Compare commits

...

356 Commits

Author SHA1 Message Date
filipi87
16133a2323 Removing the custom prompt. 2026-04-01 16:05:09 -03:00
filipi87
9d815cb5d2 Merge branch 'filipi/async_tools' into filipi/async_tools_structured_data 2026-04-01 15:50:35 -03:00
filipi87
2d87edac18 Merge branch 'main' into filipi/async_tools 2026-04-01 15:49:43 -03:00
filipi87
bce07e0c76 Merge branch 'filipi/async_tools' into filipi/async_tools_structured_data 2026-04-01 15:48:22 -03:00
filipi87
59092fe4fe Renaming the examples to match main. 2026-04-01 15:42:50 -03:00
filipi87
d515a81073 Updating the Anthropic example to use async function calls. 2026-04-01 15:31:32 -03:00
kompfner
a3c7f6c2af Merge pull request #4215 from pipecat-ai/pk/remove-openaillmcontext
Remove deprecated `OpenAILLMContext` as well as everything (code path…
2026-04-01 14:03:35 -04:00
Paul Kompfner
df68665ec1 Add changelog entries for OpenAILLMContext removal 2026-04-01 14:03:08 -04:00
filipi87
e23cb46885 Trying to structure async tool responses and improve the LLM prompt to teach it how to handle them. 2026-04-01 14:48:09 -03:00
Harshita Jain
bd6cbd7fe7 feat: add Smallest AI STT service integration (#4162)
Add SmallestSTTService using the Pulse WebSocket API for real-time
transcription. Includes SmallestSTTSettings dataclass, 32-language
support with resolve_language fallback, VAD-driven finalize signal,
and SMALLEST_TTFS_P99 latency constant.   

Also adds X-Source and X-Pipecat-Version headers to Smallest STT
and TTS WebSocket connections.
2026-04-01 13:44:04 -04:00
filipi87
72bbad51b7 Added group_parallel_tools parameter to LLMService. 2026-04-01 13:51:30 -03:00
filipi87
c066a913fe Adding changelogs for all the fixes. 2026-04-01 12:20:58 -03:00
filipi87
63bbfc3b27 Creating the concept of a group_id for the function calls. 2026-04-01 12:05:09 -03:00
filipi87
2458b9d42b Delaying the response for the get_current_weather in the openai example. 2026-04-01 10:47:29 -03:00
filipi87
4543aef3d9 Only pushing a context frame when we receive the function call result if the user is not speaking. 2026-04-01 10:45:00 -03:00
filipi87
260368b6f4 Fixing an issue where the BotOutputTransport was discarding the UninterruptibleFrames. 2026-04-01 10:32:11 -03:00
filipi87
3ad2675b24 Creating UninterruptibleProcessQueue. 2026-04-01 10:28:52 -03:00
filipi87
970d713d7a Using a JSON to send the result. 2026-04-01 10:28:03 -03:00
Mark Backman
33ef6b3174 Merge pull request #4218 from pipecat-ai/mb/rename-all-examples
Rename all examples
2026-04-01 07:15:57 -04:00
Mark Backman
3ca656cae5 Update simli name to match others 2026-03-31 22:54:21 -04:00
Mark Backman
6a84d02156 Update evals
- Removed evals for removed services
- Added eval for function-calling-deepseek.py
2026-03-31 22:13:52 -04:00
Mark Backman
080da8b94c Update eval script paths to match renamed example files 2026-03-31 22:09:42 -04:00
Mark Backman
d3021b4590 Rename example files to prepend parent folder name, preventing package shadowing
Example files like openai.py shadow installed packages when Python adds the
script directory to sys.path. Prepend the parent folder name to each example
file (e.g. openai.py -> function-calling-openai.py). Also split
thinking-and-mcp/ into separate mcp/ and thinking/ directories.
2026-03-31 22:06:01 -04:00
Paul Kompfner
92e34ea6e8 Fix potential UnboundLocalError for system_message in tracing decorator
Restore the `system_message = None` initialization that was dropped
when collapsing the OpenAILLMContext branch.
2026-03-31 21:00:51 -04:00
Paul Kompfner
ebab75765d Fix stream cancellation tests to mock get_chat_completions
The tests were mocking the removed _stream_chat_completions_*_context
methods. Update them to mock get_chat_completions instead.
2026-03-31 18:54:23 -04:00
Paul Kompfner
110c88bf92 Remove stale re-export of deleted google.openai subpackage 2026-03-31 18:53:55 -04:00
Paul Kompfner
19e521b75a Simplify LLMContextFrame handling in process_frame methods
Now that LLMContextFrame is the only frame that provides a context,
remove the intermediate `context = None` / `if context:` pattern
and handle context processing directly in the isinstance branch.
2026-03-31 18:35:48 -04:00
Paul Kompfner
394599d031 Remove deprecated OpenAILLMContext as well as everything (code paths or whole types) dependent on it (all of which were also deprecated) 2026-03-31 18:15:25 -04:00
filipi87
f7012c570c Fixed an issue in the FrameProcessor where only the current frame was checked for being an UninterruptibleFrame, not other frames in the queue. 2026-03-31 18:38:11 -03:00
filipi87
4bfa084f77 Updating the openai example to be async. 2026-03-31 17:37:39 -03:00
filipi87
780d6c476d Merge branch 'main' into filipi/async_tools 2026-03-31 17:36:40 -03:00
filipi87
dfdb92958b Fix async tool handling for compatibility with all LLMs. 2026-03-31 17:26:06 -03:00
mattie ruth backman
0f47076703 More RTVI version parsing improvements 2026-03-31 16:05:53 -04:00
mattie ruth backman
3e255f3d21 improve version format check 2026-03-31 16:05:53 -04:00
mattie ruth backman
565b9b961d add tests for rtvi versioning 2026-03-31 16:05:53 -04:00
mattie ruth backman
692c3c74d1 We should now expect clients to be version 1.0.0 with valid versioning info 2026-03-31 16:05:53 -04:00
Mark Backman
7d309b3340 Merge pull request #4208 from pipecat-ai/mb/remove-deprecated-services
Remove deprecated service module shims
2026-03-31 15:37:12 -04:00
Mark Backman
04e8444096 Add changelog for #4208 2026-03-31 15:34:16 -04:00
Mark Backman
7501effad5 Remove deprecated service module shims and old implementations
Delete deprecated import shims that only re-export from new locations:
- services/ai_services.py
- services/gemini_multimodal_live/
- services/aws_nova_sonic/
- services/openai_realtime/
- services/deepgram/{stt,tts}_sagemaker.py
- services/google/{llm_openai,llm_vertex,google}.py
- services/google/gemini_live/llm_vertex.py
- services/riva/
- services/nim/

Remove deprecated implementations replaced by newer services:
- services/openai_realtime_beta/ (use openai.realtime)
- services/google/openai/ (use google.llm)

Also removes associated examples and tests for deleted services.
2026-03-31 15:34:14 -04:00
Mark Backman
0c8ff9c4c3 Merge pull request #4209 from pipecat-ai/mb/grok-3-default
Change GrokLLMService default model to grok-3
2026-03-31 15:29:34 -04:00
Mark Backman
53f6426b0b Merge pull request #4216 from pipecat-ai/mb/add-missing-google-vertex
Add missing google-vertex.py file
2026-03-31 15:29:04 -04:00
Mark Backman
9e32ade44b Merge pull request #4203 from pipecat-ai/mb/fix-json-decode-tool-calls
Handle incomplete function call arguments from interrupted LLM streams
2026-03-31 15:28:53 -04:00
Mark Backman
2574d24400 Merge pull request #4202 from pipecat-ai/mb/fix-inworld-tts-streaming-utf8
Fix UTF-8 decode error in Inworld TTS streaming response
2026-03-31 15:28:37 -04:00
Mark Backman
27cb078716 Add missing google-vertex.py file 2026-03-31 15:25:52 -04:00
Mark Backman
ca636813a8 Merge pull request #4206 from pipecat-ai/mb/flatten-examples-dir
Move foundational examples to examples/
2026-03-31 15:23:49 -04:00
Mark Backman
47b41a0ff7 Rename services/ to voice/ and function-calling/, flatten to top level
Replace the nested services/speech/ and services/function-calling/ with
top-level voice/ and function-calling/ directories. Update eval script
paths and README to match.
2026-03-31 15:20:03 -04:00
Mark Backman
f14638a1fd Revert "Flatten services/ nesting: promote speech and function-calling to top level"
This reverts commit e1939ecd44.
2026-03-31 14:59:23 -04:00
Mark Backman
e1939ecd44 Flatten services/ nesting: promote speech and function-calling to top level
Move services/speech/* directly into services/ and services/function-calling/*
into top-level function-calling/. Update eval script paths and README.
2026-03-31 14:55:22 -04:00
Mark Backman
dc5b94f9e0 Merge pull request #4213 from pipecat-ai/mb/google-imagen-4
Update default Google Imagen model to imagen-4.0
2026-03-31 13:20:20 -04:00
Mark Backman
1d85aedcae Split features/ into audio/, observability/, and rag/ subfolders
Extract focused example groups from the catch-all features/ folder:
- audio/: audio recording, background sound, sound effects
- observability/: observer, heartbeats, sentry metrics
- rag/: mem0, gemini-rag, gemini grounding metadata

Update README to document the new folders.
2026-03-31 13:15:06 -04:00
Mark Backman
e719cbbe6d Reorganize examples into topic-based subfolders
Move 304 examples from a flat numbered directory into 14 descriptive
subfolders: getting-started, services (speech + function-calling),
transcription, vision, realtime, persistent-context,
context-summarization, update-settings (stt/tts/llm), turn-management,
thinking-and-mcp, transports, video-avatar, video-processing, and
features.

Strip numbered prefixes from filenames (e.g. 07c-interruptible-deepgram.py
becomes services/speech/deepgram.py) since the folder context makes them
redundant. Keep numbered prefixes only in getting-started/ where ordering
matters.

Update eval script paths and README to match the new structure.
2026-03-31 13:12:24 -04:00
Mark Backman
f2ce7ececc Move foundational examples to examples/ 2026-03-31 13:12:24 -04:00
kompfner
bd7496fa27 Merge pull request #4211 from pipecat-ai/pk/openai-responses-websocket-service-refactor
Introduce WebsocketLLMService and refactor OpenAIResponsesLLMService …
2026-03-31 13:02:45 -04:00
Paul Kompfner
0a8bcf58c4 Register on_connection_error event handler in WebsocketLLMService 2026-03-31 10:52:33 -04:00
Paul Kompfner
0fb45c6114 Guard _drain_cancelled_response against None websocket 2026-03-31 10:32:47 -04:00
Paul Kompfner
657a5def57 Use consistent 'inference' terminology in error messages 2026-03-31 10:17:29 -04:00
Paul Kompfner
30903042e5 Work around OpenAI Python SDK temperature bug in example 2026-03-31 10:16:30 -04:00
Mark Backman
9936ec16cb Add changelog for #4213 2026-03-31 09:28:31 -04:00
Mark Backman
212aff15c9 Update default Google Imagen model to imagen-4.0-generate-001 2026-03-31 09:16:24 -04:00
Paul Kompfner
f2b3f87661 Clarify discrete vs continuous contrast in WebsocketLLMService docstring 2026-03-30 23:46:23 -04:00
Paul Kompfner
77cfb181f6 Clarify per-inference helper usage in WebsocketLLMService docstring 2026-03-30 23:25:56 -04:00
Paul Kompfner
0b256936c6 Add ConnectionClosed to _receive_response_events raises docstring 2026-03-30 23:14:45 -04:00
Paul Kompfner
3922963c7a Extract helpers in _process_context to reduce repeated code 2026-03-30 23:10:38 -04:00
Paul Kompfner
ab9f2a35b6 Clean up TTFB metrics and previous_response state on inference failure 2026-03-30 23:04:06 -04:00
Paul Kompfner
f19d1183d8 Clean up TTFB metrics and previous_response state on retry failure 2026-03-30 23:00:22 -04:00
Paul Kompfner
9ad4fe6344 Use concrete inference language instead of abstract transaction terminology 2026-03-30 22:42:40 -04:00
Paul Kompfner
04882f6f2a Simplify _connect_websocket guard and remove unused State import 2026-03-30 22:32:08 -04:00
Paul Kompfner
712e42533d Introduce WebsocketLLMService and refactor OpenAIResponsesLLMService to use it
Add WebsocketLLMService as a base class for WebSocket-based LLM services,
parallel to WebsocketTTSService/WebsocketSTTService but codifying a
transactional request-response model rather than a continuous background
receive loop.

WebsocketLLMService provides:
- Connection lifecycle (start/stop/cancel → connect/disconnect)
- _ws_send/_ws_recv with transparent ConnectionClosed handling
  (auto-reconnect via exponential backoff → WebsocketReconnectedError)
- _ensure_connected with retry via _try_reconnect

OpenAIResponsesLLMService now inherits from WebsocketLLMService, removing
duplicated connection management code (_connect, _disconnect, _reconnect,
_ensure_connected, _ws_send, start, stop, cancel) and simplifying
_process_context from a loop with attempt tracking to a flat try/except
with a single retry.
2026-03-30 22:26:31 -04:00
Mark Backman
7d8b436018 Add changelog for #4209 2026-03-30 21:40:17 -04:00
Mark Backman
bf1856f610 Change GrokLLMService default model from grok-3-beta to grok-3
The grok-3 model is now generally available, so update the default
from the beta variant.
2026-03-30 21:39:33 -04:00
Mark Backman
248e0a4c90 Merge pull request #4207 from pipecat-ai/mb/remove-krisp
Remove docs uses of krisp optional dependency
2026-03-30 19:54:14 -04:00
Mark Backman
89dcd57577 Remove docs uses of krisp optional dependency 2026-03-30 19:50:40 -04:00
Mark Backman
32022a952e Merge pull request #4205 from pipecat-ai/mb/remove-quickstart
Remove quickstart example from repo
2026-03-30 18:58:49 -04:00
Aleix Conchillo Flaqué
65d9fcc315 Merge pull request #4204 from pipecat-ai/aleix/remove-some-deprecations
Remove deprecated APIs and modules
2026-03-30 15:32:53 -07:00
Mark Backman
b78ae40d3c Remove quickstart example from repo 2026-03-30 18:20:41 -04:00
Aleix Conchillo Flaqué
ece4d0661e update uv.lock 2026-03-30 15:06:05 -07:00
Aleix Conchillo Flaqué
82a852c1ff Add changelog for #4204 2026-03-30 15:06:05 -07:00
Aleix Conchillo Flaqué
5be1b9c8cb LLMService: remove deprecated request_image_frame() 2026-03-30 15:06:05 -07:00
Aleix Conchillo Flaqué
7913d4e188 FrameProcessor: remove deprecated wait_for_task() 2026-03-30 14:45:42 -07:00
Aleix Conchillo Flaqué
c8dd7c2b57 rtvi: remove old deprecations 2026-03-30 14:44:32 -07:00
Aleix Conchillo Flaqué
77e5f4acc1 runner(daily): remove deprecated configure_with_args() 2026-03-30 14:31:39 -07:00
Aleix Conchillo Flaqué
be8d4dfd87 TTSService: remove deprecated say() function 2026-03-30 14:29:30 -07:00
Aleix Conchillo Flaqué
bb2c60a998 transports: remove deprecated vad_enabled and vad_audio_passthrough 2026-03-30 14:28:34 -07:00
Aleix Conchillo Flaqué
7c644ed810 RTVIObserver: remove deprecated errors_enabled 2026-03-30 14:26:53 -07:00
Aleix Conchillo Flaqué
96ceec2a43 transports: remove deprecated camera_in_* and camera_out_* params 2026-03-30 14:24:40 -07:00
Aleix Conchillo Flaqué
d249473f0b AudioBufferProcessor: remove deprecated user_continuous_stream 2026-03-30 14:22:21 -07:00
Aleix Conchillo Flaqué
1da2018c85 PipelineTask: remove deprecated on_pipeline_ended/cancelled/stopped 2026-03-30 14:20:45 -07:00
Aleix Conchillo Flaqué
af126ec7cf PipelineParams: remove deprecated observers field 2026-03-30 14:18:07 -07:00
Aleix Conchillo Flaqué
340e58bf5c LLMService: remove old function call single argument 2026-03-30 14:16:18 -07:00
Aleix Conchillo Flaqué
7873159d0f LLMService: remove start_callback 2026-03-30 14:13:23 -07:00
Aleix Conchillo Flaqué
c783101741 frames: remove deprecated interruption frames 2026-03-30 14:08:42 -07:00
Aleix Conchillo Flaqué
73b8bbf963 frames: remove deprecated transport frames 2026-03-30 14:08:24 -07:00
Aleix Conchillo Flaqué
ebbe5acc8f frames: remove deprecated KeypadEntryFrame 2026-03-30 14:07:54 -07:00
Aleix Conchillo Flaqué
dd1bea2a5f audio(turn): remove FalSmartTurnAnalyzer and LocalSmartTurnAnalyzer 2026-03-30 14:04:29 -07:00
Aleix Conchillo Flaqué
136e6a58be audio(utils): remove create_default_resampler 2026-03-30 14:02:13 -07:00
Aleix Conchillo Flaqué
f0d04dde1c audio(filters): remove KrispFilter 2026-03-30 14:01:06 -07:00
Aleix Conchillo Flaqué
742a278c05 audio(filters): remove NoisereduceFilter 2026-03-30 13:58:35 -07:00
Aleix Conchillo Flaqué
b16befc9e9 transports(daily): remove deprecated frames 2026-03-30 13:56:25 -07:00
kompfner
0c11eb6fd0 Merge pull request #4141 from pipecat-ai/pk/openai-responses-websocket-service
feat: add WebSocket-based OpenAI Responses LLM service
2026-03-30 15:25:32 -04:00
Mark Backman
ea39389e03 Add changelog for #4203 2026-03-30 14:24:49 -04:00
Mark Backman
4adf0fd585 Handle incomplete function call arguments from interrupted LLM streams
When a user interruption causes the LLM chunk stream to exit early,
function call arguments may be incomplete JSON. Wrap json.loads() in
try/except JSONDecodeError to skip malformed function calls with a
warning instead of crashing. Fixes #2461.
2026-03-30 14:24:04 -04:00
Mark Backman
465b9bcbc6 Add changelog for #4202 2026-03-30 14:16:21 -04:00
Mark Backman
3f4814cf84 Fix UTF-8 decode error in Inworld TTS streaming response
Buffer raw bytes and only decode after splitting on newline boundaries,
preventing multi-byte UTF-8 characters from being split at chunk edges.

