Commit Graph

4386 Commits

Author SHA1 Message Date
Aleix Conchillo Flaqué
ca7b0650c2 examples: capture camera or screen. allow setting framerate 2025-05-29 13:16:44 -07:00
Aleix Conchillo Flaqué
67dd146038 output transports cleanup 2025-05-29 13:16:44 -07:00
Aleix Conchillo Flaqué
fb66df2efd use uvloop as the new event loop on Linux and macOS 2025-05-29 11:24:21 -07:00
Aleix Conchillo Flaqué
d203789490 add STT result field to TranscriptionFrame/InterimTranscriptionFrame 2025-05-29 11:21:44 -07:00
Aleix Conchillo Flaqué
7ea0e31cd4 Merge pull request #1924 from pipecat-ai/aleix/new-examples-package
add new pipecat.examples package and make runner public
2025-05-29 11:18:22 -07:00
Mark Backman
803b3f2cc4 fix: update websocket_client to use FrameProcessorSetup 2025-05-29 14:02:47 -04:00
Aleix Conchillo Flaqué
1788ba6c5c add new pipecat.examples package and make runner public 2025-05-29 10:43:12 -07:00
Philipp Eisen
5209bd3d9f Fix implemetation of ttfb_ms metric 2025-05-29 19:24:04 +02:00
Aleix Conchillo Flaqué
5676920a6a BaseOutputTransport/DailyTransport: allow sending DTMF keypress 2025-05-29 09:33:29 -07:00
Aleix Conchillo Flaqué
513221d9fd added OutputDTMFUrgentFrame 2025-05-29 09:32:57 -07:00
Mark Backman
bee242b781 Safer check when using the no_operation tool 2025-05-29 10:25:20 -04:00
Mark Backman
fa1c98ff29 Add no_op tool to AWSBedrockLLMService 2025-05-29 10:25:19 -04:00
Matej Marinko
ee5fea4221 Fix auto finalization cycle 2025-05-29 14:58:35 +02:00
Matej Marinko
db7b60cfe9 Auto finalize fix 2025-05-29 13:24:53 +02:00
alexflorensa
0c2efb312c fix(deepgram-stt) Set model name to Deepgram STT 2025-05-29 12:08:16 +02:00
Matej Marinko
51b79bd6a1 Minor code style changes 2025-05-29 10:11:11 +02:00
Matej Marinko
95fe762776 Fix typo 2025-05-29 09:23:37 +02:00
Aleix Conchillo Flaqué
05ae3a1703 DailyTransport: expose start_transcription/stop_transcription 2025-05-28 11:21:26 -07:00
Matej Marinko
2968c846ce Add Soniox STT service 2025-05-28 09:35:21 +02:00
Filipi Fuchter
bf31bce440 Updated SmallWebRTCTransport to align with how other transports handle on_client_disconnected 2025-05-27 17:42:52 -07:00
Aleix Conchillo Flaqué
071a9307c9 transport(websocket): do not require a frame serializer 2025-05-27 17:42:52 -07:00
Aleix Conchillo Flaqué
ecf878e14d DailyTransport: allow requesting video frames with any framerate 2025-05-27 17:42:50 -07:00
Aleix Conchillo Flaqué
4eed335bc7 PipelineTask: check if pipeline has already been cancelled 2025-05-27 17:40:39 -07:00
Aleix Conchillo Flaqué
2e57bb74d2 BaseObject: do not raise exception if event handler not registered 2025-05-27 17:40:35 -07:00
Aleix Conchillo Flaqué
0a39769cd0 Merge pull request #1901 from pipecat-ai/aleix/deepgram-sdk-4.1.0
pyproject: update deepgram-sdk to 4.1.0
2025-05-27 17:20:54 -07:00
Aleix Conchillo Flaqué
0099f60d29 deepgram: fix an issue with user provided LiveOptions 2025-05-27 15:19:17 -07:00
Filipi da Silva Fuchter
eaf9f20c56 Merge pull request #1898 from pipecat-ai/tavus_transport_fixes
Fixing TavusTransport with some TTS services.
2025-05-27 16:33:19 -03:00
Aleix Conchillo Flaqué
6242278abd Merge pull request #1895 from pipecat-ai/aleix/more-avoid-mutable-default-values
more avoiding mutable default constructor values
2025-05-27 10:08:42 -07:00
Mark Backman
30850a431a Merge pull request #1886 from pipecat-ai/mb/add-otel-llm-output
Add LLM response tracing to OTel tracing
2025-05-27 11:58:57 -04:00
Mark Backman
ebb3d1cfd3 fix: Update GoogleLLMContext to_standard_messages to be compatible with google-genai 2025-05-27 11:20:21 -04:00
Filipi Fuchter
acce9489d7 Creating the silence based on the chunk size. 2025-05-27 11:26:34 -03:00
Filipi Fuchter
3d442620f9 Removing not used imports. 2025-05-27 10:42:06 -03:00
Martin Schweiger
f1d7eb8565 final touches 2025-05-27 21:29:50 +08:00
Filipi Fuchter
798b935ff6 The remaining audio should not be sent as done. 2025-05-27 10:28:40 -03:00
Filipi Fuchter
3039a1444e Refactoring the TavusVideoService to match the same the behavior of the bot started speaking and bot stopped speaking. 2025-05-27 10:26:41 -03:00
Mark Backman
aa7d15beb3 fix: move LLM call outside tracing try block to prevent double execution 2025-05-26 22:06:31 -04:00
Filipi Fuchter
2b3d2cb342 Fixing the issue when using the TavusVideoService with some TTS services. 2025-05-26 22:55:20 -03:00
Filipi Fuchter
5a58357429 Fixing the issue when using the TavusTransport with some TTS services. 2025-05-26 18:34:52 -03:00
Mark Backman
e13c9fd42e Add PlivoFrameSerializer 2025-05-26 11:00:01 -04:00
Aleix Conchillo Flaqué
7a767e680c more avoiding mutable default constructor values 2025-05-25 21:00:35 -07:00
Martin Schweiger
320b52eb1e Merge branch 'm-ods/assemblyai-universal-streaming' of https://github.com/m-ods/pipecat into m-ods/assemblyai-universal-streaming 2025-05-26 09:13:08 +08:00
Martin Schweiger
428cee75c5 Add User-Agent header to AssemblyAI websocket connection 2025-05-26 09:10:55 +08:00
Martin Schweiger
5479a55b2c Add websockets dependency to assemblyai extra 2025-05-26 09:08:56 +08:00
aristid
09ba319f3e Merge branch 'main' into google-streaming-tts 2025-05-24 17:16:22 +02:00
ezun-kim
3da711ba8b Fix SSE server connection handling for MCP client
### Summary
This PR improves the MCP (Model Context Protocol) client's SSE (Server-Sent Events) server connection handling by replacing the generic string parameter with a proper `SseServerParameters` class.

