Skip to content

Conversation

@dkalinowski
Copy link
Collaborator

🛠 Summary

Changes which are not following OpenAI harmony format, but workaround model accuracy issues

  • tool call appears in wrong channel (analysis), not only in commentary as supposed to
  • TBD

🧪 Checklist

  • Unit tests added.

ToolCalls_t toolCalls;
for (const auto& msg : messages) {
if (startsWith(msg.getChannel(), "commentary")) {
if (startsWith(msg.getChannel(), "commentary") || startsWith(msg.getChannel(), "analysis")) {
Copy link
Collaborator

Choose a reason for hiding this comment

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

Just a comment. For GenAI/OV testing purposes it could be useful to be able to turn off such workarounds, to check accuracy fixes, otherwise we may just mask those issues

@dkalinowski dkalinowski changed the title GPT-OSS parser hacks [do not merge] GPT-OSS parser hacks Nov 18, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants