async gemini support

[Cruppelt]

🚀	This description was created by Ellipsis for commit b1602cb

Summary:

This PR introduces asynchronous support to the from_gemini function in the instructor/client_gemini.py, allowing it to handle both synchronous and asynchronous operations.

Key points:

• Added async support in from_gemini function in instructor/client_gemini.py
• Overloads added for from_gemini to handle sync and async modes
• Conditional logic based on use_async to return appropriate instructor type
• Uses generate_content_async for async and generate_content for sync

---

Generated with ❤️ by ellipsis.dev

[ellipsis-dev]

👍 Looks good to me! Reviewed everything up to b1602cb in 1 minute and 44 seconds

More details

• Looked at 78 lines of code in 1 files
• Skipped 0 files when reviewing.
• Skipped posting 1 drafted comments based on config settings.

1. instructor/client_gemini.py:31

• Draft comment:
  The function no longer enforces that the mode must be instructor.Mode.GEMINI_JSON as it did in the previous version. If this is intentional, ensure that all calling code is aware of this change. If not, consider adding an assertion to enforce this.

    assert mode == instructor.Mode.GEMINI_JSON, "Mode must be instructor.Mode.GEMINI_JSON"

This suggestion applies to the entire function definition.

• Reason this comment was not posted:
  Confidence of 0% on close inspection, compared to threshold of 85%.

Workflow ID: wflow_Q3xAfLc2GXsIwv1e

---

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

[Cruppelt]

lmk thoughts on the use_async - The other clients have different Async clients that we can check the instance of but in this case Google/Gemini has the same client.

[jxnl]

Honestly, this looks good. Have you tested it locally yet?

[Cruppelt]

Honestly, this looks good. Have you tested it locally yet?

Yes, typehead working too.

[ssonal]

retry.py retry_async logic needs to also be updated to handle ValidationErrors. Gemini uses a different messages format.

[Cruppelt]

retry.py retry_async logic needs to also be updated to handle ValidationErrors. Gemini uses a different messages format.

Have an example that will break? Happy to add

[ssonal]

retry.py retry_async logic needs to also be updated to handle ValidationErrors. Gemini uses a different messages format.

Have an example that will break? Happy to add

Just saying that

instructor/instructor/retry.py

Lines 159 to 169 in 3d8d449

	except (ValidationError, JSONDecodeError) as e:
	logger.debug(f"Error response: {response}")
	if mode in {Mode.GEMINI_JSON}:
	kwargs["contents"].extend(reask_messages(response, mode, e))
	else:
	kwargs["messages"].extend(reask_messages(response, mode, e))
	if mode in {Mode.ANTHROPIC_TOOLS, Mode.ANTHROPIC_JSON}:
	kwargs["messages"] = merge_consecutive_messages(
	kwargs["messages"]
	)
	raise e

and

instructor/instructor/retry.py

Lines 225 to 232 in 3d8d449

	except (ValidationError, JSONDecodeError) as e:
	logger.debug(f"Error response: {response}", e)
	kwargs["messages"].extend(reask_messages(response, mode, e))
	if mode in {Mode.ANTHROPIC_TOOLS, Mode.ANTHROPIC_JSON}:
	kwargs["messages"] = merge_consecutive_messages(
	kwargs["messages"]
	)
	raise e

should match.

GEMINI_JSON mode requires different handling of args for retries.