Skip to content

AWS: ignore chain of thought for agent speech #1603

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
jayeshp19 wants to merge 1 commit into
base: 0.x
Choose a base branch
from
Open

AWS: ignore chain of thought for agent speech #1603

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 5 additions & 0 deletions .changeset/nervous-vans-help.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,5 @@
---
"livekit-plugins-aws": patch
---

ignore chain of thought for agent speech
5 changes: 4 additions & 1 deletion livekit-plugins/livekit-plugins-aws/livekit/plugins/aws/llm.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -16,6 +16,7 @@

import asyncio
import os
import re
from dataclasses import dataclass
from typing import Any, Literal, MutableSet, Union

Expand Down Expand Up @@ -266,11 +267,13 @@ def _parse_chunk(self, request_id: str, chunk: dict) -> llm.ChatChunk | None:
self._text += delta["text"]
elif "contentBlockStop" in chunk:
if self._text:
# ignore chain of thought
text = re.sub(r"<thinking>.*?</thinking>", "", self._text)
Copy link
Member

@davidzhao davidzhao Mar 5, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

should this be AWS specific? or a general thing we should do in the LLM superclass? maybe an option: remove_thinking?

also is it guaranteed the chain of thought appears in the same chunk?

Copy link
Member

@theomonnom theomonnom Mar 5, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think we should keep the thinking tokens (in v1.0 at least).

It should be the responsability of the tts_node to remove those tokens.

Tho not against merging this PR for main as a temporary solution

Copy link
Member

@davidzhao davidzhao Mar 5, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

we need a util to strip tags in a streaming fashion

chat_chunk = llm.ChatChunk(
request_id=request_id,
choices=[
llm.Choice(
delta=llm.ChoiceDelta(content=self._text, role="assistant"),
delta=llm.ChoiceDelta(content=text, role="assistant"),
index=chunk["contentBlockStop"]["contentBlockIndex"],
)
],
Expand Down
Loading