fix: convert_messages drops tool_call_id field (and other fields) #1164

[chiradeepv]

Please use conventional commits to describe your changes. For example, feat: add new feature or fix: fix a bug. If you are unsure, leave the title as ... and AI will handle it.

Describe your changes

As described in #1164, when the message is of the form {"role": "tool"...}, convert_messages drops essential fields.
...

Issue ticket number and link

#1164

Checklist before requesting a review

• [x ] I have performed a self-review of my code
• If it is a core feature, I have added thorough tests.
• If it is a core feature, I have added documentation.

---

Important

Fixes convert_messages in multimodal.py to retain all fields in message dictionaries when content is a string.

• Behavior:
  • Fixes convert_messages in multimodal.py to retain all fields in message dictionaries when content is a string.
• Misc:
  • Minor comment and formatting adjustments in multimodal.py.

^{This description was created by for 23f938c. It will automatically update as commits are pushed.}

[ellipsis-dev]

👍 Looks good to me! Reviewed everything up to 23f938c in 9 seconds

More details

• Looked at 13 lines of code in 1 files
• Skipped 0 files when reviewing.
• Skipped posting 2 drafted comments based on config settings.

1. instructor/multimodal.py:359

• Draft comment:
  The change from {"role": role, "content": content} to message ensures that all fields in the message, such as 'tool_call_id', are preserved. This is consistent with the PR's intent to prevent dropping essential fields.
• Reason this comment was not posted:
  Confidence changes required: 0%
  The change in line 359 is intended to preserve the entire message object, including any additional fields like 'tool_call_id'. This is a valid change as it aligns with the PR description and issue convert_messages throws away required data (openai tool calling) #1164.

2. instructor/multimodal.py:359

• Draft comment:
  Ensure that the message object is consistently transformed before appending to converted_messages. This will maintain a consistent data structure.
• Reason this comment was not posted:
  Confidence changes required: 80%
  The function convert_messages has a line where it appends the entire message object to converted_messages without any transformation. This is inconsistent with the rest of the function where the content is transformed before appending. This could lead to inconsistent data structures in converted_messages.

Workflow ID: wflow_RGZ9KxdEUEX0frlu

---

You can customize Ellipsis with 👍 / 👎 feedback, review rules, user-specific overrides, quiet mode, and more.

[chris-sanders]

+1 I was trying to do this too and hit this same issue. As far as I can tell you can't use tool_calling with OpenAI until this is landed, can we get a release with this?

I've tested this edit and it resolves the issue for me.

Edit: nevermind this didn't fully handle it I'm going to open a PR with a few more changes.

[chris-sanders]

Ok here's the solution that works now for me with 1.6.4 as a base: #1194