### Changes
- **Breaking Change**: Changed `server_params` type from `Union[StdioServerParameters, str]` to `Union[StdioServerParameters, SseServerParameters]`
- Added import for `SseServerParameters` from `mcp.client.session_group`
- Updated SSE client connection to use structured parameters instead of a simple URL string
- Fixed error message to correctly reflect the expected parameter types
- Improved logging by changing info-level log to debug-level for consistency

### Details

#### Before
The SSE client connection only accepted a URL string:
```python
async with self._client(self._server_params) as (read, write):
```

#### After
Now properly unpacks SSE server parameters:
```python
async with self._client(
    url=self._server_params.url,
    headers=self._server_params.headers,
    timeout=self._server_params.timeout,
    sse_read_timeout=self._server_params.sse_read_timeout
) as (read, write):
```

### Benefits
- **Type Safety**: Stronger type checking with dedicated `SseServerParameters` class
- **Extended Configuration**: Support for custom headers (authentication), timeouts, and SSE-specific settings
- **Better Error Messages**: Clear type error messages when incorrect parameters are provided
- **Improved Debugging**: Debug logging of SSE server parameters for troubleshooting

### Migration Guide
Users need to update their SSE server initialization:
```python
# Before
client = MCPClient("https://example.com/sse")

# After
from mcp.client.session_group import SseServerParameters
client = MCPClient(SseServerParameters(
    url="https://example.com/sse",
    headers={"Authorization": "Bearer token"},
    timeout=30,
    sse_read_timeout=60
))
```

### Testing
- [ ] Tested with StdioServerParameters (unchanged behavior)
- [ ] Tested with SseServerParameters with various configurations
- [ ] Verified error handling for invalid parameter types

---

This is a necessary change to support production-ready SSE connections with proper authentication and timeout handling.
2025-05-24 22:35:57 +09:00
Mark Backman
6f524fb816 Add LLM response to OTel tracing 2025-05-24 09:15:39 -04:00
unknown
d3e2a9e5c0 Change default voice and fix formatting 2025-05-24 15:14:39 +02:00
Aleix Conchillo Flaqué
940926b5ec TavusVideoService: no need to enable audio/video outputs 2025-05-23 19:29:34 -07:00
Aleix Conchillo Flaqué
85c096df0b DailyTransport: create audio/video input tasks when input flag is enabled 2025-05-23 19:28:18 -07:00
Filipi Fuchter
1ebca35313 Queuing the app messages if the SmallWebrtcTransport is not connected yet. 2025-05-23 23:03:04 -03:00