wx44wx/pipecat
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Improve TTSService handling of long LLM token outputs

Browse Source
This commit is contained in:
Mark Backman
2025-11-24 18:59:26 -05:00
parent b52b108932
commit 4699dc2345
3 changed files with 67 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
CHANGELOG.md
View File

@@ -88,6 +88,13 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
### Changed
- Improved `TTSService` to properly handle buffered sentences at the end of LLM
responses. Previously, when an LLM response ended, any complete sentences
remaining in the aggregator's buffer would be sent to TTS as one large
chunk. Now `TTSService` continues processing aggregation by repeatedly calling
`aggregate("")` until all buffered text has been processed, ensuring each
sentence is sent to TTS individually for better interruption points.
- Updated `daily-python` to 0.22.0.
- `BaseTextAggregator` changes: