Fix handling of snapshot_ids ("gpt-4-turbo-2024-04-09" and "gpt-4o-2024-05-13") and alias "gpt-4-turbo".

[st4r0]

Currently snapshots "gpt-4o-2024-05-13" and "gpt-4-turbo-2024-04-09" wrongfully return model costs for "gpt-4".
The same is the case for "gpt-4-turbo", which is inconsistent with the OpenAI API that has "gpt-4-turbo" pointing to "gpt-4-turbo-2024-04-09" (https://platform.openai.com/docs/models/gpt-4-turbo-and-gpt-4).

These commits would fix these issue for now. However, if OpenAI decides to change pricing so that "gpt-4-turbo-2024-04-09" and "gpt-4-turbo-preview" (pointing to "gpt-4-0125-preview") will differ in cost, the logic will not hold. I suggest adding costs for each snapshot_id and adding tests to ensure that they are correct.

If you agree, I'd be more than happy to prepare a PR.

---

🚀 This description was created by Ellipsis for commit 665596a

Summary:

This PR fixes incorrect cost retrieval for specific GPT-4 model snapshots and aliases, ensuring consistency with OpenAI's API.

Key points:

• Updated get_model_cost in instructor/cli/usage.py to correctly handle costs for gpt-4-turbo-2024-04-09, gpt-4o-2024-05-13, and alias gpt-4-turbo.
• Suggested future improvements for dynamic cost handling based on potential changes in OpenAI's pricing.

---

Generated with ❤️ by ellipsis.dev

[sweep-ai]

Sweep: PR Review

Sweep has finished reviewing your pull request.

instructor/cli/usage.py

tag. Added support for recognizing "gpt-4-turbo" and "gpt-4o" model prefixes in the get_model_cost function.

No issues found with the reviewed changes

---

[ellipsis-dev]

👍 Looks good to me! Reviewed everything up to 665596a in 1 minute and 51 seconds

More details

• Looked at 18 lines of code in 1 files
• Skipped 0 files when reviewing.
• Skipped posting 1 drafted comments based on config settings.

1. instructor/cli/usage.py:89

• Draft comment:
  The alias 'gpt-4-turbo' should point to 'gpt-4-turbo-2024-04-09' as per the OpenAI API documentation, not 'gpt-4-turbo-preview'.

    elif model.startswith("gpt-4-turbo"):
        return MODEL_COSTS["gpt-4-turbo-2024-04-09"]

• Reason this comment was not posted:
  Confidence of 80% on close inspection, compared to threshold of 85%.

Workflow ID: wflow_exELsg3cNQf2vMKP

---

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

[st4r0]

Actually I just realized that there is no way to check the actual costs the way I had imagined, so writing tests for those might not work.

[st4r0]

Now the startswith logic could be removed...