Fixes #3538
2026-03-30 14:15:06 -04:00
Paul Kompfner
0efef19d60 Fix code review issues in WebSocket Responses service
- Use finally block in _disconnect to ensure state is always cleaned
  up, even if websocket.close() throws — prevents stale cancellation
  state (e.g. _cancel_pending_response) from polluting a new connection
- Catch ConnectionClosed in _drain_cancelled_response alongside
  TimeoutError — prevents _needs_drain from staying True and bricking
  the service on every subsequent inference attempt
- Fall back to OPENAI_API_KEY env var when api_key is not passed,
  since the WebSocket connection uses raw websockets (not the
  AsyncOpenAI client which handles this automatically)
- Use _clear_cancellation_state() instead of piecemeal resets where
  appropriate
2026-03-30 10:54:47 -04:00
Mark Backman
87b8f38a48 Merge pull request #4198 from pipecat-ai/mb/readme-update-2026-03-30
Add missing services to README available services table
2026-03-30 10:46:52 -04:00
Mark Backman
e1a3ddbb57 Add missing services to README available services table
Adds Kokoro (TTS), LiveKit and WhatsApp (Transport), Genesys
(Serializers), and Krisp Viva and RNNoise (Audio Processing).
2026-03-30 10:06:14 -04:00
Paul Kompfner
b5683556d4 Remove duplicate entries in run-release-evals.py, which appeared after a rebase 2026-03-30 10:03:43 -04:00
Paul Kompfner
26f85687d6 Handle response cancellation by draining before next inference
Instead of trying to filter stale events inline (unreliable — the API
doesn't provide a way to correlate events to a specific response),
drain remaining events from a cancelled response before starting the
next one. On cancellation, send response.cancel and set a drain flag.
At the start of the next _process_context, read and discard events
until a terminal event arrives, ensuring a clean connection. Falls
back to reconnecting if draining times out.
2026-03-30 09:59:03 -04:00
Paul Kompfner
670ce30a1c Document why HTTP variant doesn't use previous_response_id
Over HTTP, previous_response_id requires store=True (30-day OpenAI-side
conversation storage). The WebSocket variant avoids this via a
connection-local in-memory cache that works with store=False. Add
comments explaining this in both class docstrings, at the store=False
parameter, and in the adapter's previous_response_id note.
2026-03-30 09:59:03 -04:00
Paul Kompfner
1c8d31de70 Add trace logging for previous_response_id decisions and fix example
Add detailed trace-level logging to _apply_previous_response_optimization
showing why the optimization was applied or fell back to full context,
including the relevant data for debugging.

Use append_to_context=False for the filler TTSSpeakFrame in the
function-calling example to avoid altering the conversation history
and breaking the previous_response_id prefix match.
2026-03-30 09:59:03 -04:00
Paul Kompfner
9defff2a34 Skip server-known output items in previous_response_id optimization
When using previous_response_id, the server already knows its own
output from the previous response. Store the raw response output and,
on the next call, compare it against the items following the matched
input prefix — checking role and text content for messages, and call_id
for function calls. If the items match, skip them and send only truly
new input (user messages, tool results). Falls back to full context if
either the prefix or the output comparison fails.
2026-03-30 09:59:03 -04:00
Paul Kompfner
59d28f9fd2 Add changelog for WebSocket OpenAI Responses service 2026-03-30 09:59:03 -04:00
Paul Kompfner
f2a8a9e753 Add WebSocket-based OpenAI Responses LLM service with previous_response_id optimization
Introduce a WebSocket variant of the OpenAI Responses API service that
maintains a persistent connection to wss://api.openai.com/v1/responses
for lower-latency inference. The WebSocket variant automatically uses
previous_response_id to send only incremental context when possible,
falling back to full context on reconnection or cache miss.

The WebSocket variant becomes the new default OpenAIResponsesLLMService,
and the HTTP variant is renamed to OpenAIResponsesHttpLLMService. Both
share a private base class with common settings, parameter building,
and run_inference (always HTTP) logic.
2026-03-30 09:58:56 -04:00
Mark Backman
d1eb2699f3 Merge pull request #4192 from pipecat-ai/mb/update-langchain
Update langchain dependencies to latest major versions
2026-03-30 08:54:41 -04:00
Mark Backman
2e0f5fc6e9 Merge pull request #4194 from pipecat-ai/mb/update-community-integrations-package-convention
Add pipecat-{vendor} package naming convention to community guide
2026-03-30 08:52:28 -04:00
Mark Backman
dd3ca6fbba Merge pull request #4191 from pipecat-ai/mb/remove-openpipe
Remove OpenPipe integration
2026-03-30 08:52:14 -04:00
Mark Backman
171692aa30 Add pipecat-{vendor} package naming convention to community guide
Formalizes the package naming pattern that most community contributors
already follow organically, improving discoverability on PyPI.
2026-03-29 12:39:20 -04:00
Mark Backman
81ddd103f9 Fix KeyError on context messages without role in RTVI observer
Use dict.get() instead of direct key access to handle context messages
that don't have a 'role' key, such as tool results.
2026-03-29 10:28:00 -04:00
Mark Backman
8c9e189394 Fix langchain imports for langchain 1.x compatibility
ChatPromptTemplate moved from langchain.prompts to langchain_core.prompts
in langchain 1.x.
2026-03-29 10:27:48 -04:00
Mark Backman
b6579dc763 Update uv lock with latest versions of Pygments and cryptography 2026-03-29 10:20:45 -04:00
Mark Backman
abd63336e4 Add changelog for #4192 2026-03-29 10:18:52 -04:00
Mark Backman
ccb9dc20f8 Update langchain dependencies to latest major versions
Update langchain 0.3→1.2, langchain-community 0.3→0.4, and
langchain-openai 0.3→1.1. This also unblocks openai>=2.26 which
was previously constrained by the now-removed openpipe package.
2026-03-29 10:17:28 -04:00
Mark Backman
2177e28ee1 Remove OpenPipe integration
OpenPipe was acquired by CoreWeave in September 2025. The Python package
hasn't been updated since June 2025 and the repo since 2024. The openpipe
package caps openai<=1.97.1, creating dependency conflicts with other
extras. Remove the dead integration to clean up the codebase.
2026-03-29 10:12:35 -04:00
Mark Backman
3eb7c2bcd9 Merge pull request #4187 from OmerCohenAviv/fix/heartbeat-monitor-configurable
Fix heartbeat monitor timeout not respecting custom heartbeat interval
2026-03-29 09:31:12 -04:00
Mark Backman
878940f94e Merge pull request #4189 from Arindam200/main
Add NebiusLLMService for Nebius Token Factory
2026-03-29 09:03:06 -04:00
Mark Backman
a3aeafcb2d Alphabetize nebius entry in pyproject.toml extras 2026-03-29 08:58:01 -04:00
Mark Backman
63254fe337 Add NebiusLLMService with developer role and tool support fixes
- Add Nebius LLM service wrapping OpenAI-compatible Token Factory API
- Set supports_developer_role = False (Nebius rejects developer role)
- Default to openai/gpt-oss-120b model (supports function calling)
- Add Nebius function-calling example and env.example entry
- Fix Sarvam developer role support
- Update examples to use developer role for intro messages
2026-03-29 08:50:11 -04:00
Arindam200
39919f7889 Add NebiusLLMService for Nebius Token Factory
Adds an OpenAI-compatible LLM service for Nebius Token Factory, supporting
open-source models (Meta Llama, Qwen, DeepSeek) via their OpenAI-compatible
REST API at https://api.tokenfactory.nebius.com/v1/.
2026-03-29 14:35:46 +05:30
OmercohenAviv
f2e0f5d20c move wait_time out of loop 2026-03-29 00:05:21 +03:00
OmercohenAviv
2724ef6d6f non optional 2026-03-28 12:12:02 +03:00
OmercohenAviv
33fb8852e6 ruff 2026-03-28 12:05:30 +03:00
OmercohenAviv
5fe48da2fb Merge branch 'main' into fix/heartbeat-monitor-configurable 2026-03-28 11:57:23 +03:00
OmercohenAviv
dccd98ec8a test 2026-03-28 11:53:51 +03:00
Aleix Conchillo Flaqué
a84c69858e Merge pull request #4185 from pipecat-ai/changelog-0.0.108
Release 0.0.108 - Changelog Update
2026-03-27 21:47:53 -07:00
aconchillo
ca224219dc Update changelog for version 0.0.108 2026-03-27 21:43:37 -07:00
Aleix Conchillo Flaqué
83dc979d19 Merge pull request #4186 from pipecat-ai/mb/fix-websocket-disconnect-race-condition
Fix FastAPI WebSocket disconnect race condition
2026-03-27 21:40:21 -07:00
Aleix Conchillo Flaqué
fc76b3f2fb update pyproject.toml and uv.lock 2026-03-27 21:36:03 -07:00
Mark Backman
4670370dbb Add changelog for #4186 2026-03-28 00:02:44 -04:00
Mark Backman
47e53890e3 Fix FastAPI WebSocket disconnect race condition causing pipeline hang
When the remote side disconnects while send() is in flight, send() was
setting _closing=True. This prevented the receive loop from firing
on_client_disconnected, causing the pipeline to hang waiting for a
disconnect signal that never came.

The fix removes _closing from send() (that flag means we initiated the
close) and instead checks Starlette application_state in _can_send()
to suppress subsequent sends after a failure.

Fixes #3912
2026-03-28 00:01:25 -04:00
Aleix Conchillo Flaqué
195180b6f4 Merge pull request #4184 from pipecat-ai/aleix/fix-sarvam-examples-role
Fix Sarvam examples to use 'user' role instead of 'developer'
2026-03-27 20:34:59 -07:00
Aleix Conchillo Flaqué
8b64166bb7 Fix Sarvam examples to use 'user' role instead of 'developer'
Sarvam uses the OpenAI-compatible API but does not support the
'developer' role, causing errors. Use 'user' role instead.
2026-03-27 20:33:25 -07:00
Aleix Conchillo Flaqué
1d18995435 Merge pull request #4183 from pipecat-ai/aleix/fix-task-scheduling
Yield after create_task to ensure timer tasks are scheduled
2026-03-27 20:32:32 -07:00
Aleix Conchillo Flaqué
ea7324b2ba Add changelog for #4183 2026-03-27 19:03:55 -07:00
Aleix Conchillo Flaqué
52ed7137af Yield after create_task to ensure timer tasks are scheduled
Add `await asyncio.sleep(0)` after `create_task()` calls in
UserIdleController, SpeechTimeoutUserTurnStopStrategy,
TurnAnalyzerUserTurnStopStrategy, and UserTurnCompletionLLMServiceMixin
so the event loop schedules the newly created timer tasks before the
caller continues.
2026-03-27 19:03:23 -07:00
kompfner
b33df03724 Merge pull request #4179 from pipecat-ai/pk/fix-gemini-live-vertex
Don't send history_config for Gemini Live Vertex (unsupported)
2026-03-27 17:34:29 -04:00
Paul Kompfner
28fbe1db08 Don't send history_config for Gemini Live Vertex (unsupported) 2026-03-27 17:30:47 -04:00
kompfner
9240e92d9f Merge pull request #4177 from pipecat-ai/pk/tweak-26i-for-gemini-3.1-flash-live-support
Tweak 26i example system instruction for Gemini 3.1 Flash Live compat…
2026-03-27 17:20:06 -04:00
Paul Kompfner
5caf53f086 Tweak 26i example system instruction for Gemini 3.1 Flash Live compatibility
Gemini 3.1 Flash Live won't reliably report ending its turn until
after it says something following a tool call. Restructure the system
instruction so the model says goodbye *after* calling
end_conversation, and add a comment explaining the deferred EndFrame
behavior that makes this work.
2026-03-27 17:13:17 -04:00
Mark Backman
ac2716811c Merge pull request #4176 from pipecat-ai/mb/fix-websocket-rtvi-messages
Fix RTVI events not delivered over WebSocket transports
2026-03-27 16:50:37 -04:00
Mark Backman
d313d56776 Fix RTVI events not delivered over WebSocket transports
The base serializer filters out RTVI protocol messages by default
(ignore_rtvi_messages=True) to prevent them from being sent over
telephony media streams. ProtobufFrameSerializer is used by WebSocket
transports, which are the delivery channel for these messages, so
disable the filter there.
2026-03-27 16:47:11 -04:00
kompfner
159776f106 Merge pull request #4175 from pipecat-ai/pk/gemini-live-dropped-support-for-text-modality
Warn when TEXT modality is set for Gemini Live, and remove 26d text example
2026-03-27 16:26:36 -04:00
kompfner
a23803478f Merge pull request #4171 from pipecat-ai/pk/fix-gemini-3.1-flash-live-video
Gate Gemini Live sending real-time input messages to the API until it…
2026-03-27 16:26:03 -04:00
Mark Backman
bae193ab4d Merge pull request #4172 from pipecat-ai/mb/rime-tts-fixes
Fix Rime TTS stop-frame handling and handle done message
2026-03-27 16:22:25 -04:00
Paul Kompfner
04adb697be Warn when TEXT modality is set for Gemini Live, and remove 26d text example
All recent Gemini Live models (including the default
gemini-2.5-flash-native-audio-preview-12-2025, and going at least as
far back as gemini-2.5-flash-native-audio-preview-09-2025) only
support AUDIO as a response modality. We considered using
`modalities=TEXT` as a Pipecat-level signal to suppress audio output
frames (so developers could pair Gemini Live with an external TTS),
but the output transcription from the API arrives too late relative
to the audio to be useful for driving an external TTS service.

For now, just log a warning when a TEXT modality is configured
(at init or via set_model_modalities) and proceed as normal. The 26d
text-modality example is removed since it no longer represents a
viable configuration.
2026-03-27 16:21:15 -04:00
Mark Backman
4f9c8a6860 Merge pull request #4174 from pipecat-ai/fix/deepgram-sdk-6.1.0-compat
Fix Deepgram STT compatibility with deepgram-sdk 6.1.0
2026-03-27 15:11:43 -04:00
Mark Backman
a1a29b3933 Add changelog for #4174 2026-03-27 14:50:12 -04:00
Mark Backman
0798803c70 Bump deepgram-sdk minimum version to 6.1.0 2026-03-27 14:46:17 -04:00
Mark Backman
6422661d08 Fix Deepgram STT compatibility with deepgram-sdk 6.1.0
The SDK now requires explicit message objects for send_keep_alive,
send_close_stream, and send_finalize instead of no-arg calls.
2026-03-27 14:40:48 -04:00
Mark Backman
ed94b65d83 Merge pull request #4173 from pipecat-ai/filipi/updating_inworld_examples
Removing the models from the Inworld example so we can use the default model.
2026-03-27 14:02:55 -04:00
filipi87
f9670b9601 Removing the models from the Inworld example so we can use the default model. 2026-03-27 14:23:20 -03:00
OmercohenAviv
de8ba68589 Fix heartbeat monitor timeout not respecting custom heartbeat interval
The heartbeat monitor timeout (`HEARTBEAT_MONITOR_SECS`) was a static
module-level constant that never derived from the user-configurable
`heartbeats_period_secs`. This meant overriding the heartbeat interval
had no effect on the monitor window, causing spurious warnings or
delayed detection depending on the configured interval.

Add a new `heartbeats_monitor_secs` parameter to `PipelineParams` so
the monitor timeout is independently configurable (defaults to 10s).
The monitor handler now reads from the instance param instead of the
hard-coded constant.

Made-with: Cursor
2026-03-27 19:41:06 +03:00
Paul Kompfner
5b2991f47f Gate Gemini Live sending real-time input messages to the API until it's ready, i.e. after we've sent the initial conversation history (or determined that we don't need to).
This fixes the 26c example when using Gemini 3.1 Flash Live, which seems to be more strict about not receiving real-time input (at least, video messages) before conversation history.
2026-03-27 12:41:05 -04:00
Mark Backman
fc3186dc0d Add changelog entries for PR #4172 2026-03-27 12:38:53 -04:00
Mark Backman
1808b447c9 Handle done message from Rime TTS to avoid stop-frame timeout
Rime's WebSocket API sends a done message when synthesis completes.
Handle it to stop TTFB metrics, push TTSStoppedFrame, and remove the
audio context immediately instead of relying on the 3-second
stop_frame_timeout_s fallback.
2026-03-27 12:37:03 -04:00
Mark Backman
70df9d3fe4 Fix duplicate TTSStoppedFrame in TTS service timeout path 2026-03-27 12:07:37 -04:00
Filipi da Silva Fuchter
a8bfc23d3a Merge pull request #4167 from pipecat-ai/filipi/inworld_improvements
InworldTTSService improvements.
2026-03-27 11:15:14 -04:00
filipi87
e2870fc2ac Changing to debug the log when we are not able to append audio to the context. 2026-03-27 12:12:16 -03:00
filipi87
e851f8c1d5 Adding changelog entry for the fix. 2026-03-27 12:11:35 -03:00
filipi87
b31bece617 Not trying to recreate the context. 2026-03-27 12:06:21 -03:00
kompfner
9e350bcc2f Merge pull request #4147 from pipecat-ai/cb/gemini-transcript-fixes
Fix Gemini Live to handle bundled server_content fields
2026-03-27 11:00:10 -04:00
Paul Kompfner
9c2594c484 Remove brittle test 2026-03-27 10:56:39 -04:00
Mark Backman
900fc88430 Merge pull request #4128 from pipecat-ai/mb/end-of-turn-assembly 2026-03-27 10:47:09 -04:00
filipi87
4ef5ac6f0c InworldTTSService improvements. 2026-03-27 11:33:32 -03:00
Mark Backman
cbb3d99493 Merge pull request #4166 from pipecat-ai/mb/fix-example-ordering-56
Fix example numbering, add LemonSlice to evals
2026-03-27 10:29:07 -04:00
Filipi da Silva Fuchter
fb1996cedc Merge pull request #4143 from pipecat-ai/cb/sagemaker-flux
Add Deepgram Flux STT service for AWS SageMaker
2026-03-27 10:27:49 -04:00
Filipi da Silva Fuchter
95c55ec6c3 Merge pull request #4145 from pipecat-ai/filipi/tts_improvements_remove_reset
TTS improvements.
2026-03-27 10:24:59 -04:00
Mark Backman
a45de9af7f Merge pull request #4161 from tanmayc25/fix/lemonslice-missing-dtmf-callback
fix(lemonslice): add missing on_dtmf_event callback in DailyCallbacks construction
2026-03-27 10:19:54 -04:00
Mark Backman
5e61a57582 Fix changelog entry for #4161 2026-03-27 10:16:25 -04:00
Mark Backman
d8b0ed18fd Fix example numbering, add LemonSlice to evals 2026-03-27 10:11:37 -04:00
Mark Backman
789275a57b Merge pull request #4164 from pipecat-ai/mb/update-community-integrations-guide
docs: update COMMUNITY_INTEGRATIONS.md for accuracy
2026-03-27 09:38:31 -04:00
Filipi da Silva Fuchter
38c961a363 Merge pull request #4113 from inworld-ai/ian/lang-timestamps
fix(inworld): fallback to full text when TTS timestamps are not received
2026-03-27 09:34:05 -04:00
Mark Backman
41a86a51bf docs: update COMMUNITY_INTEGRATIONS.md for accuracy
- Replace deprecated TTS classes (AudioContextWordTTSService, WordTTSService)
  with current hierarchy (WebsocketTTSService, InterruptibleTTSService, TTSService)
- Add WebsocketSTTService and SDK-based STTService categories
- Fix LLM section: document _process_context, adapter_class, remove deprecated
  create_context_aggregator guidance, add thought frames for reasoning models
- Fix Vision section: run_vision takes UserImageRawFrame not LLMContext,
  yields Vision*Frame types not TextFrame
- Fix push_error API: takes (error_msg, exception) not ErrorFrame
- Fix frame name: TTSRawAudioFrame → TTSAudioRawFrame
- Remove stale v13+ version reference
- Clarify @traced_stt method convention
2026-03-27 09:22:32 -04:00
Filipi da Silva Fuchter
e1bfa4cf21 Merge pull request #4152 from vpalmisano/vpalmisano-patch-1
Fix audio transcript check in base_llm.py
2026-03-27 08:34:15 -04:00
filipi87
537d57449e Fixing the format and including the changelog. 2026-03-27 09:29:46 -03:00
Tanmay Chaudhari
33e146decd fix(lemonslice): add missing on_dtmf_event callback in DailyCallbacks construction
DailyCallbacks gained a required on_dtmf_event field in PR #4047.
PR #4079 fixed this for TavusTransportClient but
LemonSliceTransportClient.setup() was not updated, causing a pydantic
ValidationError at pipeline setup time.
2026-03-27 12:06:26 +05:30
Mark Backman
eee47deb34 Merge pull request #4060 from alpsencer/fix/empty-tool-call-arguments
fix(openai): handle tool calls with empty/null arguments
2026-03-26 22:04:37 -04:00
Mark Backman
21a729ae5d Merge pull request #4146 from pipecat-ai/mb/gemini-live-local-vad 2026-03-26 17:48:21 -04:00
Filipi da Silva Fuchter
1870f4010e Merge pull request #4158 from pipecat-ai/filipi/flux_refactor
Creating a base class, DeepgramFluxSTTBase, to reuse Deepgram Flux logic
2026-03-26 17:33:35 -04:00
filipi87
28683a7296 Moving flux_stt.py to deepgram/flux/sagemaker/stt.py 2026-03-26 17:43:51 -03:00
filipi87
0e504d876d Creating a base class DeepgramFluxSTTBase so we can reuse Deepgram Flux logic. 2026-03-26 17:37:37 -03:00
Mark Backman
5c51981207 Merge pull request #4149 from pipecat-ai/mb/fix-service-switcher-passthrough-errors
Fix ServiceSwitcher reacting to pass-through ErrorFrames
2026-03-26 16:34:45 -04:00
Mark Backman
a13c4d1248 Narrow ServiceSwitcher error check to active service only
Only trigger handle_error for ErrorFrames originating from the active
service, not any managed service. This prevents edge cases where errors
from a non-active service could incorrectly trigger failover.
2026-03-26 15:28:19 -04:00
filipi87
ca1b4ad124 Organizing the methods from Deepgram Flux and Flux SageMaker in the same position. 2026-03-26 16:05:17 -03:00
Mark Backman
533dcdba3f Merge pull request #4154 from pipecat-ai/mb/deprecate-sambanova-stt
Remove SambaNovaSTTService
2026-03-26 14:10:14 -04:00
Mark Backman
7eec03cb77 Merge pull request #4156 from pipecat-ai/mb/mem0-improvements
fix(mem0): improve Mem0 service reliability and add get_memories() method
2026-03-26 14:09:34 -04:00
Mark Backman
83911dced6 docs: add changelog entries for #4156 2026-03-26 13:30:00 -04:00
Mark Backman
4e4a8c45d5 build(mem0): bump mem0ai dependency to >=1.0.8,<2 2026-03-26 13:28:41 -04:00
Mark Backman
9c6d51c570 feat(mem0): add get_memories() convenience method to Mem0MemoryService
Expose a public method for retrieving all stored memories outside the
pipeline, avoiding the need for callers to reimplement client branching,
OR filter construction, and asyncio.to_thread wrapping. Simplify the
example get_initial_greeting() to use it.
2026-03-26 13:28:41 -04:00
Mark Backman
9152d85824 fix(mem0): filter to user/assistant roles before storing in Mem0
Mem0 API only accepts user and assistant roles. Filter out system,
developer, and other roles before calling add() to avoid 400 errors.
2026-03-26 13:28:41 -04:00
Mark Backman
6a87d0e87d fix(mem0): make memory service non-blocking and use position parameter
Move blocking Mem0 API calls off the event loop using asyncio.to_thread().
Store messages as a fire-and-forget background task via create_task() since
the result is not needed. Insert memory messages at the configured position
in the context instead of always appending.

