.
- @pipedog/transport-websocket: websocketConnect on Workers connects with fetch() and sends
  additionalHeaders.

Patch Changes

Updated dependencies []:
- @pipedog/asyncio@0.4.0
- @pipedog/resampler-wasm@0.4.0
- @pipedog/sentencex-wasm@0.4.0

All @pipedog/* packages are released at 0.4.0:

Assets 3

v0.3.0

github-actions released this 21 Oct 07:23

Immutable

v0.3.0

42e09c5

Minor Changes

#8 6aa3dac Thanks @polvallverdu! - First providers and transports.
- @pipedog/google: GoogleSTTService (Cloud Speech v2, Chirp 3), GoogleTTSService (Chirp 3 HD
  streaming), GoogleHttpTTSService, GeminiTTSService and GeminiSTTService (Gemini Live,
  gemini-3.5-transcribe-live). The Gemini classes run everywhere, Workers included; the Cloud
  classes need the optional @google-cloud/speech / @google-cloud/text-to-speech peers (Bun,
  Node, Deno).
- @pipedog/langchain: LangchainProcessor for any LangChain Runnable, LangGraph graphs included.
- @pipedog/transport-websocket: WebsocketServerTransport, WebsocketClientTransport,
  FastAPIWebsocketTransport (bring your own socket: Bun, Deno, Node ws, Hono, Cloudflare
  WebSocketPair) and RTVIClientTransport.
- @pipedog/transport-smallwebrtc: SmallWebRTCTransport and its request handler, compatible with
  Pipecat's @pipecat-ai/small-webrtc-transport client (Bun, Node, Deno).
- Serializers, byte-identical to upstream: ProtobufFrameSerializer, RTVIClientSerializer,
  TelnyxFrameSerializer.
- @pipedog/core: utils/py_json (Python's json.dumps/json.loads) and utils/base64. The
  protobuf codec rejects invalid UTF-8 in string fields, and FrameSerializer.shouldIgnoreFrame
  raises for a transport message that is not a dict, as upstream does.

Patch Changes

Updated dependencies []:
- @pipedog/asyncio@0.3.0
- @pipedog/resampler-wasm@0.3.0
- @pipedog/sentencex-wasm@0.3.0

All @pipedog/* packages are released at 0.3.0:

Assets 3

v0.2.0

github-actions released this 20 Oct 22:00

Immutable

v0.2.0

869e2ec

Minor Changes

#4 f59cf26 Thanks @polvallverdu! - SOXRAudioResampler and SOXRStreamAudioResampler now resample instead of throwing "not ported yet" when the rates differ. This also fixes createFileResampler(), createStreamResampler(), the μ-law/A-law helpers at other rates, loadDtmfAudio() at rates other than 8 kHz, LocalSmartTurnAnalyzerV3 at rates other than 16 kHz, and output transports and TTS services whose rates differ. The resampling goes through a small engine interface in @pipedog/core/audio/resamplers/engine:
- The default engine is @pipedog/resampler-wasm, which core now depends on. It loads lazily on first use, and on Cloudflare Workers it loads through the package's workerd export condition.
- If WebAssembly cannot be loaded, core logs one warning and falls back to a pure-TypeScript windowed-sinc engine with soxr-like presets (createSincResamplerEngine()).
- setResamplerEngine(engine) lets apps inject their own engine, for example a native soxr binding. getResamplerEngine() resolves to the engine in use.
The quality presets (VHQ by default, then HQ, MQ, LQ and QQ), clearing after inactivity, flush() and reset() behave as before. Equal rates still return the input unchanged.

@pipedog/resampler-wasm: initResampler() falls back to an embedded copy of the module (src/wasm_bytes.ts) when the .wasm file next to the package cannot be read, for example after bundling.

@pipedog/resampler-wasm: the quality presets now follow libsoxr's quality table. The stopband starts at the Nyquist frequency of the lower rate. MQ, HQ and VHQ are flat to about 0.91 of it and differ mainly in attenuation: about 111, 138 and at least 157 dB. LQ (about 112 dB) and QQ (about 75 dB) have earlier roll-offs. The filter is stretched when downsampling, so it no longer leaks aliases at large ratios: a 5 kHz tone resampled from 48 kHz to 8 kHz is now 87 to 163 dB down, depending on the preset. Processing is now done in f64. Output samples, filter delays (about 8.5 ms at VHQ) and qualityInfo() values change; the output lengths do not.
#4 f59cf26 Thanks @polvallverdu! - Fill four gaps in core that upstream tests exercise:
- Resampling works: SOXRAudioResampler, SOXRStreamAudioResampler and soxrResample run on
  @pipedog/resampler-wasm (now a dependency) instead of throwing, so TTS services, output
  transports and the audio buffer processor can convert sample rates. Call await initSoxr()
  before the synchronous soxrResample; the async resamplers do it themselves.
- encodeImage() JPEG-encodes raw L/LA/RGB/RGBA/RGBX pixels with a built-in baseline
  encoder (utils/jpeg_encoder), so raw image frames reach LLMs and classifiers.
- utils/yaml: includeLoader() / addIncludeConstructor() work, with a PyYAML-style
  SafeLoader (YAML 1.1 resolvers, per-class constructors) and load() on top of the yaml
  package. Included files are read with an injected readFile.
- LocalFileStorage saves, loads, deletes and trims files through an injected fs
  (node:fs/promises fits as is).
#4 f59cf26 Thanks @polvallverdu! - Core is fully ported (all 361 upstream modules): the text transforms that make text speakable for
TTS (numbers, currency, dates, percentages, phone numbers, units, acronyms, email, markdown,
replacements, pronunciations, VoiceFormatter) and the WebSocket proxy client. Numbers are spoken
with num2words-ts, as upstream uses num2words; Unicode digits (e.g. "٥٠") are recognised and
read like upstream does.
#6 72b8cb8 Thanks @polvallverdu! - Local ONNX models. @pipedog/onnx is a small inference backend over onnxruntime-node (native,
fastest) or onnxruntime-web (WASM), both optional peer dependencies; createOnnxBackend() picks
the native one when it is installed. On Cloudflare Workers the workerd export condition runs
onnxruntime-web from its .wasm imported as a module. @pipedog/audio-silero ports
SileroVADAnalyzer with the upstream silero_vad.onnx included, and
@pipedog/audio-smart-turn provides createLocalSmartTurnAnalyzerV3(), core's analyzer wired to
an ONNX backend and the upstream smart-turn-v3.2-cpu.onnx. On Workers both models are passed in
as bytes. In core, VADAnalyzer.voiceConfidence() may now return a promise.
#6 72b8cb8 Thanks @polvallverdu! - OpenTelemetry tracing, as in Pipecat's pipecat-ai[tracing] extra, on Bun, Node, Deno and Cloudflare Workers.

