"<break" tag leaks into conversation message when used in .say

[marioplumbarius]

Bug Description

When I add <break> tag to the text argument from .say.
The agent says the message without mentioning the <break> tag.
But the conversation message contains the <break> tag.

Expected Behavior

When agent uses session.say with a text that includes <break>, the <break> should not be present in the conversation message.

Reproduction Steps

Steps to reproduce the behavior:
1. one client is connected to the room (in addition to agent)
2. clients and agent exchange messages
3. when agent uses `session.say` with a text that includes `<break ...`, the `<break ...` tag leaks into the message

Operating System

MacOS, Linux

Models Used

Azure-Speech, Cartesia, OpenAI

Package Versions

**Server**
- Version: [0.x.x] (unsure what this is)
- Environment: EKS
- Dependencies: azure-speech, cartesia, gpt-4.1

**Client**
- SDK: python
- Version:
  - livekit = "1.0.20"
  - livekit-agents = {version = "1.3.6", extras = ["turn-detector", "silero", "cartesia", "azure"]}
  - livekit-api = "1.1.0"
  - livekit-plugins-noise-cancellation = "0.2.5"
  - livekit-plugins-openai = "1.3.6"
  - livekit-protocol = "1.1.1"

Session/Room/Call IDs

No response

Proposed Solution

Additional Context

No response

Screenshots and Recordings

No response

[longcw]

you can cleanup the tags in a custom transcription_node https://docs.livekit.io/agents/logic/nodes/#transcription-node