Closes #1741
2026-03-26 13:28:41 -04:00
Mark Backman
fe0633ecd1 Add 14s to release evals 2026-03-26 12:27:27 -04:00
Mark Backman
ca2bfd6f12 Remove SambaNovaSTTService
SambaNova no longer offers speech-to-text audio models.
2026-03-26 12:22:06 -04:00
kompfner
345ccc0abe Merge pull request #4148 from pipecat-ai/khk/gemini-transcription-fixes-addon
Fix bundled Gemini Live transcription ordering
2026-03-26 11:33:10 -04:00
namanbansal013
800fd6a916 Changelog entry for the websocket word context leak. 2026-03-26 11:52:34 -03:00
filipi87
d286991257 Changelog entry for the changes involving add_word_timestamp. 2026-03-26 11:51:31 -03:00
namanbansal013
a06bf47ed2 Discard any pre-audio word timestamps from the interrupted turn. 2026-03-26 11:42:24 -03:00
Mark Backman
5ad4aa9bea Merge pull request #4153 from pipecat-ai/mb/deepgram-stt-try-except
Handle Deepgram SDK 6.x send_media() exceptions
2026-03-26 10:15:21 -04:00
filipi87
c4466ba678 Adding changelog for the InterruptibleTTSService race condition fix 2026-03-26 10:58:57 -03:00
filipi87
df602b900d Preventing a race condition in the InterruptibleTTSServices in cases where run_tts has been invoked but the BotStartedSpeakingFrame has not yet been received. 2026-03-26 10:39:44 -03:00
Mark Backman
c331c75d66 Add tests for send_media() exception handling in DeepgramSTTService 2026-03-26 09:20:58 -04:00
filipi87
f7ec6befe1 Invoking superclass method when audio context is interrupted or completed. 2026-03-26 10:14:21 -03:00
Mark Backman
6a6ee8d563 Merge pull request #4150 from pipecat-ai/mb/resolve-dependabot-2026-03-25
Bump nltk minimum version to 3.9.4 to resolve CVE-2026-33230
2026-03-26 09:10:47 -04:00
Mark Backman
259f5e124c Add changelog for #4153 2026-03-26 08:48:45 -04:00
Mark Backman
cfe91d11ec Handle Deepgram SDK 6.x send_media() exceptions
Deepgram SDK 6.x surfaces connection errors from send_media() instead
of silently swallowing them. This causes error floods when the WebSocket
disconnects since every queued audio frame hits the dead connection.

Wrap send_media() in try/except: on failure, log one warning and set
self._connection = None so subsequent frames skip until the existing
_connection_handler reconnects.
2026-03-26 08:45:42 -04:00
Vittorio Palmisano
467184e63e Fix audio transcript check in base_llm.py
In some cases the openai provider could answer with a `chunk.choices[0].delta.audio = None`, so the process context fails with error:
```
pipecat/services/openai/base_llm.py:552): Error during completion: 'NoneType' object has no attribute 'get'
```
2026-03-26 13:09:36 +01:00
Mark Backman
af566ac936 Merge pull request #4151 from ajmeraharsh/fix/livekit-clear-audio-queue-on-interruption
fix(livekit): clear AudioSource buffer on interruption
2026-03-26 00:52:26 -04:00
ajmeraharsh
62484a4fc3 fix(livekit): clear AudioSource buffer on interruption
When an InterruptionFrame arrives, the Python-side audio task is
cancelled but frames already submitted to rtc.AudioSource continue
playing from its internal buffer. This causes the bot to keep speaking
for several seconds after being interrupted.

Fix by overriding process_frame in LiveKitOutputTransport to call
audio_source.clear_queue() on InterruptionFrame, immediately flushing
the buffered audio.
2026-03-26 09:47:00 +05:30
Mark Backman
7fef3b01eb Merge pull request #4142 from pipecat-ai/mb/grok-move-to-xai-module
Consolidate Grok services into xai module
2026-03-25 23:32:18 -04:00
Mark Backman
6d1918f12a Update GROK_API_KEY to XAI_API_KEY 2026-03-25 23:23:58 -04:00
Mark Backman
e58740e948 Bump nltk minimum version to 3.9.4 to resolve CVE-2026-33230 2026-03-25 23:16:46 -04:00
Mark Backman
ddfe44940d Add changelog for #4149 2026-03-25 22:54:25 -04:00
Mark Backman
fdbdbc8be3 Fix ServiceSwitcher reacting to pass-through ErrorFrames from other pipeline stages
ErrorFrames propagating upstream from downstream processors (e.g. TTS) would
enter the ServiceSwitcher via process_frame, traverse the active service sub-pipeline,
and reach push_frame where they incorrectly triggered failover. Now only errors whose
processor is one of the managed services trigger handle_error. Also fix the log in
handle_error to attribute errors to the actual source processor rather than the
current active_service.

Closes #4139
2026-03-25 22:53:04 -04:00
Kwindla Hultman Kramer
3cd7d882fb Fix bundled Gemini Live transcription ordering 2026-03-25 18:56:00 -07:00
Chad Bailey
2d78533d77 Add changelog for Gemini Live server_content fix 2026-03-25 23:42:42 +00:00
Chad Bailey
c1dd44f947 Fix Gemini Live message handling to process all server_content fields
Gemini 3.x can bundle multiple fields (e.g. model_turn and
output_transcription) on the same server_content message. The previous
elif chain would only process the first matching field and silently
drop the rest. Switch to independent if checks so every field is
handled.
2026-03-25 23:42:07 +00:00
Mark Backman
9db15e7942 Add changelog entry for #4146 2026-03-25 18:07:05 -04:00
Mark Backman
503e5e9106 Fix Gemini Live local VAD by sending correct activity events to server
When Gemini Live was configured with local VAD (server-side VAD disabled),
the service was listening for the wrong frame types and not sending
ActivityStart/ActivityEnd events to the server. Now it listens for
VADUserStartedSpeakingFrame/VADUserStoppedSpeakingFrame and sends the
appropriate activity signals when local VAD is in use.

Also removes the unnecessary local SileroVADAnalyzer from server-side VAD
examples and adds a new 26a example demonstrating local VAD configuration.
2026-03-25 18:00:13 -04:00
filipi87
2ff4b3f4a3 Improving docstring based on the recent changes. 2026-03-25 17:52:05 -03:00
filipi87
b4096f9a11 Refactoring to remove "Reset" and "TTSStoppedFrame" from word. 2026-03-25 17:47:24 -03:00
filipi87
c4253a7d98 Refactoring to invoke append_to_audio_context instead of direct queue put. 2026-03-25 17:21:55 -03:00
Filipi da Silva Fuchter
2441c4f801 Merge pull request #4135 from pipecat-ai/filipi/audio_buffer
Fixed audio crackling and popping artifacts in AudioBufferProcessor
2026-03-25 15:40:17 -04:00
Mark Backman
a7a55dd30e Merge pull request #4136 from pipecat-ai/mb/bump-package-version-nvidia
Upgrade protobuf to 6.x for nvidia-riva-client compatibility
2026-03-25 15:27:48 -04:00
Mark Backman
de6a7223ba Suppress verbose gRPC C-core logging in nvidia services
Set GRPC_VERBOSITY=ERROR by default so users do not see noisy fork
handler and abseil warnings from the gRPC C library. Users can still
override by setting GRPC_VERBOSITY themselves.
2026-03-25 15:23:54 -04:00
Mark Backman
165932e1cc Add changelog for #4136 2026-03-25 15:23:54 -04:00
Mark Backman
1f0d9ad01a Upgrade protobuf to 6.x for nvidia-riva-client 2.25.1 compatibility
nvidia-riva-client 2.25.1 ships with gencode compiled against protobuf
6.31.1, which requires a runtime >= 6.31.1. Update protobuf from 5.29.6
to >=6.31.1,<7 and grpcio-tools from 1.67.1 to 1.78.0 to match.
Regenerate frames_pb2.py with the new compiler.
2026-03-25 15:23:53 -04:00
Chad Bailey
052075c244 updated changelog 2026-03-25 19:12:37 +00:00
Chad Bailey
a8d0e1de9f Update changelog filename with PR number 2026-03-25 19:10:20 +00:00
Chad Bailey
4f0b2066c0 Add Deepgram Flux STT service for AWS SageMaker
Add DeepgramFluxSageMakerSTTService that combines SageMaker's HTTP/2
transport with Flux's JSON turn detection protocol (StartOfTurn,
EndOfTurn, EagerEndOfTurn, TurnResumed). Includes mid-stream Configure
support, silence watchdog, and an example bot.
2026-03-25 19:09:52 +00:00
filipi87
413dbaf974 Automated tests to validate the silence injection guards. 2026-03-25 16:05:58 -03:00
Ian Lee
5645909d34 [inworld] add falbback for empty timestamps from server 2026-03-25 11:55:09 -07:00
filipi87
da3f184316 Automated tests to validate the silence injection guards. 2026-03-25 15:38:21 -03:00
filipi87
e5a2723632 Fixed audio crackling and popping artifacts in AudioBufferProcessor. 2026-03-25 15:29:50 -03:00
Mark Backman
4ee4002d5d Merge pull request #4137 from pipecat-ai/mb/language-string-log-level-debug
Downgrade unrecognized language string log from warning to debug
2026-03-25 12:26:46 -04:00
Mark Backman
54a17ab1f3 Add changelog for #4142 2026-03-25 12:22:37 -04:00
Mark Backman
1c99a537b2 Consolidate Grok services into xai module
Both GrokLLMService and XAIHttpTTSService use the same xAI API (api.x.ai),
so move Grok source files into the xai module. Leave deprecation shims in
the old grok/ paths for backward compatibility.
2026-03-25 12:07:40 -04:00
Mark Backman
ff5d055b3c Merge pull request #4031 from niczy/xai-tts-service
Add xAI TTS service
2026-03-25 10:57:08 -04:00
Mark Backman
adc003d6c7 Code review cleanup 2026-03-25 10:53:07 -04:00
Nicholas Zhao
bbd14de9c5 Address PR review: rename to XAIHttpTTSService, add language map, clean up API
- Rename XAITTSService → XAIHttpTTSService and XAITTSSettings → XAIHttpTTSSettings
- Add language_to_xai_language() with explicit LANGUAGE_MAP using resolve_language()
- Remove deprecated InputParams, params, voice, language init params
- Remove XAI_DEFAULT_SAMPLE_RATE and XAI_PCM_CODEC constants; add encoding param
- Set sample_rate=None default (picked up from PipelineParams or user)
- Use Language.EN enum instead of string "en" for default language
- Add changelog/4031.added.md
- Add 07e-interruptible-xai.py foundational example
- Update 14g-function-calling-grok.py to use XAIHttpTTSService
- Register 07e in run-release-evals.py
2026-03-25 10:46:54 -04:00
Nicholas Zhao
02b97035f8 Add xAI TTS service 2026-03-25 10:45:15 -04:00
Mark Backman
f470ff193e Update language tests to expect debug instead of warning 2026-03-25 10:26:10 -04:00
Mark Backman
7bc8b89a54 Add changelog for #4137 2026-03-25 10:21:44 -04:00
Mark Backman
a8eff6fbbf Downgrade unrecognized language string log from warning to debug
Service-specific language strings like Deepgram's "multi" are valid
pass-through values, not issues worth warning about.
2026-03-25 10:20:36 -04:00
kompfner
86e086c6b5 Merge pull request #4130 from pipecat-ai/pk/realtime-services-init-v-context-system-instructions-cleanup
Prefer init-provided system instructions in realtime services
2026-03-25 09:13:52 -04:00
Paul Kompfner
4bdfe1cf31 Add changelog for realtime system instruction preference change 2026-03-24 17:34:50 -04:00
Paul Kompfner
bb33045389 Add system instruction conflict resolution tests for realtime adapters
Test that OpenAI Realtime, Grok Realtime, and Nova Sonic adapters
prefer init-provided system_instruction over context-provided, warn
on conflicts, and don't warn for developer messages.
2026-03-24 17:30:35 -04:00
Paul Kompfner
ac2b1ecd47 Prefer init-provided system instruction in Grok Realtime
Add system_instruction parameter to the Grok Realtime adapter's
get_llm_invocation_params() and call _resolve_system_instruction() to
prefer init-provided over context-provided system instructions and
warn on conflicts. Previously context-provided took precedence.

Update the Grok Realtime example to use settings.system_instruction
instead of session_properties.instructions.
2026-03-24 17:29:19 -04:00
Paul Kompfner
e7dd84b552 Prefer init-provided system instruction in OpenAI Realtime
Add system_instruction parameter to the OpenAI Realtime adapter's
get_llm_invocation_params() and call _resolve_system_instruction() to
prefer init-provided over context-provided system instructions and
warn on conflicts. Previously context-provided took precedence.
2026-03-24 17:21:53 -04:00
Paul Kompfner
39329aaddb Prefer init-provided system instruction in Nova Sonic
Add system_instruction parameter to the Nova Sonic adapter's
get_llm_invocation_params() and call _resolve_system_instruction() to
prefer init-provided over context-provided system instructions and
warn on conflicts. Previously context-provided took precedence.

Remove the service-side fallback logic, as the adapter now handles
resolution.
2026-03-24 17:18:44 -04:00
Paul Kompfner
56a56a4174 Prefer init-provided system instruction in Gemini Live
Pass self._system_instruction_from_init to the adapter's
get_llm_invocation_params(), which calls _resolve_system_instruction()
to prefer init-provided over context-provided system instructions and
warn on conflicts. Previously context-provided took precedence.

Also fix the reconnect check to only reconnect when the resolved
system instruction actually differs from what the initial connection
used, avoiding unnecessary reconnects.
2026-03-24 17:06:56 -04:00
kompfner
b80328e038 Merge pull request #4125 from pipecat-ai/pk/gemini-live-endframe-deferral-issue
Gemini Live: fix EndFrame-deferral hang
2026-03-24 17:02:46 -04:00
kompfner
3a80be760b Merge pull request #4089 from pipecat-ai/pk/system-and-developer-message-handling-update
Centralize system message handling in adapters; add developer message support
2026-03-24 16:24:11 -04:00
Mark Backman
b66c892100 Add changelog for #4128 2026-03-24 16:15:00 -04:00
Mark Backman
6c30371295 Fix Deepgram Flux event handler docstring to match implementation
Update documented event signatures to include transcript argument
where the code actually passes it. Remove stale on_speech_started
and on_utterance_end entries that were never registered.
2026-03-24 16:12:25 -04:00
Mark Backman
ddf6a41854 Add on_end_of_turn event handler to AssemblyAI STT
Fires after the final transcript is pushed in both Pipecat and
AssemblyAI turn detection modes, giving users a reliable hook
that arrives after all transcript frames. Matches the existing
Deepgram Flux on_end_of_turn pattern.
2026-03-24 16:11:35 -04:00
Paul Kompfner
e0c49927cf Remove hard-coded model overrides from Together and Groq examples
Prefer service defaults — the hard-coded models we were using are no
longer available on these providers.
2026-03-24 16:05:15 -04:00
Paul Kompfner
45926a7135 Update Together.ai default model to openai/gpt-oss-20b
The previous default (meta-llama/Meta-Llama-3.1-8B-Instruct-Turbo) is
no longer available as a serverless Together.ai model and now requires a
custom deployment. The new default is openai/gpt-oss-20b, one of
Together's recommended models for small & fast use-cases.
2026-03-24 16:05:15 -04:00
Paul Kompfner
8c678c1c98 Set supports_developer_role = False for more OpenAI-compatible services
DeepSeek, Mistral, OLLama, Qwen, SambaNova, and Together don't support
the "developer" message role.
2026-03-24 16:05:15 -04:00
Paul Kompfner
4c121332cf Convert developer messages to user for Cerebras (and lay groundwork for other incompatible services)
OpenAI-compatible services that don't support the "developer" message
role can now set supports_developer_role = False on the service class.
BaseOpenAILLMService passes this as convert_developer_to_user to the
adapter, which converts developer messages to user messages before
sending them to the API. Applied to Cerebras and Perplexity.