@pipedog/core: @opentelemetry/api is now an optional peer dependency. When it is installed, core imports it on its own (a dynamic import() started when utils/tracing/setup loads, with no top-level await), and isTracingAvailable() turns true, like upstream's is_tracing_available(). Without it, core runs as before and tracing stays a no-op. PipelineWorker awaits that import in its async setup before deciding whether to trace, so construction order does not matter (a worker built at module top level traces too); loadOpenTelemetry() exposes the same wait. registerOpenTelemetry() still registers the modules explicitly and now wins over the automatic import. The structural OpenTelemetry types accept the real @opentelemetry/api and @opentelemetry/sdk-trace-base modules (array attribute values may hold null/undefined, as in OpenTelemetry). Span names, attributes and hierarchy are unchanged and match upstream's.

@pipedog/tracing (new): setupTracing({ serviceName, exporter, consoleExport }), matching upstream's setup_tracing(): it registers @opentelemetry/api, @opentelemetry/sdk-trace-base and @opentelemetry/resources with core and installs a global tracer provider with batch span processors for the given exporter and, optionally, the console. It re-exports OTLPTraceExporter (OTLP/HTTP JSON, from @opentelemetry/exporter-trace-otlp-http), ConsoleSpanExporter, InMemorySpanExporter and the span processors. forceFlushTracing() and shutdownTracing() export what the batch processors still hold, e.g. ctx.waitUntil(forceFlushTracing()) on Cloudflare Workers. @opentelemetry/api is a peer dependency.

Patch Changes

#7 122b72f Thanks @polvallverdu! - @pipedog/asyncio now schedules like asyncio, so ported code interleaves exactly as it does in
Python. Waking a waiter (a future's result, Event.set(), Queue.put(), a task finishing, a
timer) no longer resumes it on the next microtask: the wakeup goes on a FIFO ready queue
(callSoon), and queued wakeups run one at a time, each once the microtask queue is empty. A task
therefore runs until it awaits something that is not done; get() on a non-empty queue, put() on
an unbounded one, wait() on a set event and awaiting a done task do not let any other task run,
and sleep(0) yields exactly one turn. The turn is driven by a MessageChannel on Node,
setImmediate on Bun and Deno, and a body read on Cloudflare Workers.
- New: Future (asyncio.Future: setResult, setException, cancel, done callbacks; awaitable),
  callSoon / callLater returning a cancellable Handle, isFuture. Task extends Future.
- Event, Queue (and PriorityQueue, LifoQueue), Lock, Semaphore, gather, wait,
  shield, waitFor and timeout are ported from CPython on top of Future, with its wakeup
  rules: a woken get() takes whatever is at the head when it runs (so a PriorityQueue hands
  out the smallest item of a burst), a cancellation that arrives after the wakeup was queued wins,
  gather raises a cancellation once its children are done, and addDoneCallback callbacks run
  on the ready queue.
- waitFor(fn, t) and timeout(t, fn) run fn in the current task, like Python 3.12+: on
  expiry the task is cancelled and the CancelledError becomes TimeoutError. When every
  cancellation request has been withdrawn with uncancel(), task.signal is a fresh signal.
- Queue uses O(1) removal from the front (it used Array.shift(), which is O(n) on V8 and made
  long queues quadratic on Node, Deno and Workers).
- 32 interleaving scenarios recorded from CPython 3.14 (test/python/scenarios.py) are replayed in
  the tests and in the compat smoke test on all four runtimes.
@pipedog/core: LLMAssistantAggregator runs the LLM once for a burst of
LLMMessagesAppendFrame / FunctionCallResultFrame instead of once per frame. Its
hasQueuedFrame(...) check now sees the later frames of the burst queued, as upstream does, so
there are no more duplicate responses. Other frame orderings that depend on asyncio scheduling
(system frames overtaking queued data frames, bus priority, interruption handling) now match
upstream too.
Updated dependencies [122b72f, [f59cf26](h...

Assets 3

v0.10.0

Minor Changes

Patch Changes

Uh oh!

v0.9.0

Minor Changes

Patch Changes

Uh oh!

v0.8.0

Patch Changes

Uh oh!

v0.7.0

Minor Changes

Patch Changes

Uh oh!

v0.6.1

Patch Changes

Uh oh!

v0.6.0

Minor Changes

Patch Changes

Uh oh!

v0.5.0

Patch Changes

Uh oh!

v0.4.0

Minor Changes

Patch Changes

Uh oh!

v0.3.0

Minor Changes

Patch Changes

Uh oh!

v0.2.0

Minor Changes

Patch Changes

Uh oh!