anthropic force tool

[Cruppelt]

🚀 This description was created by Ellipsis for commit 5a3bff4

Summary:

This PR introduces a new tool choice in the anthropic tools mode and updates the anthropic library version to 0.26.0.

Key points:

• Updated handle_response_model in /instructor/process_response.py to add a new tool choice for anthropic tools mode.
• Updated anthropic library version from 0.23.1 to 0.26.0 in pyproject.toml.
• Ensured consistency of anthropic version across different sections in pyproject.toml.

---

Generated with ❤️ by ellipsis.dev

[sweep-ai]

Sweep: PR Review

instructor/process_response.py

A new line was added to set the tool_choice key in the new_kwargs dictionary when the mode is ANTHROPIC_TOOLS.

Potential Issues

Sweep isn't 100% sure if the following are issues or not but they may be worth taking a look at.

• The new line assumes that response_model has an attribute __name__, which may not be true for all possible response_model types, potentially causing an AttributeError.

instructor/instructor/process_response.py

Line 283 in 5a3bff4

new_kwargs["tool_choice"] = {"type": "tool", "name": response_model.__name__}


View Diff

---

The following files were not reviewed because our filter identified them as typically non-human-readable or less important files (e.g., dist files, package.json, images). If this is an error, please let us know.

• poetry.lock
• pyproject.toml

[ellipsis-dev]

👍 Looks good to me! Reviewed everything up to 5a3bff4 in 1 minute and 11 seconds

More details

• Looked at 45 lines of code in 2 files
• Skipped 1 files when reviewing.
• Skipped posting 1 drafted comments based on config settings.

1. instructor/process_response.py:283

• Draft comment:
  Using response_model.__name__ directly can lead to an AttributeError if response_model is not a class or does not have a __name__ attribute. Consider checking if response_model has the attribute __name__ before accessing it, or handle this potential issue gracefully.

            new_kwargs['tool_choice'] = {'type': 'tool', 'name': getattr(response_model, '__name__', str(response_model))}

• Reason this comment was not posted:
  Confidence of 30% on close inspection, compared to threshold of 85%.

Workflow ID: wflow_JRyAz2b7oR3lK4Vk

---

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

[Cruppelt]

Local testing seems fine. Followed the guide from https://github.com/anthropics/anthropic-cookbook/blob/main/tool_use/tool_choice.ipynb to force a too.