Also removes the now-redundant developer→user conversion step from
PerplexityLLMAdapter (handled by the parent adapter via the flag).
2026-03-24 16:05:15 -04:00
Paul Kompfner
74686f9190 Add changelog for Gemini Live system_instruction fix 2026-03-24 16:05:15 -04:00
Paul Kompfner
19bcc8620c Fix Gemini Live not honoring settings.system_instruction
_system_instruction_from_init was being set from the deprecated
`system_instruction` constructor parameter instead of
`self._settings.system_instruction`, so system instructions provided
via settings were silently ignored.
2026-03-24 16:05:15 -04:00
Paul Kompfner
0530722c58 Convert developer messages to user in Perplexity adapter
Perplexity doesn't support the "developer" role. Developer messages are
now converted to "user" before other transformations are applied.
2026-03-24 16:05:15 -04:00
Paul Kompfner
0d1b834770 Add developer message support to realtime adapters
OpenAI Realtime, Grok Realtime, and AWS Nova Sonic adapters now convert
"developer" role messages to "user" (consistent with all other non-OpenAI
adapters). Previously these messages were silently dropped. Adds starter
unit tests for all three realtime adapters.
2026-03-24 16:05:15 -04:00
Paul Kompfner
7a0f7b58d1 Remove bit of unintentionally-left-in debugging logic 2026-03-24 16:05:15 -04:00
Paul Kompfner
5806a3f0fa Use "developer" role for remaining developer-intent messages in examples 2026-03-24 16:05:04 -04:00
Paul Kompfner
27fabfc1b3 Improve warning message wording and formatting 2026-03-24 16:02:42 -04:00
Paul Kompfner
d779a5b4ea Use "developer" role for programmatic conversation-kickoff messages
These messages are developer instructions to the assistant (e.g. "Please
introduce yourself to the user"), not simulated user input. The
"developer" role is semantically correct for this purpose.
2026-03-24 16:02:42 -04:00
Paul Kompfner
2bb36b5b66 Update changelog for developer message simplification 2026-03-24 16:02:42 -04:00
Paul Kompfner
e0bc9c73c6 Add Anthropic interruptible example (07e) and register in release evals 2026-03-24 16:02:42 -04:00
Paul Kompfner
2135557689 Simplify: don't promote developer messages to system instruction
Developer messages are now always converted to "user" in non-OpenAI
adapters, never promoted to the system instruction. This removes an
inconsistency where adding an unrelated message to context would change
whether a developer message got promoted.

Simplifications:
- Rename _extract_initial_system_or_developer → _extract_initial_system
- Return Optional[str] instead of Tuple (role is always "system")
- Drop initial_context_message_role from _resolve_system_instruction
- Drop system_role fields from all ConvertedMessages dataclasses
2026-03-24 16:02:42 -04:00
Paul Kompfner
a0393b9af6 Fix: warn on system_instruction conflict even with single system message
When the only message in context was a system message,
_extract_initial_system_or_developer would convert it to "user" (to
prevent empty history) without warning about the conflict with
system_instruction. Now warns inline before converting, with a message
explaining both the conflict and the user-role conversion.
2026-03-24 16:02:42 -04:00
Paul Kompfner
64ba013b68 Move OpenAI Responses adapter tests into test_get_llm_invocation_params.py
Consolidates all adapter get_llm_invocation_params tests in one file.
Adds new tests for developer message handling in the Responses adapter.
2026-03-24 16:02:42 -04:00
Paul Kompfner
7377d88cf5 Move system_instruction tests into test_get_llm_invocation_params.py 2026-03-24 16:02:42 -04:00
Paul Kompfner
3bbec0a2c8 Broaden docstring: all non-OpenAI providers need non-empty messages 2026-03-24 16:02:42 -04:00
Paul Kompfner
e29a63e1ae Improve _extract_initial_system_or_developer docstring clarity 2026-03-24 16:02:42 -04:00
Paul Kompfner
45178972d7 Fix stale docstring in PerplexityLLMAdapter 2026-03-24 16:02:42 -04:00
Paul Kompfner
bb7199d143 Add changelog entries for #4089 2026-03-24 16:02:42 -04:00
Paul Kompfner
d4dea30407 Centralize system message handling in adapters; add developer message support
Two goals:

1. Centralize system_instruction vs context system message resolution into
   the LLM adapters. This eliminates duplication between in-pipeline and
   out-of-band (run_inference) code paths across ~16 locations in service
   llm.py files.

2. Add support for "developer" role messages in conversation context, which
   is facilitated by the above centralization.

Shared helpers on BaseLLMAdapter:
- _extract_initial_system_or_developer: extracts/converts messages[0]
  based on role and whether system_instruction is provided
- _resolve_system_instruction: warns on conflicts between system_instruction
  and context system messages, returns the effective instruction

Developer message handling (new):
- Non-OpenAI adapters: an initial "developer" message is promoted to the
  system instruction when no system_instruction is provided; otherwise it
  is converted to "user". Subsequent "developer" messages are always
  converted to "user". No conflict warning is emitted for developer
  messages (unlike "system" messages).
- OpenAI adapter: "developer" messages pass through in conversation
  history without triggering conflict warnings.
- OpenAI Responses adapter: "developer" messages are kept as "developer"
  role (same as "system", which is also converted to "developer" for the
  Responses API).

Other behavior changes:
- Gemini: "initial" system message detection now checks messages[0] only
  (previously searched anywhere in the list)
- Bedrock: a lone system message is now converted to "user" instead of
  being extracted to an empty message list (matches existing Anthropic
  behavior)
2026-03-24 16:02:42 -04:00
Mark Backman
b49bf1c83f Merge pull request #4127 from pipecat-ai/mb/tts-text-frame-ordering
Fix LLMFullResponseEndFrame racing ahead of final TTSTextFrame
2026-03-24 15:39:06 -04:00
Mark Backman
1b0f7ecb0e Merge pull request #4126 from pipecat-ai/mb/fix-tts-flush-phantom-contexts
Fix TTS flush creating phantom contexts on ElevenLabs
2026-03-24 15:33:58 -04:00
Mark Backman
8e57dd67a2 Add changelog for #4127 2026-03-24 15:10:48 -04:00
Mark Backman
5d71de8aad Fix LLMFullResponseEndFrame racing ahead of final TTSTextFrame
Route LLMFullResponseEndFrame through the serialization queue instead
of pushing it directly downstream when push_text_frames is enabled.
This ensures the frame is emitted only after the audio context is
fully drained, preserving correct ordering relative to TTSTextFrames.

Previously, the final sentence TTSTextFrame would arrive at the
LLMAssistantAggregator after LLMFullResponseEndFrame, causing it to
be dropped from the conversation context (especially with RTVI text
input where no subsequent interruption would flush the orphaned text).
2026-03-24 15:09:42 -04:00
Paul Kompfner
dc56cb2ccc Gemini Live: reset _bot_is_responding when releasing deferred EndFrame
Without this, the released EndFrame re-enters process_frame, sees
_bot_is_responding is still True, defers again, and loops indefinitely.
2026-03-24 15:01:07 -04:00
Paul Kompfner
063955b7eb Gemini Live: clean up EndFrame deferral state on disconnect
Cancel the deferral timeout task and clear the pending EndFrame during
disconnect, which could otherwise be left dangling after a
CancelFrame-triggered shutdown.
2026-03-24 14:30:14 -04:00
Mark Backman
e05bd54743 Add changelog for #4126 2026-03-24 13:43:07 -04:00
Mark Backman
35f52f70ab Fix TTS flush creating phantom contexts on providers like ElevenLabs
When an interruption arrives before any LLM text reaches run_tts, the
turn context ID exists but was never registered via create_audio_context.
Calling flush_audio for this unregistered context sends a message to the
provider (e.g. ElevenLabs) with a context_id it has never seen, which
implicitly creates a server-side context that is never closed. After
enough rapid interruptions these phantom contexts accumulate and exceed
the providers limit (ElevenLabs: 5 simultaneous contexts, 1008 policy
violation).

Guard the flush call with audio_context_available so it only fires when
the context was actually opened.

Fixes #4114
2026-03-24 13:42:01 -04:00
Paul Kompfner
d05eb02b98 Add changelog for #4125 2026-03-24 12:54:50 -04:00
Paul Kompfner
4abd4d031d Gemini Live: add safety timeout to EndFrame deferral to prevent indefinite pipeline hang
When an EndFrame arrives while the bot is mid-response, it is deferred
until turn_complete is received. If turn_complete never arrives, the
EndFrame gets stuck forever and the pipeline hangs indefinitely.

Add a 30-second timeout: if turn_complete hasn't arrived by then, the
deferred EndFrame is released anyway with a warning log. The timeout
is cancelled if turn_complete arrives normally.
2026-03-24 12:50:07 -04:00
Paul Kompfner
7e42998e9e Gemini Live: fix potential EndFrame-deferral hang by handling turn_complete without usage_metadata
We observed a case where a deferred EndFrame was never released in
Gemini Live, causing the pipeline to hang indefinitely. The EndFrame
deferral mechanism waits for _handle_msg_turn_complete to set
_bot_is_responding back to False, but turn_complete messages were only
processed if they also contained usage_metadata. If Gemini ever sent
turn_complete without usage_metadata, the message would be silently
dropped and the deferred EndFrame would never be released.

Now turn_complete is always handled regardless of usage_metadata
presence, with usage_metadata processing only when available.

Note: we have not actually observed a turn_complete without
usage_metadata in practice, so this is a theoretical fix for the
EndFrame-deferral hang. The actual root cause of the observed hang
may lie elsewhere.
2026-03-24 12:32:14 -04:00
Filipi da Silva Fuchter
28eb4544d3 Merge pull request #4122 from pipecat-ai/filipi/inworld_follow_up
Invoking on_turn_context_created when we receive a TTSSpeakFrame.
2026-03-24 12:28:00 -04:00
Filipi da Silva Fuchter
b45dcb1ae0 Merge pull request #4028 from inworld-ai/ian/close-on-turn-complete
fix(inworld): close context at end of turn instead of relying on idle timeout
2026-03-24 12:07:51 -04:00
Mark Backman
6eb988b729 Merge pull request #4092 from harshitajain165/harshita/smallest-tts-only
Add Smallest AI TTS service integration
2026-03-24 11:54:34 -04:00
Mark Backman
f68b3222b3 Fix SmallestTTSService to use InterruptibleTTSService audio context system
- Route audio through audio contexts (append_to_audio_context) instead of
  pushing frames directly, enabling proper turn management and interruptions
- Add push_stop_frames and push_start_frame so the base class handles
  TTSStartedFrame/TTSStoppedFrame lifecycle
- Remove manual context_id tracking (self._context_id) in favor of
  get_active_audio_context_id()
- Don't call remove_audio_context on "complete" — Smallest sends one
  per request, not per turn; let the base class timeout handle cleanup
- Guard v2-only params (consistency, similarity, enhancement) so they
  aren't sent to lightning-v3.1
- Remove request_id from request payload (not a documented request field)
- Add flush_audio override to send flush to WebSocket
2026-03-24 11:46:28 -04:00
filipi87
3274235ea1 Adding missing changelog entry. 2026-03-24 12:42:56 -03:00
filipi87
05b9c514fb Invoking TTSSpeakFrame when we receive a TTSSpeakFrame. 2026-03-24 12:39:28 -03:00
Filipi da Silva Fuchter
03c0d7c345 Merge pull request #4013 from inworld-ai/ian/prewarm-context-inworld-v2
[inworld] Pre-open WebSocket TTS context on LLM response start
2026-03-24 11:37:28 -04:00
Filipi da Silva Fuchter
0783edb185 Merge pull request #4120 from pipecat-ai/filipi/krisp-viva-vad-support
Added cleanup() method to VADAnalyzer base class
2026-03-24 11:26:53 -04:00
Mark Backman
51d28b4a9f Code review fixes 2026-03-24 11:21:04 -04:00
kompfner
cf083b8411 Merge pull request #4078 from pipecat-ai/cb/gemini-updates
Updates for Gemini Live
2026-03-24 11:18:00 -04:00
Harshita Jain
099814d74a Add Smallest AI TTS service integration
Adds SmallestTTSService, a WebSocket-based TTS service using Smallest AI's
Lightning v3.1 model. Follows current Pipecat service conventions:

- SmallestTTSSettings dataclass with runtime-updatable settings (voice,
  language, speed, etc.)
- Reconnects on model change; keepalive every 30s to prevent idle timeout
- TTS settings default to None so the API applies its own defaults
- Model enum: SmallestTTSModel.LIGHTNING_V3_1

Includes a foundational example (07zl-interruptible-smallest.py) using
Deepgram STT + Smallest TTS + OpenAI LLM.

STT integration will follow in a separate PR once the hallucination/finalize
behaviour is resolved.

Made-with: Cursor
2026-03-24 11:11:10 -04:00
Mark Backman
dd45843c42 Merge pull request #4117 from m-ods/feat/assemblyai-domain-param
feat(assemblyai): add domain parameter for Medical Mode
2026-03-24 11:02:01 -04:00
Mark Backman
fe15d8654b Add changelog for #4117 2026-03-24 10:57:55 -04:00
Paul Kompfner
68a440ae2e Move inference_on_context_initialization comment to constructor level 2026-03-24 10:49:45 -04:00
Paul Kompfner
8109ab6135 Further tweaks and improvements to Gemini 3 support in Gemini Live
Gets Gemini 3 support to the point where it works with:
- The "legacy" pattern from the previous (removed) 26- example
- inference_on_context_initialization=True (the default)
- inference_on_context_initialization=False
2026-03-24 10:45:41 -04:00
Filipi da Silva Fuchter
f311a0b6e4 Merge pull request #4084 from pipecat-ai/filipi/refactor_stop_frame
Refactoring the way we automatically push TTSStoppedFrame.
2026-03-24 10:06:02 -04:00
filipi87
9df8985d60 Refactoring the way we automatically push TTSStoppedFrame. 2026-03-24 11:00:06 -03:00
filipi87
b3a25e0ebe Adding changelog entry for cleanup method. 2026-03-24 10:53:07 -03:00
filipi87
02cfb129d3 Invoke cleanup method on VAD analyzer. 2026-03-24 10:49:14 -03:00
filipi87
311afef7da Fixing Krisp Viva example. 2026-03-24 10:48:22 -03:00
Filipi da Silva Fuchter
5ed183d215 Merge pull request #4022 from krispai/krisp-viva-vad-support
Draft Implementation for Krisp VIVA VAD.
2026-03-24 09:44:32 -04:00
Mark Backman
5c3d3aea2b Merge pull request #4115 from pipecat-ai/mb/user-turn-stop-warnings
Warn when VAD stop_secs misconfiguration may degrade turn detection
2026-03-24 09:32:20 -04:00
Mark Backman
0651569a4e Merge pull request #4119 from Alex-wuhu/novita-integration
feat: add Novita AI as LLM provider
2026-03-24 09:30:05 -04:00
Mark Backman
bf04ea2043 Add changelog for #4119 2026-03-24 09:23:40 -04:00
Mark Backman
aa0b49d69f Code review fixes 2026-03-24 09:22:08 -04:00
Alex-wuhu
8c6f4a8d7b Add Novita AI LLM service provider 2026-03-24 09:20:50 -04:00
Mark Backman
bbaa5971c4 Merge pull request #3981 from dhruvladia-sarvam/feat/sarvam-llm-integration
Sarvam LLM Integration
2026-03-24 09:00:01 -04:00
Mark Backman
cdd8c3e5bb Fix examples 2026-03-24 08:53:56 -04:00
Mark Backman
1c8a8f51d4 Code review fixes 2026-03-24 08:46:03 -04:00
dhruvladia-sarvam
349b8645f3 Merge branch 'main' into feat/sarvam-llm-integration 2026-03-24 16:34:12 +05:30
dhruvladia-sarvam
696196e30c alignment with pr 4081 2026-03-24 16:29:58 +05:30
Garegin Harutyunyan
dacffccd3a fixed runtime issue. 2026-03-24 12:56:19 +04:00
Martin Schweiger
f21b262969 feat(assemblyai): add domain parameter for Medical Mode
Add `domain` field to AssemblyAISTTSettings to support AssemblyAI's
streaming API `domain` query parameter, enabling specialized recognition
modes like Medical Mode (`medical-v1`).

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-24 13:09:42 +08:00
Mark Backman
42cab7eea0 Add changelog entry for #4115 2026-03-23 18:01:04 -04:00
Mark Backman
483b643b07 Warn when VAD stop_secs misconfiguration may degrade turn detection
Add warnings in SpeechTimeoutUserTurnStopStrategy and
TurnAnalyzerUserTurnStopStrategy when stop_secs differs from the
recommended default (0.2s) or when stop_secs >= STT p99 latency,
which collapses the STT wait timeout to 0s. Document the stop_secs=0.2
assumption in stt_latency.py.
2026-03-23 17:57:51 -04:00
Mark Backman
d314e2831a Simplify 26 name, update evals 2026-03-23 15:46:13 -04:00
dhruvladia-sarvam
0f6cc231cf removing error wraps and model validation check 2026-03-24 00:06:15 +05:30
Chad Bailey
844555c520 removed old Gemini Live example 2026-03-23 18:31:36 +00:00
dhruvladia-sarvam
3428a4c6ad fixes post PR 4081 2026-03-23 23:45:27 +05:30
Garegin Harutyunyan
f8c7414ea7 format fix. 2026-03-23 18:58:19 +04:00
Garegin Harutyunyan
f1f51de962 Merge branch 'main' into krisp-viva-vad-support 2026-03-23 18:35:58 +04:00
Garegin Harutyunyan
c32240e14b Fixed review comments. 2026-03-23 17:44:48 +04:00
Ian Lee
e9f3086ea3 add on_turn_context_created hook instead 2026-03-20 10:33:25 -07:00
Chad Bailey
05157129e2 added changelog 2026-03-18 23:31:18 +00:00
Chad Bailey
4a0411cbc4 disabled single responses for gemini 3 live models 2026-03-18 23:23:45 +00:00
Chad Bailey
6cd39b8b42 updates 2026-03-18 23:04:22 +00:00
Chad Bailey
38d7882f0f updated context seeding to allow gemini 3.1 to greet the user 2026-03-18 21:28:17 +00:00
Yavuz Alp Sencer ÖZTÜRK
9a55eb67cf fix(openai): handle tool calls with empty/null arguments
When an LLM returns a tool call with no arguments (arguments=null in
the streaming chunks), the tool call is silently dropped because:

1. `tool_call.function.arguments` is None, so nothing is accumulated
   and `arguments` stays as "" (empty string)
2. `if function_name and arguments:` treats "" as falsy, skipping the
   entire tool call execution

OpenAI always sends arguments="{}" even for parameterless tools,
masking this bug. But vLLM, Ollama, and other OpenAI-compatible
providers may omit arguments entirely when the tool schema has no
required parameters, causing tool calls to be silently ignored.

