From 03002ad685a201e40f3b91d14c4ea446b933d0bd Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Aleix=20Conchillo=20Flaqu=C3=A9?= Date: Fri, 25 Apr 2025 11:40:35 -0700 Subject: [PATCH] LLMUserContextAggregator: reduce aggregation_timeout to 0.5 --- CHANGELOG.md | 7 +++++++ src/pipecat/processors/aggregators/llm_response.py | 2 +- 2 files changed, 8 insertions(+), 1 deletion(-) diff --git a/CHANGELOG.md b/CHANGELOG.md index 92ee7d565..65b692e69 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -25,6 +25,13 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0 ### Changed +- Changed the user aggregator timeout for late transcriptions from 1.0s to 0.5s + (`LLMUserAggregatorParams.aggregation_timeout`). Sometimes, the STT services + might give us more than one transcription which could come after the user + stopped speaking. We still want to include these additional transcriptions + with the first one because it's part of the user turn. This is what this + timeout is helpful with. + - Short utterances not detected by VAD while the bot is speaking are now ignored. This reduces the amount of bot interruptions significantly providing a more natural conversation experience. diff --git a/src/pipecat/processors/aggregators/llm_response.py b/src/pipecat/processors/aggregators/llm_response.py index 2d9421634..fa163ac0a 100644 --- a/src/pipecat/processors/aggregators/llm_response.py +++ b/src/pipecat/processors/aggregators/llm_response.py @@ -50,7 +50,7 @@ from pipecat.utils.time import time_now_iso8601 @dataclass class LLMUserAggregatorParams: - aggregation_timeout: float = 1.0 + aggregation_timeout: float = 0.5 @dataclass