Skip to content

common : fix Hermes/Qwen tool-call parsing to stop wrapper leaks #16378

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

Jump to bottom
Closed
wants to merge 3 commits into from
Closed

common : fix Hermes/Qwen tool-call parsing to stop wrapper leaks #16378

wants to merge 3 commits into from

Conversation

@laundrevity
Copy link

@laundrevity laundrevity commented Oct 2, 2025 •
edited
Loading

Fixes #15012

(in the sense that XML no longer shows up in content, and qwen-code works with Qwen3-Coder-30B-A3B-Instruct-UD_Q4_K_XL)

  • teach the Hermes/Qwen parser to strip <tool_call>/<function...> wrappers when the model nests a function block inside them
  • add a helper that removes wrapper-only suffix deltas so streaming outputs no longer emit XML fragments for tool calls
  • cover newline + nested <tool_call><function...> cases in the chat parser tests

@laundrevity laundrevity requested a review from ggerganov as a code owner October 2, 2025 01:13
@github-actions github-actions bot added the testing Everything test related label Oct 2, 2025
@laundrevity laundrevity mentioned this pull request Oct 2, 2025
Open
4 tasks
@laundrevity laundrevity closed this Oct 2, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@ggerganov ggerganov Awaiting requested review from ggerganov ggerganov is a code owner

Assignees

No one assigned

Labels

testing Everything test related

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

Feature Request: Qwen3-Coder Tool Call Parser

1 participant

@laundrevity