Fix: check only `function_name` (not `arguments`) and default empty
arguments to "{}" so `json.loads` produces an empty dict. Apply the
same fallback for intermediate tool calls in multi-tool responses.
2026-03-17 19:44:59 +03:00
dhruvladia-sarvam
8a4f6b486e wrapper fixes 2026-03-17 02:47:47 +05:30
dhruvladia-sarvam
8745f20330 fix llm wrapper redundancy and restore run_inference parity 2026-03-15 22:24:06 +05:30
Ian Lee
3e5be23bd8 fix(inworld): close context at end of turn instead of relying on idle timeout
The Inworld WS TTS plugin previously relied on the base TTS service's 3-second AUDIO_CONTEXT_TIMEOUT to detect when audio was done, then sent close_context in on_audio_context_completed. This added unnecessary latency before TTSStoppedFrame was emitted.

The original implementation likely borrowed this idea from the 11labs' impelementation. But it's likely better to mirror the Cartesia plugin pattern where on_audio_context_completed is a no-op because the server signals completion directly.

Now close_context is sent in on_turn_context_completed (right after flush_context), so the server responds with contextClosed immediately after the last audio byte. The existing receive handler already calls remove_audio_context on contextClosed, which exits the audio context handler cleanly.
2026-03-13 12:52:07 -07:00
Garegin Harutyunyan
33f042b500 format fix. 2026-03-13 19:32:39 +04:00
Garegin Harutyunyan
0722784f3a tests for VAD. 2026-03-13 19:30:03 +04:00
Garegin Harutyunyan
cbc1c275b3 num_frames_required() implementation. 2026-03-13 18:28:22 +04:00
Garegin Harutyunyan
14ca70f13e Fixed format issue. 2026-03-13 18:22:56 +04:00
Garegin Harutyunyan
f7568a91b1 Draft Implementation for Krisp VIVA VAD. 2026-03-13 18:12:21 +04:00
Ian Lee
dfe5fec8f9 [inworld] prewarm context on llm response start 2026-03-12 15:34:57 -07:00
dhruvladia-sarvam
dc0386937a Initial 2026-03-11 02:27:57 +05:30
591 changed files with 14370 additions and 19777 deletions

View File

@@ -144,7 +144,7 @@ class InputParams(BaseModel):
#### Examples
Validated against `examples/foundational/07-interruptible.py`:
Validated against `examples/07-interruptible.py`:
- Proper `create_transport()` usage
- Correct pipeline structure

View File

@@ -42,7 +42,7 @@ jobs:
- name: Test uv sync with all extras
run: |
uv sync --group dev --all-extras --no-extra krisp
uv sync --group dev --all-extras
- name: Verify installation
run: |

View File

@@ -1,51 +0,0 @@
name: Sync Quickstart to pipecat-quickstart repo
on:
push:
branches: [main]
paths:
- 'examples/quickstart/**'
workflow_dispatch: # Manual trigger
jobs:
sync-quickstart:
runs-on: ubuntu-latest
steps:
- name: Checkout main repo
uses: actions/checkout@v4
with:
fetch-depth: 0
- name: Checkout quickstart repo
uses: actions/checkout@v4
with:
repository: pipecat-ai/pipecat-quickstart
token: ${{ secrets.QUICKSTART_SYNC_TOKEN }}
path: quickstart-repo
- name: Sync files (excluding uv.lock and README.md)
run: |
# Copy all files except uv.lock and README.md
find examples/quickstart -type f \
-not -name "README.md" \
-not -name "uv.lock" \
-exec cp {} quickstart-repo/ \;
- name: Commit and push changes
run: |
cd quickstart-repo
git config user.name "GitHub Action"
git config user.email "action@github.com"
git add .
# Only commit if there are changes
if ! git diff --staged --quiet; then
git commit -m "Sync from pipecat main repo
Updated files from examples/quickstart/
Commit: ${{ github.sha }}
"
git push
else
echo "No changes to sync"
fi

View File

@@ -1,8 +1,13 @@
repos:
- repo: https://github.com/astral-sh/ruff-pre-commit
rev: v0.12.1
- repo: local
hooks:
- id: ruff
language_version: python3
args: [--fix]
name: ruff
entry: uv run ruff check --fix
language: system
types: [python]
- id: ruff-format
name: ruff-format
entry: uv run ruff format
language: system
types: [python]

View File

@@ -11,7 +11,7 @@ build:
jobs:
post_install:
- pip install uv
- UV_PROJECT_ENVIRONMENT=$READTHEDOCS_VIRTUALENV_PATH uv sync --group docs --all-extras --no-extra krisp --no-extra gstreamer --no-extra local_smart_turn --no-extra moondream --no-extra riva --no-extra mlx-whisper
- UV_PROJECT_ENVIRONMENT=$READTHEDOCS_VIRTUALENV_PATH uv sync --group docs --all-extras --no-extra gstreamer --no-extra local_smart_turn --no-extra moondream --no-extra riva --no-extra mlx-whisper
sphinx:
configuration: docs/api/conf.py

View File

@@ -7,6 +7,308 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
<!-- towncrier release notes start -->
## [0.0.108] - 2026-03-27
### Added
- Added `SarvamLLMService` with support for `sarvam-30b`, `sarvam-30b-16k`,
`sarvam-105b` and `sarvam-105b-32k`.
(PR [#3978](https://github.com/pipecat-ai/pipecat/pull/3978))
- Added `on_turn_context_created(context_id)` hook to `TTSService`. Override
this to perform provider-specific setup (e.g. eagerly opening a server-side
context) before text starts flowing. Called each time a new turn context ID
is created.
(PR [#4013](https://github.com/pipecat-ai/pipecat/pull/4013))
- Added `XAIHttpTTSService` for text-to-speech using xAI's HTTP TTS API.
(PR [#4031](https://github.com/pipecat-ai/pipecat/pull/4031))
- Added support for "developer" role messages in conversation context across
all LLM adapters. For non-OpenAI services (Anthropic, Google, AWS Bedrock),
"developer" messages are converted to "user" messages (use
`system_instruction` to set the system instruction). For OpenAI services,
"developer" messages pass through in conversation history. For the Responses
API, they are kept as "developer" role (matching the existing "system" →
"developer" conversion).
(PR [#4089](https://github.com/pipecat-ai/pipecat/pull/4089))
- Added `SmallestTTSService`, a WebSocket-based TTS service integration with
Smallest AI's Waves API. Supports the Lightning v2 and v3.1 models with
configurable voice, language, speed, consistency, similarity, and enhancement
settings.
(PR [#4092](https://github.com/pipecat-ai/pipecat/pull/4092))
- Added warnings in turn stop strategies when `VADParams.stop_secs` differs
from the recommended default (0.2s) or when `stop_secs >= STT p99 latency`,
which collapses the STT wait timeout to 0s and may cause delayed turn
detection. The warnings guide developers to re-run the
[stt-benchmark](https://github.com/pipecat-ai/stt-benchmark) with their VAD
settings.
(PR [#4115](https://github.com/pipecat-ai/pipecat/pull/4115))
- Added `domain` parameter to `AssemblyAISTTSettings` for specialized
recognition modes such as Medical Mode (`domain="medical-v1"`).
(PR [#4117](https://github.com/pipecat-ai/pipecat/pull/4117))
- Added `NovitaLLMService` for using Novita AI's LLM models via their
OpenAI-compatible API.
(PR [#4119](https://github.com/pipecat-ai/pipecat/pull/4119))
- Added `cleanup()` method to `VADAnalyzer` and `VADController` so VAD analyzer
resources are properly released when no longer needed. Custom `VADAnalyzer`
subclasses can override `cleanup()` to free any held resources.
(PR [#4120](https://github.com/pipecat-ai/pipecat/pull/4120))
- Added `on_end_of_turn` event handler to `AssemblyAISTTService`. This fires
after the final transcript is pushed, providing a reliable hook for
end-of-turn logic that doesn't race with `TranscriptionFrame`. Works in both
Pipecat and AssemblyAI turn detection modes.
(PR [#4128](https://github.com/pipecat-ai/pipecat/pull/4128))
- Added `DeepgramFluxSageMakerSTTService` for running Deepgram Flux
speech-to-text on AWS SageMaker endpoints. Use with
`ExternalUserTurnStrategies` to take advantage of Flux's turn detection.
(PR [#4143](https://github.com/pipecat-ai/pipecat/pull/4143))
- Added `Mem0MemoryService.get_memories()` convenience method for retrieving
all stored memories outside the pipeline (e.g. to build a personalized
greeting at connection time). This avoids the need to manually handle client
type branching, filter construction, and async wrapping.
(PR [#4156](https://github.com/pipecat-ai/pipecat/pull/4156))
### Changed
- Added context prewarming path for `InworldTTSService` to improve first audio
latency.
(PR [#4013](https://github.com/pipecat-ai/pipecat/pull/4013))
- Added `KrispVivaVadAnalyzer` for Voice Activity Detection using the Krisp
VIVA SDK (requires `krisp_audio`).
(PR [#4022](https://github.com/pipecat-ai/pipecat/pull/4022))
- Modified `InworldTTSService` to close context at end of turn instead of
relying on idle timeout.
(PR [#4028](https://github.com/pipecat-ai/pipecat/pull/4028))
- Added Gemini 3 support to the Gemini Live service.
(PR [#4078](https://github.com/pipecat-ai/pipecat/pull/4078))
- `TTSService`: the default `stop_frame_timeout_s` (idle time before an
automatic `TTSStoppedFrame` is pushed when `push_stop_frames=True`) has
changed from `2.0` to `3.0` seconds.
(PR [#4084](https://github.com/pipecat-ai/pipecat/pull/4084))
- ⚠️ `GeminiLLMAdapter` now only treats `messages[0]` as the initial system
message, matching all other adapters. Previously it searched for the first
"system" message anywhere in the conversation history. A "system" message
appearing later in the list will now be converted to "user" instead of being
extracted as the system instruction.
(PR [#4089](https://github.com/pipecat-ai/pipecat/pull/4089))
- Fixed `InworldTtsService` to fallback to full text when TTS timestamps are
not received.
(PR [#4113](https://github.com/pipecat-ai/pipecat/pull/4113))
- ⚠️ Realtime services (Gemini Live, OpenAI Realtime, Grok Realtime, Nova
Sonic) now prefer `system_instruction` from service settings over an initial
system message in the LLM context, matching the behavior of non-realtime
services. Previously, context-provided system instructions took precedence. A
warning is now logged when both are set.
(PR [#4130](https://github.com/pipecat-ai/pipecat/pull/4130))
- Bumped `nvidia-riva-client` minimum version to `>=2.25.1`.
(PR [#4136](https://github.com/pipecat-ai/pipecat/pull/4136))
- Upgraded `protobuf` from 5.x to 6.x (`>=6.31.1,<7`).
(PR [#4136](https://github.com/pipecat-ai/pipecat/pull/4136))
- Unrecognized language strings (e.g. Deepgram's `"multi"`) no longer produce a
warning at startup. The log message has been downgraded to debug level since
these are valid service-specific values that are passed through correctly.
(PR [#4137](https://github.com/pipecat-ai/pipecat/pull/4137))
- `GrokLLMService` and `GrokRealtimeLLMService` now live in the
`pipecat.services.xai` module alongside `XAIHttpTTSService`, since all three
use the same xAI API. Update imports from `pipecat.services.grok.*` to
`pipecat.services.xai.*` (e.g. `from pipecat.services.xai.llm import
GrokLLMService`).
(PR [#4142](https://github.com/pipecat-ai/pipecat/pull/4142))
- ⚠️ Bumped `mem0ai` dependency from `~=0.1.94` to `>=1.0.8,<2`. Users of the
`mem0` extra will need to update their mem0ai package.
(PR [#4156](https://github.com/pipecat-ai/pipecat/pull/4156))
### Deprecated
- `pipecat.services.grok.llm`, `pipecat.services.grok.realtime.llm`, and
`pipecat.services.grok.realtime.events` are deprecated. The old import paths
still work but emit a `DeprecationWarning`; use `pipecat.services.xai.llm`,
`pipecat.services.xai.realtime.llm`, and
`pipecat.services.xai.realtime.events` instead.
(PR [#4142](https://github.com/pipecat-ai/pipecat/pull/4142))
### Removed
- ⚠️ `TTSService.add_word_timestamps()` no longer supports the `"Reset"` and
`"TTSStoppedFrame"` sentinel strings. If you have a custom TTS service that
called `await self.add_word_timestamps([("Reset", 0)])` or `await
self.add_word_timestamps([("TTSStoppedFrame", 0), ("Reset", 0)], ctx_id)`,
replace them with `await self.append_to_audio_context(ctx_id,
TTSStoppedFrame(context_id=ctx_id))` and let `_handle_audio_context` manage
the word-timestamp reset automatically.
(PR [#4145](https://github.com/pipecat-ai/pipecat/pull/4145))
- Removed `SambaNovaSTTService`. SambaNova no longer offers speech-to-text
audio models. Use another STT provider instead.
(PR [#4154](https://github.com/pipecat-ai/pipecat/pull/4154))
### Fixed
- Fixed Gemini Live (`GoogleGeminiLiveLLMService`) not honoring
`settings.system_instruction`. The system instruction was being read from a
deprecated constructor parameter instead of the settings object, causing it
to be silently ignored.
(PR [#4089](https://github.com/pipecat-ai/pipecat/pull/4089))
- Fixed `AWSBedrockLLMAdapter` sending an empty message list to the API when
the only message in context was a system message. The lone system message is
now converted to "user" role instead of being extracted, matching the
existing Anthropic adapter behavior.
(PR [#4089](https://github.com/pipecat-ai/pipecat/pull/4089))
- Fixed Gemini Live pipeline hanging indefinitely when an `EndFrame` was
deferred while waiting for the bot to finish responding and `turn_complete`
never arrived. As a possible root-cause fix, `turn_complete` messages are now
handled even if they lack `usage_metadata`. As a fallback, the deferred
`EndFrame` now has a 30-second safety timeout.
(PR [#4125](https://github.com/pipecat-ai/pipecat/pull/4125))
- Fixed ElevenLabs WebSocket disconnections (1008 "Maximum simultaneous
contexts exceeded") caused by rapid user interruptions. When interruptions
arrived before any TTS text was generated, phantom contexts were created on
the ElevenLabs server that were never closed, eventually exceeding the
5-context limit.
(PR [#4126](https://github.com/pipecat-ai/pipecat/pull/4126))
- Fixed the final sentence being dropped from the conversation context when
using RTVI text input with non-word-timestamp TTS services. The
`LLMFullResponseEndFrame` was racing ahead of the last `TTSTextFrame`,
causing the `LLMAssistantAggregator` to finalize the context before the final
sentence arrived.
(PR [#4127](https://github.com/pipecat-ai/pipecat/pull/4127))
- Fixed audio crackling and popping in recordings when both user and bot are
speaking. `AudioBufferProcessor` no longer injects silence into a track's
buffer while that track is actively producing audio, preventing mid-utterance
interruptions in the recorded output.
(PR [#4135](https://github.com/pipecat-ai/pipecat/pull/4135))
- Fixed websocket TTS word timestamps so interrupted contexts cannot leak stale
words or backward PTS values into later turns.
(PR [#4145](https://github.com/pipecat-ai/pipecat/pull/4145))
- Fixed a race condition in `InterruptibleTTSService` where, if `run_tts` had
been invoked but `BotStartedSpeakingFrame` had not yet been received, a user
interruption could allow stale audio to leak through.
(PR [#4145](https://github.com/pipecat-ai/pipecat/pull/4145))
- Fixed Gemini Live local VAD mode (`GeminiVADParams(disabled=True)` with
external VAD) not working. The bot now correctly detects user speech and
signals turn boundaries to the Gemini API.
(PR [#4146](https://github.com/pipecat-ai/pipecat/pull/4146))
- Fixed Gemini Live message handling to process all `server_content` fields
independently. Gemini 3.x can bundle multiple fields (e.g. `model_turn` and
`output_transcription`) on the same message, but the previous `elif` chain
only processed the first match, silently dropping the rest.
(PR [#4147](https://github.com/pipecat-ai/pipecat/pull/4147))
- Fixed `ServiceSwitcher` with `ServiceSwitcherStrategyFailover` incorrectly
triggering failover when `ErrorFrame`s from other pipeline stages (e.g. TTS)
propagated upstream through the switcher. Previously, any non-fatal error
passing through would be misattributed to the active service and trigger an
unwanted service switch. Now only errors originating from the switcher's own
managed services trigger failover.
(PR [#4149](https://github.com/pipecat-ai/pipecat/pull/4149))
- Fixed `LiveKitOutputTransport` not clearing the `rtc.AudioSource` internal
buffer on interruption, causing the bot to continue speaking for several
seconds after being interrupted.
(PR [#4151](https://github.com/pipecat-ai/pipecat/pull/4151))
- Fixed a crash in OpenAI LLM processing when the provider returns
`chunk.choices[0].delta.audio = None`, which caused `'NoneType' object has no
attribute 'get'` errors during audio transcript handling.
(PR [#4152](https://github.com/pipecat-ai/pipecat/pull/4152))
- Fixed error floods in `DeepgramSTTService` when the WebSocket connection
drops. With Deepgram SDK 6.x, `send_media()` raises exceptions on a dead
connection instead of silently failing, causing every queued audio frame to
log an error. Now `send_media()` failures are caught gracefully — a single
warning is logged and audio frames are skipped until the existing
reconnection logic restores the connection.
(PR [#4153](https://github.com/pipecat-ai/pipecat/pull/4153))
- `Mem0MemoryService` no longer blocks the event loop during memory storage and
retrieval. All Mem0 API calls now run in a background thread, and message
storage is fire-and-forget so it doesn't delay downstream processing.
(PR [#4156](https://github.com/pipecat-ai/pipecat/pull/4156))
- Fixed `Mem0MemoryService` failing to store messages when the context
contained system or developer role messages. The Mem0 API only accepts user
and assistant roles, so other roles are now filtered out before storing.
(PR [#4156](https://github.com/pipecat-ai/pipecat/pull/4156))
- Added missing `on_dtmf_event` callback to `LemonSliceTransportClient.setup()`
`DailyCallbacks` construction, fixing a `ValidationError` at pipeline setup
time.
(PR [#4161](https://github.com/pipecat-ai/pipecat/pull/4161))
- Fixed an issue in `InworldTTSService` where, in cases of fast interruption,
we would continue receiving audio from the previous context.
(PR [#4167](https://github.com/pipecat-ai/pipecat/pull/4167))
- Fixed a word timestamp interleaving issue in `InworldTTSService` when
processing multiple sentences.
(PR [#4167](https://github.com/pipecat-ai/pipecat/pull/4167))
- Fixed duplicate `TTSStoppedFrame` being pushed in TTS services using
`push_stop_frames=True`. When the stop-frame timeout fired, a second
`TTSStoppedFrame` could be pushed after the normal one at context completion.
(PR [#4172](https://github.com/pipecat-ai/pipecat/pull/4172))
- ⚠️ Fixed `DeepgramSTTService` compatibility with deepgram-sdk 6.1.0. The SDK
now requires explicit message objects for `send_keep_alive()`,
`send_close_stream()`, and `send_finalize()`. The minimum deepgram-sdk
version is now 6.1.0.
(PR [#4174](https://github.com/pipecat-ai/pipecat/pull/4174))
- Fixed RTVI events not being delivered to clients when using WebSocket
transports. `ProtobufFrameSerializer` now sets `ignore_rtvi_messages=False`
by default.
(PR [#4176](https://github.com/pipecat-ai/pipecat/pull/4176))
- Fixed a timing issue where turn detection timer tasks (idle controller,
speech timeout, turn analyzer, and turn completion) could miss their first
tick because the newly created asyncio task was not yet scheduled when the
caller continued.
(PR [#4183](https://github.com/pipecat-ai/pipecat/pull/4183))
- Fixed `FastAPIWebsocketTransport` intermittently hanging on shutdown when the
remote side (e.g. Twilio) disconnects while audio is being sent. A race
condition between the send and receive paths could cause the
`on_client_disconnected` callback to be skipped, leaving the pipeline waiting
for a disconnect signal that never came.
(PR [#4186](https://github.com/pipecat-ai/pipecat/pull/4186))
### Performance
- `RimeTTSService` now handles Rime's `done` WebSocket message to complete
audio contexts immediately, eliminating the 3-second idle timeout that
previously added latency at the end of each utterance.
(PR [#4172](https://github.com/pipecat-ai/pipecat/pull/4172))
## [0.0.107] - 2026-03-23
### Added

