Edits to batch.py to support Anthropic batch API

[economy]

I noticed the structure of the JSONL file produced by create_from_messages was incorrect for the Anthropic API. This change allows submission of batch jobs to Anthropic by creating the correct JSONL structure

---

Important

create_from_messages in batch.py now supports Anthropic batch API by correctly formatting JSONL files.

• Behavior:
  • create_from_messages in batch.py now supports Anthropic batch API by formatting JSONL files correctly.
  • Uses a conditional check for "claude" in the model name to determine if Anthropic format is needed.
• Implementation:
  • For Anthropic, constructs a JSON object with custom_id, params including model, max_tokens, temperature, and messages.
  • Writes formatted JSONL to file for Anthropic and OpenAI formats.
• Misc:
  • Minor restructuring of file writing logic in create_from_messages.

^{This description was created by for b69522c. It will automatically update as commits are pushed.}

[ellipsis-dev]

👍 Looks good to me! Reviewed everything up to b69522c in 26 seconds

More details

• Looked at 53 lines of code in 1 files
• Skipped 0 files when reviewing.
• Skipped posting 4 drafted comments based on config settings.

1. instructor/batch.py:126

• Draft comment:
  The logic use_anthropic = "claude" in model.lower() is fragile. Consider using a more robust method to determine the API format, such as a configuration or a mapping of model names to formats.
• Reason this comment was not posted:
  Confidence changes required: 50%
  The code for handling the Anthropic format in create_from_messages is correct, but the logic for determining the format based on the model name is fragile. It assumes that any model name containing 'claude' should use the Anthropic format, which might not be future-proof or accurate for all cases.

2. instructor/batch.py:157

• Draft comment:
  The logic use_anthropic = "claude" in model.lower() is fragile. Consider using a more robust method to determine the API format, such as a configuration or a mapping of model names to formats.
• Reason this comment was not posted:
  Confidence changes required: 50%
  The code for handling the Anthropic format in create_from_messages is correct, but the logic for determining the format based on the model name is fragile. It assumes that any model name containing 'claude' should use the Anthropic format, which might not be future-proof or accurate for all cases.

3. instructor/batch.py:129

• Draft comment:
  The logic for writing messages to a file is repeated for both Anthropic and OpenAI formats. Consider refactoring to eliminate redundancy and adhere to the DRY principle.
• Reason this comment was not posted:
  Confidence changes required: 80%
  The code has repeated logic for handling the writing of messages to a file, which violates the DRY principle.

4. instructor/batch.py:131

• Draft comment:
  The JSON parsing logic here is complex and could benefit from comments explaining the structure being parsed. This applies to similar logic on line 149 as well.
• Reason this comment was not posted:
  Confidence changes required: 70%
  The code has multiple instances of complex JSON parsing logic that could benefit from comments for clarity.

Workflow ID: wflow_8B7rKNEOkfw2TTg7

---

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