View File

@@ -10,7 +10,7 @@ Pipecat is an open-source Python framework for building real-time voice and mult
```bash
# Setup development environment
uv sync --group dev --all-extras --no-extra gstreamer --no-extra krisp
uv sync --group dev --all-extras --no-extra gstreamer
# Install pre-commit hooks
uv run pre-commit install

View File

@@ -23,7 +23,7 @@ Create your integration following the patterns and examples shown in the "Integr
Your repository must contain these components:
- **Source code** - Complete implementation following Pipecat patterns
- **Foundational example** - Single file example showing basic usage (see [Pipecat examples](https://github.com/pipecat-ai/pipecat/tree/main/examples/foundational))
- **Foundational example** - Single file example showing basic usage (see [Pipecat examples](https://github.com/pipecat-ai/pipecat/tree/main/examples))
- **README.md** - Must include:
- Introduction and explanation of your integration
- Installation instructions
@@ -65,12 +65,25 @@ Once your PR is submitted, post in the `#community-integrations` Discord channel
#### Websocket-based Services
**Base class:** `WebsocketSTTService`
**Use for:** Services where you manage the websocket connection directly. Combines `STTService` with `WebsocketService` for automatic reconnection and keepalive support.
**Examples:**
- [CartesiaSTTService](https://github.com/pipecat-ai/pipecat/blob/main/src/pipecat/services/cartesia/stt.py)
- [ElevenLabsRealtimeSTTService](https://github.com/pipecat-ai/pipecat/blob/main/src/pipecat/services/elevenlabs/stt.py)
#### SDK-based Streaming Services
**Base class:** `STTService`
**Use for:** Streaming services where the provider's Python SDK manages the connection internally.
**Examples:**
- [DeepgramSTTService](https://github.com/pipecat-ai/pipecat/blob/main/src/pipecat/services/deepgram/stt.py)
- [SpeechmaticsSTTService](https://github.com/pipecat-ai/pipecat/blob/main/src/pipecat/services/speechmatics/stt.py)
- [GoogleSTTService](https://github.com/pipecat-ai/pipecat/blob/main/src/pipecat/services/google/stt.py)
#### File-based Services
@@ -108,55 +121,59 @@ Once your PR is submitted, post in the `#community-integrations` Discord channel
#### Key requirements:
- **Frame sequence:** Output must follow this frame sequence pattern:
- `LLMFullResponseStartFrame` - Signals the start of an LLM response
- `LLMTextFrame` - Contains LLM content, typically streamed as tokens
- `LLMFullResponseEndFrame` - Signals the end of an LLM response
- **`_process_context(self, context: LLMContext)`** — The main method that processes an LLM context and generates a response. Each LLM service overrides `process_frame` to extract context from `LLMContextFrame` and calls `_process_context`.
- **Context aggregation:** Implement context aggregation to collect user and assistant content:
- Aggregators come in pairs with a `user()` instance and `assistant()` instance
- Context must adhere to the `LLMContext` universal format
- Aggregators should handle adding messages, function calls, and images to the context
- **`adapter_class`** — Class attribute pointing to a `BaseLLMAdapter` subclass. Defaults to `OpenAILLMAdapter`. Non-OpenAI services must implement their own adapter (see `src/pipecat/adapters/base_llm_adapter.py`) with methods:
- `get_llm_invocation_params(context)` — Extract provider-specific params from universal context
- `to_provider_tools_format(tools_schema)` — Convert standard tools to provider format
- `get_messages_for_logging(context)` — Format messages for logging
- Reference adapters: `src/pipecat/adapters/services/` (anthropic, gemini, bedrock, etc.)
- **Frame sequence:** Output must follow this frame sequence pattern:
- `LLMFullResponseStartFrame` — Signals the start of an LLM response
- `LLMTextFrame` — Contains LLM content, typically streamed as tokens
- `LLMFullResponseEndFrame` — Signals the end of an LLM response
- **Thought frames (reasoning models):** If the model supports extended thinking / chain-of-thought, emit thought frames alongside the response:
- `LLMThoughtStartFrame` — Signals the start of a thought
- `LLMThoughtTextFrame` — Contains thought content, streamed as tokens
- `LLMThoughtEndFrame` — Signals the end of a thought
- **Context aggregation** is handled by the framework via `LLMContext` + `LLMContextAggregatorPair`. The LLM service just processes context it receives — no need to implement aggregators.
### TTS (Text-to-Speech) Services
#### AudioContextWordTTSService
#### WebsocketTTSService
**Use for:** Websocket-based services supporting word/timestamp alignment
**Use for:** Websocket-based streaming services (with or without word timestamps)
**Example:**
**Examples:**
- [CartesiaTTSService](https://github.com/pipecat-ai/pipecat/blob/main/src/pipecat/services/cartesia/tts.py)
- [ElevenLabsTTSService](https://github.com/pipecat-ai/pipecat/blob/main/src/pipecat/services/elevenlabs/tts.py)
#### InterruptibleTTSService
**Use for:** Websocket-based services without word/timestamp alignment, requiring disconnection on interruption
**Use for:** Websocket-based services without word timestamps that reconnect on interruption (e.g. don't support a context ID or interruption message)
**Example:**
- [SarvamTTSService](https://github.com/pipecat-ai/pipecat/blob/main/src/pipecat/services/sarvam/tts.py)
#### WordTTSService
**Use for:** HTTP-based services supporting word/timestamp alignment
**Example:**
- [ElevenLabsHttpTTSService](https://github.com/pipecat-ai/pipecat/blob/main/src/pipecat/services/elevenlabs/tts.py)
#### TTSService
**Use for:** HTTP-based services without word/timestamp alignment
**Use for:** HTTP-based services (word timestamps are supported in the base class)
**Example:**
**Examples:**
- [GoogleHttpTTSService](https://github.com/pipecat-ai/pipecat/blob/main/src/pipecat/services/google/tts.py)
- [OpenAITTSService](https://github.com/pipecat-ai/pipecat/blob/main/src/pipecat/services/openai/tts.py)
#### Key requirements:
- For websocket services, use asyncio WebSocket implementation (required for v13+ support)
- For websocket services, use asyncio WebSocket implementation
- Handle idle service timeouts with keepalives
- TTSServices push both audio (`TTSRawAudioFrame`) and text (`TTSTextFrame`) frames
- TTS services push both audio (`TTSAudioRawFrame`) and text (`TTSTextFrame`) frames
### Telephony Serializers
@@ -200,14 +217,25 @@ Vision services process images and provide analysis such as descriptions, object
#### Key requirements:
- Must implement `run_vision` method that takes an `LLMContext` and returns an `AsyncGenerator[Frame, None]`
- The method processes the latest image in the context and yields frames with analysis results
- Typically yields `TextFrame` objects containing descriptions or answers
- Must implement `run_vision` method that takes a `UserImageRawFrame` and returns an `AsyncGenerator[Frame, None]`
- The method processes the image frame and yields frames with analysis results
- Must yield the frame sequence: `VisionFullResponseStartFrame`, `VisionTextFrame`, `VisionFullResponseEndFrame`
## Implementation Guidelines
### Naming Conventions
#### Package and Repository Naming
Use the `pipecat-{vendor}` naming convention for your PyPI package and repository:
- `pipecat-{vendor}` — for single-service integrations (e.g., `pipecat-deepdub`)
- `pipecat-{vendor}-{type}` — when a vendor offers multiple service types (e.g., `pipecat-upliftai-stt`, `pipecat-upliftai-tts`)
This convention makes community packages easily discoverable via PyPI search and clearly identifies them as part of the Pipecat ecosystem.
#### Class Naming
- **STT:** `VendorSTTService`
- **LLM:** `VendorLLMService`
- **TTS:**
@@ -381,7 +409,7 @@ Note that `self.sample_rate` is a `@property` set in the TTSService base class,
Use Pipecat's tracing decorators:
- **STT:** `@traced_stt` - decorate a function that handles `transcript`, `is_final`, `language` as args
- **STT:** `@traced_stt` - decorate `_handle_transcription(self, transcript, is_final, language)` (the standard method name convention)
- **LLM:** `@traced_llm` - decorate the `_process_context()` method
- **TTS:** `@traced_tts` - decorate the `run_tts()` method
@@ -389,8 +417,9 @@ Use Pipecat's tracing decorators:
### Packaging and Distribution
- Name your package `pipecat-{vendor}` (see [Naming Conventions](#naming-conventions))
- Use [uv](https://docs.astral.sh/uv/) for packaging (encouraged)
- Consider releasing to PyPI for easier installation
- Publish to PyPI for easier installation
- Follow semantic versioning principles
- Maintain a changelog
@@ -403,17 +432,15 @@ For REST-based communication, use aiohttp. Pipecat includes this as a required d
- Wrap API calls in appropriate try/catch blocks
- Handle rate limits and network failures gracefully
- Provide meaningful error messages
- When errors occur, raise exceptions AND push `ErrorFrame`s to notify the pipeline:
- When errors occur, raise exceptions AND push errors to notify the pipeline:
```python
from pipecat.frames.frames import ErrorFrame
try:
# Your API call
result = await self._make_api_call()
except Exception as e:
# Push error frame to pipeline
await self.push_error(ErrorFrame(error=f"{self} error: {e}"))
# Push error upstream to notify the pipeline
await self.push_error(f"{self} error: {e}", exception=e)
# Raise or handle as appropriate
raise
```

View File

@@ -8,7 +8,7 @@
**Pipecat** is an open-source Python framework for building real-time voice and multimodal conversational agents. Orchestrate audio and video, AI services, different transports, and conversation pipelines effortlessly—so you can focus on what makes your agent unique.
> Want to dive right in? Try the [quickstart](https://docs.pipecat.ai/getting-started/quickstart).
> Want to dive right in? Run `pipecat init quickstart` or follow the [quickstart guide](https://docs.pipecat.ai/getting-started/quickstart).
## 🚀 What You Can Build
@@ -80,25 +80,25 @@ Catch new features, interviews, and how-tos on our [Pipecat TV](https://www.yout
<a href="https://github.com/pipecat-ai/pipecat-examples/tree/main/storytelling-chatbot"><img src="https://raw.githubusercontent.com/pipecat-ai/pipecat-examples/main/storytelling-chatbot/image.png" width="400" /></a>
<br/>
<a href="https://github.com/pipecat-ai/pipecat-examples/tree/main/translation-chatbot"><img src="https://raw.githubusercontent.com/pipecat-ai/pipecat-examples/main/translation-chatbot/image.png" width="400" /></a>&nbsp;
<a href="https://github.com/pipecat-ai/pipecat/blob/main/examples/foundational/12-describe-video.py"><img src="https://github.com/pipecat-ai/pipecat/blob/main/examples/foundational/assets/moondream.png" width="400" /></a>
<a href="https://github.com/pipecat-ai/pipecat/blob/main/examples/vision/vision-moondream.py"><img src="https://github.com/pipecat-ai/pipecat/blob/main/examples/assets/moondream.png" width="400" /></a>
</p>
## 🧩 Available services
| Category | Services |
| ------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| Speech-to-Text | [AssemblyAI](https://docs.pipecat.ai/server/services/stt/assemblyai), [AWS](https://docs.pipecat.ai/server/services/stt/aws), [Azure](https://docs.pipecat.ai/server/services/stt/azure), [Cartesia](https://docs.pipecat.ai/server/services/stt/cartesia), [Deepgram](https://docs.pipecat.ai/server/services/stt/deepgram), [ElevenLabs](https://docs.pipecat.ai/server/services/stt/elevenlabs), [Fal Wizper](https://docs.pipecat.ai/server/services/stt/fal), [Gladia](https://docs.pipecat.ai/server/services/stt/gladia), [Google](https://docs.pipecat.ai/server/services/stt/google), [Gradium](https://docs.pipecat.ai/server/services/stt/gradium), [Groq (Whisper)](https://docs.pipecat.ai/server/services/stt/groq), [NVIDIA Riva](https://docs.pipecat.ai/server/services/stt/riva), [OpenAI (Whisper)](https://docs.pipecat.ai/server/services/stt/openai), [SambaNova (Whisper)](https://docs.pipecat.ai/server/services/stt/sambanova), [Sarvam](https://docs.pipecat.ai/server/services/stt/sarvam), [Soniox](https://docs.pipecat.ai/server/services/stt/soniox), [Speechmatics](https://docs.pipecat.ai/server/services/stt/speechmatics), [Whisper](https://docs.pipecat.ai/server/services/stt/whisper) |
| LLMs | [Anthropic](https://docs.pipecat.ai/server/services/llm/anthropic), [AWS](https://docs.pipecat.ai/server/services/llm/aws), [Azure](https://docs.pipecat.ai/server/services/llm/azure), [Cerebras](https://docs.pipecat.ai/server/services/llm/cerebras), [DeepSeek](https://docs.pipecat.ai/server/services/llm/deepseek), [Fireworks AI](https://docs.pipecat.ai/server/services/llm/fireworks), [Gemini](https://docs.pipecat.ai/server/services/llm/gemini), [Grok](https://docs.pipecat.ai/server/services/llm/grok), [Groq](https://docs.pipecat.ai/server/services/llm/groq), [Mistral](https://docs.pipecat.ai/server/services/llm/mistral), [NVIDIA NIM](https://docs.pipecat.ai/server/services/llm/nim), [Ollama](https://docs.pipecat.ai/server/services/llm/ollama), [OpenAI](https://docs.pipecat.ai/server/services/llm/openai), [OpenRouter](https://docs.pipecat.ai/server/services/llm/openrouter), [Perplexity](https://docs.pipecat.ai/server/services/llm/perplexity), [Qwen](https://docs.pipecat.ai/server/services/llm/qwen), [SambaNova](https://docs.pipecat.ai/server/services/llm/sambanova) [Together AI](https://docs.pipecat.ai/server/services/llm/together) |
| Text-to-Speech | [Async](https://docs.pipecat.ai/server/services/tts/asyncai), [AWS](https://docs.pipecat.ai/server/services/tts/aws), [Azure](https://docs.pipecat.ai/server/services/tts/azure), [Camb AI](https://docs.pipecat.ai/server/services/tts/camb), [Cartesia](https://docs.pipecat.ai/server/services/tts/cartesia), [Deepgram](https://docs.pipecat.ai/server/services/tts/deepgram), [ElevenLabs](https://docs.pipecat.ai/server/services/tts/elevenlabs), [Fish](https://docs.pipecat.ai/server/services/tts/fish), [Google](https://docs.pipecat.ai/server/services/tts/google), [Gradium](https://docs.pipecat.ai/server/services/tts/gradium), [Groq](https://docs.pipecat.ai/server/services/tts/groq), [Hume](https://docs.pipecat.ai/server/services/tts/hume), [Inworld](https://docs.pipecat.ai/server/services/tts/inworld), [LMNT](https://docs.pipecat.ai/server/services/tts/lmnt), [MiniMax](https://docs.pipecat.ai/server/services/tts/minimax), [Neuphonic](https://docs.pipecat.ai/server/services/tts/neuphonic), [NVIDIA Riva](https://docs.pipecat.ai/server/services/tts/riva), [OpenAI](https://docs.pipecat.ai/server/services/tts/openai), [Piper](https://docs.pipecat.ai/server/services/tts/piper), [Resemble](https://docs.pipecat.ai/server/services/tts/resemble), [Rime](https://docs.pipecat.ai/server/services/tts/rime), [Sarvam](https://docs.pipecat.ai/server/services/tts/sarvam), [Speechmatics](https://docs.pipecat.ai/server/services/tts/speechmatics), [XTTS](https://docs.pipecat.ai/server/services/tts/xtts) |
| Speech-to-Speech | [AWS Nova Sonic](https://docs.pipecat.ai/server/services/s2s/aws), [Gemini Multimodal Live](https://docs.pipecat.ai/server/services/s2s/gemini), [Grok Voice Agent](https://docs.pipecat.ai/server/services/s2s/grok), [OpenAI Realtime](https://docs.pipecat.ai/server/services/s2s/openai), [Ultravox](https://docs.pipecat.ai/server/services/s2s/ultravox), |
| Transport | [Daily (WebRTC)](https://docs.pipecat.ai/server/services/transport/daily), [FastAPI Websocket](https://docs.pipecat.ai/server/services/transport/fastapi-websocket), [SmallWebRTCTransport](https://docs.pipecat.ai/server/services/transport/small-webrtc), [WebSocket Server](https://docs.pipecat.ai/server/services/transport/websocket-server), Local |
| Serializers | [Exotel](https://docs.pipecat.ai/server/utilities/serializers/exotel), [Plivo](https://docs.pipecat.ai/server/utilities/serializers/plivo), [Twilio](https://docs.pipecat.ai/server/utilities/serializers/twilio), [Telnyx](https://docs.pipecat.ai/server/utilities/serializers/telnyx), [Vonage](https://docs.pipecat.ai/server/utilities/serializers/vonage) |
| Video | [HeyGen](https://docs.pipecat.ai/server/services/video/heygen), [LemonSlice](https://docs.pipecat.ai/server/services/video/lemonslice), [Tavus](https://docs.pipecat.ai/server/services/video/tavus), [Simli](https://docs.pipecat.ai/server/services/video/simli) |
| Memory | [mem0](https://docs.pipecat.ai/server/services/memory/mem0) |
| Vision & Image | [fal](https://docs.pipecat.ai/server/services/image-generation/fal), [Google Imagen](https://docs.pipecat.ai/server/services/image-generation/google-imagen), [Moondream](https://docs.pipecat.ai/server/services/vision/moondream) |
| Audio Processing | [Silero VAD](https://docs.pipecat.ai/server/utilities/audio/silero-vad-analyzer), [Krisp](https://docs.pipecat.ai/server/utilities/audio/krisp-filter), [Koala](https://docs.pipecat.ai/server/utilities/audio/koala-filter), [ai-coustics](https://docs.pipecat.ai/server/utilities/audio/aic-filter) |
| Analytics & Metrics | [OpenTelemetry](https://docs.pipecat.ai/server/utilities/opentelemetry), [Sentry](https://docs.pipecat.ai/server/services/analytics/sentry) |
| Community | [Browse community integrations →](https://docs.pipecat.ai/server/services/community-integrations) |
| Category | Services |
| ------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| Speech-to-Text | [AssemblyAI](https://docs.pipecat.ai/server/services/stt/assemblyai), [AWS](https://docs.pipecat.ai/server/services/stt/aws), [Azure](https://docs.pipecat.ai/server/services/stt/azure), [Cartesia](https://docs.pipecat.ai/server/services/stt/cartesia), [Deepgram](https://docs.pipecat.ai/server/services/stt/deepgram), [ElevenLabs](https://docs.pipecat.ai/server/services/stt/elevenlabs), [Fal Wizper](https://docs.pipecat.ai/server/services/stt/fal), [Gladia](https://docs.pipecat.ai/server/services/stt/gladia), [Google](https://docs.pipecat.ai/server/services/stt/google), [Gradium](https://docs.pipecat.ai/server/services/stt/gradium), [Groq (Whisper)](https://docs.pipecat.ai/server/services/stt/groq), [NVIDIA Riva](https://docs.pipecat.ai/server/services/stt/riva), [OpenAI (Whisper)](https://docs.pipecat.ai/server/services/stt/openai), [Sarvam](https://docs.pipecat.ai/server/services/stt/sarvam), [Soniox](https://docs.pipecat.ai/server/services/stt/soniox), [Speechmatics](https://docs.pipecat.ai/server/services/stt/speechmatics), [Whisper](https://docs.pipecat.ai/server/services/stt/whisper) |
| LLMs | [Anthropic](https://docs.pipecat.ai/server/services/llm/anthropic), [AWS](https://docs.pipecat.ai/server/services/llm/aws), [Azure](https://docs.pipecat.ai/server/services/llm/azure), [Cerebras](https://docs.pipecat.ai/server/services/llm/cerebras), [DeepSeek](https://docs.pipecat.ai/server/services/llm/deepseek), [Fireworks AI](https://docs.pipecat.ai/server/services/llm/fireworks), [Gemini](https://docs.pipecat.ai/server/services/llm/gemini), [Grok](https://docs.pipecat.ai/server/services/llm/grok), [Groq](https://docs.pipecat.ai/server/services/llm/groq), [Mistral](https://docs.pipecat.ai/server/services/llm/mistral), [Nebius](https://docs.pipecat.ai/server/services/llm/nebius), [Novita](https://docs.pipecat.ai/server/services/llm/novita), [NVIDIA NIM](https://docs.pipecat.ai/server/services/llm/nvidia), [Ollama](https://docs.pipecat.ai/server/services/llm/ollama), [OpenAI](https://docs.pipecat.ai/server/services/llm/openai), [OpenRouter](https://docs.pipecat.ai/server/services/llm/openrouter), [Perplexity](https://docs.pipecat.ai/server/services/llm/perplexity), [Qwen](https://docs.pipecat.ai/server/services/llm/qwen), [SambaNova](https://docs.pipecat.ai/server/services/llm/sambanova), [Sarvam](https://docs.pipecat.ai/server/services/llm/sarvam), [Together AI](https://docs.pipecat.ai/server/services/llm/together) |
| Text-to-Speech | [Async](https://docs.pipecat.ai/server/services/tts/asyncai), [AWS](https://docs.pipecat.ai/server/services/tts/aws), [Azure](https://docs.pipecat.ai/server/services/tts/azure), [Camb AI](https://docs.pipecat.ai/server/services/tts/camb), [Cartesia](https://docs.pipecat.ai/server/services/tts/cartesia), [Deepgram](https://docs.pipecat.ai/server/services/tts/deepgram), [ElevenLabs](https://docs.pipecat.ai/server/services/tts/elevenlabs), [Fish](https://docs.pipecat.ai/server/services/tts/fish), [Google](https://docs.pipecat.ai/server/services/tts/google), [Gradium](https://docs.pipecat.ai/server/services/tts/gradium), [Groq](https://docs.pipecat.ai/server/services/tts/groq), [Hume](https://docs.pipecat.ai/server/services/tts/hume), [Inworld](https://docs.pipecat.ai/server/services/tts/inworld), [Kokoro](https://docs.pipecat.ai/server/services/tts/kokoro), [LMNT](https://docs.pipecat.ai/server/services/tts/lmnt), [MiniMax](https://docs.pipecat.ai/server/services/tts/minimax), [Neuphonic](https://docs.pipecat.ai/server/services/tts/neuphonic), [NVIDIA Riva](https://docs.pipecat.ai/server/services/tts/riva), [OpenAI](https://docs.pipecat.ai/server/services/tts/openai), [Piper](https://docs.pipecat.ai/server/services/tts/piper), [Resemble](https://docs.pipecat.ai/server/services/tts/resemble), [Rime](https://docs.pipecat.ai/server/services/tts/rime), [Sarvam](https://docs.pipecat.ai/server/services/tts/sarvam), [Smallest](https://docs.pipecat.ai/server/services/tts/smallest), [Speechmatics](https://docs.pipecat.ai/server/services/tts/speechmatics), [xAI](https://docs.pipecat.ai/server/services/tts/xai), [XTTS](https://docs.pipecat.ai/server/services/tts/xtts) |
| Speech-to-Speech | [AWS Nova Sonic](https://docs.pipecat.ai/server/services/s2s/aws), [Gemini Multimodal Live](https://docs.pipecat.ai/server/services/s2s/gemini), [Grok Voice Agent](https://docs.pipecat.ai/server/services/s2s/grok), [OpenAI Realtime](https://docs.pipecat.ai/server/services/s2s/openai), [Ultravox](https://docs.pipecat.ai/server/services/s2s/ultravox), |
| Transport | [Daily (WebRTC)](https://docs.pipecat.ai/server/services/transport/daily), [FastAPI Websocket](https://docs.pipecat.ai/server/services/transport/fastapi-websocket), [LiveKit (WebRTC)](https://docs.pipecat.ai/server/services/transport/livekit), [SmallWebRTCTransport](https://docs.pipecat.ai/server/services/transport/small-webrtc), [WebSocket Server](https://docs.pipecat.ai/server/services/transport/websocket-server), [WhatsApp](https://docs.pipecat.ai/server/services/transport/whatsapp), Local |
| Serializers | [Exotel](https://docs.pipecat.ai/server/services/serializers/exotel), [Genesys](https://docs.pipecat.ai/server/services/serializers/genesys), [Plivo](https://docs.pipecat.ai/server/services/serializers/plivo), [Twilio](https://docs.pipecat.ai/server/services/serializers/twilio), [Telnyx](https://docs.pipecat.ai/server/services/serializers/telnyx), [Vonage](https://docs.pipecat.ai/server/services/serializers/vonage) |
| Video | [HeyGen](https://docs.pipecat.ai/server/services/video/heygen), [LemonSlice](https://docs.pipecat.ai/server/services/transport/lemonslice), [Tavus](https://docs.pipecat.ai/server/services/video/tavus), [Simli](https://docs.pipecat.ai/server/services/video/simli) |
| Memory | [mem0](https://docs.pipecat.ai/server/services/memory/mem0) |
| Vision & Image | [fal](https://docs.pipecat.ai/server/services/image-generation/fal), [Google Imagen](https://docs.pipecat.ai/server/services/image-generation/google-imagen), [Moondream](https://docs.pipecat.ai/server/services/vision/moondream) |
| Audio Processing | [Silero VAD](https://docs.pipecat.ai/server/utilities/audio/silero-vad-analyzer), [Krisp Viva](https://docs.pipecat.ai/guides/features/krisp-viva), [Koala](https://docs.pipecat.ai/server/utilities/audio/koala-filter), [ai-coustics](https://docs.pipecat.ai/server/utilities/audio/aic-filter), [RNNoise](https://docs.pipecat.ai/server/utilities/audio/rnnoise-filter) |
| Analytics & Metrics | [OpenTelemetry](https://docs.pipecat.ai/server/utilities/opentelemetry), [Sentry](https://docs.pipecat.ai/server/services/analytics/sentry) |
| Community | [Browse community integrations →](https://docs.pipecat.ai/server/services/community-integrations) |
📚 [View full services documentation →](https://docs.pipecat.ai/server/services/supported-services)
@@ -142,7 +142,7 @@ You can get started with Pipecat running on your local machine, then move your a
## 🧪 Code examples
- [Foundational](https://github.com/pipecat-ai/pipecat/tree/main/examples/foundational) — small snippets that build on each other, introducing one or two concepts at a time
- [Foundational](https://github.com/pipecat-ai/pipecat/tree/main/examples) — small snippets that build on each other, introducing one or two concepts at a time
- [Example apps](https://github.com/pipecat-ai/pipecat-examples) — complete applications that you can use as starting points for development
## 🛠️ Contributing to the framework
@@ -166,7 +166,6 @@ You can get started with Pipecat running on your local machine, then move your a
```bash
uv sync --group dev --all-extras \
--no-extra gstreamer \
--no-extra krisp \
--no-extra local \
```

1
changelog/4141.added.md Normal file
View File

@@ -0,0 +1 @@
- ⚠️ Added WebSocket-based `OpenAIResponsesLLMService` as the new default for the OpenAI Responses API. It maintains a persistent connection to `wss://api.openai.com/v1/responses` and automatically uses `previous_response_id` to send only incremental context, falling back to full context on reconnection or cache miss. The previous HTTP-based implementation is now available as `OpenAIResponsesHttpLLMService`.

View File

@@ -0,0 +1 @@
- ⚠️ Removed `OpenPipeLLMService` and the `openpipe` extra. OpenPipe was acquired by CoreWeave and the package is no longer maintained. If you were using `openpipe` as an LLM provider, switch to the underlying provider directly (e.g. `openai`). The OpenPipe interface can still be used with `OpenAILLMService` by specifying a `base_url`.

View File

@@ -0,0 +1 @@
- ⚠️ Updated `langchain` extra to require langchain 1.x (from 0.3.x), langchain-community 0.4.x (from 0.3.x), and langchain-openai 1.x (from 0.3.x). If you pin these packages in your project, update your pins accordingly.

1
changelog/4202.fixed.md Normal file
View File

@@ -0,0 +1 @@
- Fixed `InworldHttpTTSService` streaming responses crashing with `UnicodeDecodeError` when multi-byte UTF-8 characters were split across chunk boundaries. This caused TTS audio to cut off mid-sentence intermittently.

1
changelog/4203.fixed.md Normal file
View File

@@ -0,0 +1 @@
- Fixed a crash (`JSONDecodeError`) when a user interruption occurs while the LLM is streaming function call arguments. Previously, the incomplete JSON arguments were passed directly to `json.loads()`, causing an unhandled exception. Affected services: OpenAI, Google (OpenAI-compatible), and SambaNova.

View File

@@ -0,0 +1 @@
- ⚠️ Removed deprecated `observers` field from `PipelineParams`. Pass observers directly to `PipelineTask` constructor instead.

View File

@@ -0,0 +1 @@
- ⚠️ Removed deprecated `on_pipeline_ended`, `on_pipeline_cancelled`, and `on_pipeline_stopped` events from `PipelineTask`. Use `on_pipeline_finished` instead.

View File

@@ -0,0 +1 @@
- ⚠️ Removed `AudioBufferProcessor.user_continuous_stream` parameter. Use `user_audio_passthrough` instead.

View File

@@ -0,0 +1 @@
- ⚠️ Removed deprecated `camera_in_enabled`, `camera_in_is_live`, `camera_in_width`, `camera_in_height`, `camera_out_enabled`, `camera_out_is_live`, `camera_out_width`, `camera_out_height`, and `camera_out_color` transport params. Use the `video_in_*` and `video_out_*` equivalents instead.

View File

@@ -0,0 +1 @@
- ⚠️ Removed `RTVIObserver.errors_enabled` parameter.

View File

@@ -0,0 +1 @@
- ⚠️ Removed deprecated `vad_enabled` and `vad_audio_passthrough` transport params.

View File

@@ -0,0 +1 @@
- ⚠️ Removed `TTSService.say()`. Push a `TTSSpeakFrame` into the pipeline instead.

View File

@@ -0,0 +1 @@
- ⚠️ Removed `DailyRunner.configure_with_args()`. Use `PipelineRunner` with `RunnerArguments` instead.

View File

@@ -0,0 +1 @@
- ⚠️ Removed deprecated RTVI models, frames, and processor methods including `RTVIConfig`, `RTVIServiceConfig`, `RTVIServiceOptionConfig`, various `RTVI*Data` models, `RTVIActionFrame`, and `RTVIProcessor.handle_function_call`/`handle_function_call_start`. Use the updated RTVI processor API instead.

View File

@@ -0,0 +1 @@
- ⚠️ Removed `FrameProcessor.wait_for_task()`. Use `create_task()` and manage tasks with the built-in `TaskManager` instead.

View File

@@ -0,0 +1 @@
- ⚠️ Removed `KrispFilter`. The `krisp` extra has been removed from `pyproject.toml`.

View File

@@ -0,0 +1 @@
- ⚠️ Removed `LLMService.request_image_frame()`. Push a `UserImageRequestFrame` instead.

View File

@@ -0,0 +1 @@
- ⚠️ Removed `create_default_resampler()` from `pipecat.audio.utils`.

View File

@@ -0,0 +1 @@
- ⚠️ Removed `FalSmartTurnAnalyzer` and `LocalSmartTurnAnalyzer`.

View File

@@ -0,0 +1 @@
- ⚠️ Removed deprecated transport frames: `TransportMessageFrame`, `TransportMessageUrgentFrame`, `InputTransportMessageUrgentFrame`, `DailyTransportMessageFrame`, and `DailyTransportMessageUrgentFrame`. Use `OutputTransportMessageFrame`, `OutputTransportMessageUrgentFrame`, `InputTransportMessageFrame`, `DailyOutputTransportMessageFrame`, and `DailyOutputTransportMessageUrgentFrame` instead.

View File

@@ -0,0 +1 @@
- ⚠️ Removed deprecated `KeypadEntryFrame` alias.

View File

@@ -0,0 +1 @@
- ⚠️ Removed deprecated interruption frames: `StartInterruptionFrame` and `BotInterruptionFrame`. Use `InterruptionFrame` and `InterruptionTaskFrame` instead.

View File

@@ -0,0 +1 @@
- ⚠️ Removed `LLMService.start_callback` parameter. Register an `on_llm_response_start` event handler instead.

View File

@@ -0,0 +1 @@
- ⚠️ Removed single-argument function call support from `LLMService`. Functions must use named parameters instead of a single `arguments` parameter.

View File

@@ -0,0 +1 @@
- ⚠️ Removed `NoisereduceFilter`. Use system-level noise reduction or a service-based alternative instead.

View File

@@ -0,0 +1 @@
- ⚠️ Removed deprecated `pipecat.services.riva` package. Use `pipecat.services.nvidia.stt` and `pipecat.services.nvidia.tts` instead (`RivaSTTService``NvidiaSTTService`, `RivaTTSService``NvidiaTTSService`).

View File

@@ -0,0 +1 @@
- ⚠️ Removed deprecated `pipecat.services.nim` package. Use `pipecat.services.nvidia.llm` instead (`NimLLMService``NvidiaLLMService`).

View File

@@ -0,0 +1 @@
- ⚠️ Removed deprecated `pipecat.services.gemini_multimodal_live` package. Use `pipecat.services.google.gemini_live` instead. Note that class names no longer include "Multimodal" (e.g. `GeminiMultimodalLiveLLMService``GeminiLiveLLMService`).

View File

@@ -0,0 +1 @@
- ⚠️ Removed deprecated `pipecat.services.aws_nova_sonic` package. Use `pipecat.services.aws.nova_sonic` instead.

View File

@@ -0,0 +1 @@
- ⚠️ Removed deprecated `pipecat.services.openai_realtime` package. Use `pipecat.services.openai.realtime` instead.

View File

@@ -0,0 +1 @@
- ⚠️ Removed deprecated `OpenAIRealtimeBetaLLMService` and `AzureRealtimeBetaLLMService`. Use `OpenAIRealtimeLLMService` and `AzureRealtimeLLMService` from `pipecat.services.openai.realtime` and `pipecat.services.azure.realtime` instead.

View File

@@ -0,0 +1 @@
- ⚠️ Removed deprecated `pipecat.services.deepgram.stt_sagemaker` and `pipecat.services.deepgram.tts_sagemaker` modules. Use `pipecat.services.deepgram.sagemaker.stt` and `pipecat.services.deepgram.sagemaker.tts` instead.

View File

@@ -0,0 +1 @@
- ⚠️ Removed deprecated `GoogleLLMOpenAIBetaService` from `pipecat.services.google.openai`. Use `GoogleLLMService` from `pipecat.services.google.llm` instead.

View File

@@ -0,0 +1 @@
- ⚠️ Removed deprecated `pipecat.services.google.llm_vertex` module. Use `pipecat.services.google.vertex.llm` instead.

View File

@@ -0,0 +1 @@
- ⚠️ Removed deprecated `pipecat.services.google.gemini_live.llm_vertex` module. Use `pipecat.services.google.gemini_live.vertex.llm` instead.

View File

@@ -0,0 +1 @@
- ⚠️ Removed deprecated `pipecat.services.ai_services` module. Import from `pipecat.services.ai_service`, `pipecat.services.llm_service`, `pipecat.services.stt_service`, `pipecat.services.tts_service`, etc. instead.

View File

@@ -0,0 +1 @@
- Changed `GrokLLMService` default model from `grok-3-beta` to `grok-3`, now that the model is generally available.

View File

@@ -0,0 +1 @@
- `GoogleImageGenService` now defaults to `imagen-4.0-generate-001` (previously `imagen-3.0-generate-002`).

View File

@@ -0,0 +1 @@
- ⚠️ `BaseOpenAILLMService.get_chat_completions()` now accepts an `LLMContext` instead of `OpenAILLMInvocationParams`. If you override this method, update your signature accordingly.

View File

@@ -0,0 +1,22 @@
- ⚠️ Removed deprecated service-specific context and aggregator machinery, which was superseded by the universal `LLMContext` system.
Service-specific classes removed: `AnthropicLLMContext`, `AnthropicContextAggregatorPair`, `AWSBedrockLLMContext`, `AWSBedrockContextAggregatorPair`, `OpenAIContextAggregatorPair`, and their user/assistant aggregators. Also removed `create_context_aggregator()` from `LLMService`, `OpenAILLMService`, `AnthropicLLMService`, and `AWSBedrockLLMService`.
Base aggregator classes removed (from `pipecat.processors.aggregators.llm_response`): `BaseLLMResponseAggregator`, `LLMContextResponseAggregator`, `LLMUserContextAggregator`, `LLMAssistantContextAggregator`, `LLMUserResponseAggregator`, `LLMAssistantResponseAggregator`.
From the developer's point of view, migrating will usually be a matter of going from this:
```python
context = OpenAILLMContext(messages, tools)
context_aggregator = llm.create_context_aggregator(context)
```
To this:
```python
from pipecat.processors.aggregators.llm_context import LLMContext
from pipecat.processors.aggregators.llm_response_universal import LLMContextAggregatorPair
context = LLMContext(messages, tools)
context_aggregator = LLMContextAggregatorPair(context)
```

View File

@@ -0,0 +1 @@
- ⚠️ Removed deprecated frame types `LLMMessagesFrame` and `OpenAILLMContextAssistantTimestampFrame` from `pipecat.frames.frames`. Instead of `LLMMessagesFrame`, use `LLMContextFrame` with the new messages, or `LLMMessagesUpdateFrame` with `run_llm=True`.

View File

@@ -0,0 +1 @@
- ⚠️ Removed `GatedOpenAILLMContextAggregator` (from `pipecat.processors.aggregators.gated_open_ai_llm_context`). Use `GatedLLMContextAggregator` (from `pipecat.processors.aggregators.gated_llm_context`) instead.

View File

@@ -0,0 +1 @@
- ⚠️ Removed `VisionImageFrameAggregator` (from `pipecat.processors.aggregators.vision_image_frame`). Vision/image handling is now built into `LLMContext` (from `pipecat.processors.aggregators.llm_context`). See the `12*` examples for the recommended replacement pattern.

View File

@@ -0,0 +1 @@
- ⚠️ Removed deprecated compatibility modules: `pipecat.services.openai_realtime_beta` (use `pipecat.services.openai.realtime`), `pipecat.services.openai_realtime.context`, `pipecat.services.openai_realtime.frames`, `pipecat.services.openai.realtime.context`, `pipecat.services.openai.realtime.frames`, `pipecat.services.gemini_multimodal_live` (use `pipecat.services.google.gemini_live`), `pipecat.services.aws_nova_sonic.context` (use `pipecat.services.aws.nova_sonic`), `pipecat.services.google.openai` and `pipecat.services.google.llm_openai` (use `pipecat.services.google.llm`).

18
changelog/4215.removed.md Normal file
View File

@@ -0,0 +1,18 @@
- ⚠️ Removed `OpenAILLMContext`, `OpenAILLMContextFrame`, and `OpenAILLMContext.from_messages()`. Use `LLMContext` (from `pipecat.processors.aggregators.llm_context`) and `LLMContextFrame` (from `pipecat.frames.frames`) instead. All services now exclusively use the universal `LLMContext`.
From the developer's point of view, migrating will usually be a matter of going from this:
```python
context = OpenAILLMContext(messages, tools)
context_aggregator = llm.create_context_aggregator(context)
```
To this:
```python
from pipecat.processors.aggregators.llm_context import LLMContext
from pipecat.processors.aggregators.llm_response_universal import LLMContextAggregatorPair
context = LLMContext(messages, tools)
context_aggregator = LLMContextAggregatorPair(context)
```

View File

@@ -0,0 +1 @@
- Added `group_parallel_tools` parameter to `LLMService` (default `True`). When `True`, all function calls from the same LLM response batch share a group ID and the LLM is triggered exactly once after the last call completes. Set to `False` to trigger inference independently for each function call result as it arrives.

1
changelog/4217.added.md Normal file
View File

@@ -0,0 +1 @@
- Added `is_async=True` support to `register_function()` and `register_direct_function()`. When enabled, the LLM continues the conversation immediately without waiting for the function result. The result is injected back into the context as a `developer` message once available, triggering a new LLM inference at that point.

View File

@@ -0,0 +1 @@
- When multiple function calls are returned in a single LLM response, the LLM is now triggered exactly once after the last call in the batch completes, rather than waiting for all function calls.

View File

@@ -0,0 +1 @@
- Fixed `BaseOutputTransport` discarding pending `UninterruptibleFrame` items (e.g. function-call context updates) when an interruption arrived. The audio task is now kept alive and only interruptible frames are drained when uninterruptible frames are present in the queue.

View File

@@ -0,0 +1 @@
- Fixed spurious LLM inference being triggered when a function call result arrived while the user was actively speaking. The context frame is now suppressed until the user stops speaking.

1
changelog/4217.fixed.md Normal file
View File

@@ -0,0 +1 @@
- Fixed an issue where `UninterruptibleFrame` items queued in `FrameProcessor` could be incorrectly dropped on interruption. Previously only the frame currently being processed was checked; now the entire process queue is scanned so pending uninterruptible frames are always delivered.

View File

@@ -2,7 +2,7 @@
# Build docs using uv
echo "Installing dependencies with uv..."
uv sync --group docs --all-extras --no-extra krisp --no-extra gstreamer --no-extra local_smart_turn --no-extra moondream --no-extra riva --no-extra mlx-whisper
uv sync --group docs --all-extras --no-extra gstreamer --no-extra local_smart_turn --no-extra moondream --no-extra riva --no-extra mlx-whisper
# Check if sphinx-build is available
if ! uv run sphinx-build --version &> /dev/null; then

View File

@@ -48,8 +48,6 @@ autodoc_default_options = {
# Mock imports for optional dependencies
autodoc_mock_imports = [
# Krisp - has build issues on some platforms
"pipecat_ai_krisp",
"krisp",
"krisp_audio",
# System-specific GUI libraries
"_tkinter",
@@ -98,7 +96,6 @@ autodoc_mock_imports = [
"cartesia",
"camb",
"sarvamai",
"openpipe",
"openai.types.beta.realtime",
"langchain_core",
"langchain_core.messages",

View File

@@ -80,9 +80,6 @@ GOOGLE_TEST_CREDENTIALS=...
# Gradium
GRAPDIUM_API_KEY=...
# Grok
GROK_API_KEY=...
# Groq
GROQ_API_KEY=...
@@ -124,18 +121,21 @@ MINIMAX_GROUP_ID=...
# Mistral
MISTRAL_API_KEY=...
# Nebius
NEBIUS_API_KEY=...
# Neuphonic
NEUPHONIC_API_KEY=...
# Novita
NOVITA_API_KEY=...
# NVIDIA
NVIDIA_API_KEY=...
# OpenAI
OPENAI_API_KEY=...
# OpenPipe
OPENPIPE_API_KEY=...
# OpenRouter
OPENROUTER_API_KEY=...
@@ -176,6 +176,9 @@ SENTRY_DSN=...
SIMLI_API_KEY=...
SIMLI_FACE_ID=...
# Smallest
SMALLEST_API_KEY=...
# Smart turn
LOCAL_SMART_TURN_MODEL_PATH=...
FAL_SMART_TURN_API_KEY=...
@@ -209,3 +212,6 @@ WHATSAPP_TOKEN=...
WHATSAPP_WEBHOOK_VERIFICATION_TOKEN=...
WHATSAPP_PHONE_NUMBER_ID=...
WHATSAPP_APP_SECRET=...
# xAI / Grok
XAI_API_KEY=...

View File

@@ -1,31 +1,150 @@
# Pipecat Examples
This directory contains examples to help you learn how to build with Pipecat.
This directory contains examples showing how to build voice and multimodal agents with Pipecat.
## Getting Started
## Setup
New to Pipecat? Start here:
1. Follow the [README](https://github.com/pipecat-ai/pipecat/blob/main/README.md#%EF%B8%8F-contributing-to-the-framework) steps to get your local environment configured.
- **[Quickstart](quickstart/)** - Get your first voice AI bot running in 5 minutes _(coming soon)_
- **[Client/Server Web](client-server-web/)** - Learn to build web applications with Pipecat's client SDKs _(coming soon)_
- **[Phone Bot with Twilio](phone-bot-twilio/)** - Connect your bot to a phone number _(coming soon)_
> **Run from root directory**: Make sure you are running the steps from the root directory.
## Foundational Examples
> **Using local audio?**: The `LocalAudioTransport` requires a system dependency for `portaudio`. Install the dependency to use the transport.
Single-file examples that introduce core Pipecat concepts one at a time. These examples:
2. Copy the [`env.example`](../env.example) file and add API keys for services you plan to use:
- Build on each other progressively
- Focus on specific features or integrations
- Are used for testing with every Pipecat release
```bash
cp env.example .env
# Edit .env with your API keys
```
See the **[Foundational Examples README](foundational/)** for the complete list.
3. Run any example:
## More Advanced Examples
```bash
uv run python getting-started/01-say-one-thing.py
```
Ready to explore complex use cases? Visit **[pipecat-examples](https://github.com/pipecat-ai/pipecat-examples)** for:
4. Open the web interface at http://localhost:7860/client/ and click "Connect"
- Production-ready applications
- Multi-platform client implementations
- Telephony integrations
- Multimodal and creative applications
- Deployment and monitoring examples
## Running examples with other transports
Most examples support running with other transports, like Twilio or Daily.
### Daily
You need to create a Daily account at https://dashboard.daily.co/u/signup. Once signed up, you can create your own room from the dashboard and set the environment variables `DAILY_ROOM_URL` and `DAILY_API_KEY`. Alternatively, you can let the example create a room for you (still needs `DAILY_API_KEY` environment variable). Then, start any example with `-t daily`:
```bash
uv run getting-started/06-voice-agent.py -t daily
```
### Twilio
It is also possible to run the example through a Twilio phone number. You will need to setup a few things:
1. Install and run [ngrok](https://ngrok.com/download).
```bash
ngrok http 7860
```
2. Configure your Twilio phone number. One way is to setup a TwiML app and set the request URL to the ngrok URL from step (1). Then, set your phone number to use the new TwiML app.
Then, run the example with:
```bash
uv run getting-started/06-voice-agent.py -t twilio -x NGROK_HOST_NAME
```
## Directory Structure
### [`getting-started/`](./getting-started/)
Progressive introduction to Pipecat, from minimal TTS to a full voice agent with function calling.
### [`voice/`](./voice/)
Full STT + LLM + TTS voice agent pipelines showcasing different speech service providers (Deepgram, ElevenLabs, Cartesia, etc.)
### [`function-calling/`](./function-calling/)
Function calling with different LLM providers (OpenAI, Anthropic, Google, etc.)
### [`transcription/`](./transcription/)
Speech-to-text examples with various STT providers.
### [`vision/`](./vision/)
Image description and vision capabilities with different multimodal LLMs.
### [`realtime/`](./realtime/)
Realtime and multimodal live APIs (OpenAI Realtime, Gemini Live, AWS Nova Sonic, Ultravox, Grok).
### [`persistent-context/`](./persistent-context/)
Maintaining conversation context across sessions with different providers.
### [`context-summarization/`](./context-summarization/)
Summarizing conversation context to manage token limits.
### [`update-settings/`](./update-settings/)
Changing service settings at runtime, organized by service type:
- **[`stt/`](./update-settings/stt/)** — Speech-to-text settings
- **[`tts/`](./update-settings/tts/)** — Text-to-speech settings
- **[`llm/`](./update-settings/llm/)** — LLM settings
### [`turn-management/`](./turn-management/)
Turn detection, interruption handling, and user input management.
### [`thinking-and-mcp/`](./thinking-and-mcp/)
LLM thinking/reasoning modes and MCP (Model Context Protocol) tool server integration.
### [`transports/`](./transports/)
Transport layer examples (WebRTC, Daily, LiveKit).
### [`video-avatar/`](./video-avatar/)
Video avatar integrations (Tavus, HeyGen, Simli, LemonSlice).
### [`video-processing/`](./video-processing/)
Video processing, mirroring, GStreamer, and custom video tracks.
### [`audio/`](./audio/)
Audio recording, background sounds, and sound effects.
### [`observability/`](./observability/)
Pipeline monitoring: observers, heartbeats, and Sentry metrics.
### [`rag/`](./rag/)
Retrieval-augmented generation, grounding, and long-term memory (Mem0, Gemini).
### [`features/`](./features/)
Miscellaneous features: wake phrases, live translation, service switching, voice switching, and more.
## Advanced Usage
### Customizing Network Settings
```bash
uv run python <example-name> --host 0.0.0.0 --port 8080
```
### Troubleshooting
- **No audio/video**: Check browser permissions for microphone and camera
- **Connection errors**: Verify API keys in `.env` file
- **Port conflicts**: Use `--port` to change the port
For more examples, visit the [pipecat-examples repository](https://github.com/pipecat-ai/pipecat-examples).

View File

Before

Width:  |  Height:  |  Size: 63 KiB

After

Width:  |  Height:  |  Size: 63 KiB

View File

Before

Width:  |  Height:  |  Size: 1.1 MiB

After

Width:  |  Height:  |  Size: 1.1 MiB

View File

Before

Width:  |  Height:  |  Size: 871 KiB

After

Width:  |  Height:  |  Size: 871 KiB

View File

Before

Width:  |  Height:  |  Size: 868 KiB

After

Width:  |  Height:  |  Size: 868 KiB

View File

Before

Width:  |  Height:  |  Size: 868 KiB

After

Width:  |  Height:  |  Size: 868 KiB

View File

Before

Width:  |  Height:  |  Size: 870 KiB

After

Width:  |  Height:  |  Size: 870 KiB

View File

Before

Width:  |  Height:  |  Size: 871 KiB

After

Width:  |  Height:  |  Size: 871 KiB

View File

Before

Width:  |  Height:  |  Size: 871 KiB

After

Width:  |  Height:  |  Size: 871 KiB

View File

Before

Width:  |  Height:  |  Size: 872 KiB

After

Width:  |  Height:  |  Size: 872 KiB

View File

Before

Width:  |  Height:  |  Size: 868 KiB

After

Width:  |  Height:  |  Size: 868 KiB

View File

Before

Width:  |  Height:  |  Size: 33 KiB

After

Width:  |  Height:  |  Size: 33 KiB

View File

Before

Width:  |  Height:  |  Size: 30 KiB

After

Width:  |  Height:  |  Size: 30 KiB

View File

@@ -128,7 +128,9 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
logger.info(f"Re-enabling background sound and starting bot...")
await task.queue_frame(MixerEnableFrame(True))
# Kick off the conversation.
context.add_message({"role": "user", "content": "Please introduce yourself to the user."})
context.add_message(
{"role": "developer", "content": "Please introduce yourself to the user."}
)
await task.queue_frames([LLMRunFrame()])
@transport.event_handler("on_client_disconnected")

View File

@@ -211,7 +211,9 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
async def on_client_connected(transport, client):
logger.info("Client connected")
# Kick off the conversation.
context.add_message({"role": "user", "content": "Please introduce yourself to the user."})
context.add_message(
{"role": "developer", "content": "Please introduce yourself to the user."}
)
await task.queue_frames([LLMRunFrame()])
@transport.event_handler("on_client_disconnected")

View File

@@ -172,7 +172,9 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
async def on_client_connected(transport, client):
logger.info("Client connected")
# Kick off the conversation.
context.add_message({"role": "user", "content": "Please introduce yourself to the user."})
context.add_message(
{"role": "developer", "content": "Please introduce yourself to the user."}
)
await task.queue_frames([LLMRunFrame()])
@transport.event_handler("on_client_disconnected")

View File

@@ -146,7 +146,9 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
async def on_client_connected(transport, client):
logger.info("Client connected")
# Kick off the conversation.
context.add_message({"role": "user", "content": "Please introduce yourself to the user."})
context.add_message(
{"role": "developer", "content": "Please introduce yourself to the user."}
)
await task.queue_frames([LLMRunFrame()])
@transport.event_handler("on_client_disconnected")

View File

@@ -172,7 +172,9 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
async def on_client_connected(transport, client):
logger.info("Client connected")
# Kick off the conversation.
context.add_message({"role": "user", "content": "Please introduce yourself to the user."})
context.add_message(
{"role": "developer", "content": "Please introduce yourself to the user."}
)
await task.queue_frames([LLMRunFrame()])
@transport.event_handler("on_client_disconnected")

View File

@@ -120,7 +120,9 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
async def on_client_connected(transport, client):
logger.info(f"Client connected")
# Kick off the conversation.
context.add_message({"role": "user", "content": "Please introduce yourself to the user."})
context.add_message(
{"role": "developer", "content": "Please introduce yourself to the user."}
)
# Custom frames are pushed in order so they can be used for synchronization purposes.
await task.queue_frames([CustomBeforeProcessFrame(), LLMRunFrame(), CustomAfterPushFrame()])

View File

@@ -77,7 +77,6 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
groq_llm = GroqLLMService(
api_key=os.getenv("GROQ_API_KEY"),
settings=GroqLLMService.Settings(
model="meta-llama/llama-4-maverick-17b-128e-instruct",
system_instruction="You are a very helpful assistant. Your goal is to demonstrate your capabilities in detail in a creative and helpful way.",
),
)
@@ -145,10 +144,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
logger.info(f"Client connected")
# Kick off the conversation.
openai_context.add_message(
{"role": "user", "content": "Please introduce yourself to the user."}
{"role": "developer", "content": "Please introduce yourself to the user."}
)
groq_context.add_message(
{"role": "user", "content": "Please introduce yourself to the user."}
{"role": "developer", "content": "Please introduce yourself to the user."}
)
await task.queue_frames([LLMRunFrame()])

View File

@@ -155,10 +155,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
async def on_client_connected(transport, client):
logger.info("Client connected")
main_context.add_message(
{"role": "user", "content": "Please introduce yourself to the user."}
{"role": "developer", "content": "Please introduce yourself to the user."}
)
evaluator_context.add_message(
{"role": "user", "content": "Ready to evaluate user messages."}
{"role": "developer", "content": "Ready to evaluate user messages."}
)
await task.queue_frames([LLMRunFrame()])

View File

@@ -141,7 +141,9 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
async def on_client_connected(transport, client):
logger.info(f"Client connected: {client}")
# Kick off the conversation.
context.add_message({"role": "user", "content": "Please introduce yourself to the user."})
context.add_message(
{"role": "developer", "content": "Please introduce yourself to the user."}
)
await task.queue_frames([LLMRunFrame()])
@transport.event_handler("on_client_disconnected")

View File

@@ -109,7 +109,9 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
async def on_client_connected(transport, client):
logger.info(f"Client connected")
# Kick off the conversation.
context.add_message({"role": "user", "content": "Please introduce yourself to the user."})
context.add_message(
{"role": "developer", "content": "Please introduce yourself to the user."}
)
await task.queue_frames([LLMRunFrame()])
# Handle "latency-ping" messages. The client will send app messages that look like

View File

@@ -164,7 +164,9 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
async def on_client_connected(transport, client):
logger.info(f"Client connected")
# Kick off the conversation.
context.add_message({"role": "user", "content": "Please introduce yourself to the user."})
context.add_message(
{"role": "developer", "content": "Please introduce yourself to the user."}
)
await task.queue_frames([LLMRunFrame()])
await asyncio.sleep(15)
print(f"Switching to {stt_deepgram}")

View File

@@ -162,7 +162,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
# Kick off the conversation.
context.add_message(
{
"role": "user",
"role": "developer",
"content": f"Please introduce yourself to the user and let them know the languages you speak. Your initial responses should be in {tts.current_language}.",
}
)

View File

@@ -172,7 +172,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
# Kick off the conversation.
context.add_message(
{
"role": "user",
"role": "developer",
"content": f"Please introduce yourself to the user and let them know the voices you can do. Your initial responses should be as if you were a {tts.current_voice}.",
}
)

View File

@@ -120,7 +120,9 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
async def on_client_connected(transport, client):
logger.info(f"Client connected")
# Kick off the conversation.
context.add_message({"role": "user", "content": "Please introduce yourself to the user."})
context.add_message(
{"role": "developer", "content": "Please introduce yourself to the user."}
)
await task.queue_frames([LLMRunFrame()])
@transport.event_handler("on_client_disconnected")

Some files were not shown because too many files have changed in this diff Show More