Feat/adding directory and file prompt loading in genkit #3962

MengqinShen · 2025-12-17T03:46:08Z

Description here... Help the reviewer by:

linking to an issue that includes more details
if it's a new feature include samples of how to use the new feature
(optional if issue link is provided) if you fixed a bug include basic bug details

Checklist (if applicable):

PR title is following https://www.conventionalcommits.org/en/v1.0.0/
Tested (manually, unit tested, etc.)
Docs updated (updated docs or a docs bug required)

…S SDK

…st based on code review

gemini-code-assist · 2025-12-17T03:46:36Z

Summary of Changes

Hello @MengqinShen, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request significantly expands Genkit's prompt management capabilities by introducing a robust system for loading prompts directly from file directories. This new feature supports structured prompt organization with variants and reusable partials, enhancing modularity and developer workflow. It also refines the internal prompt registration and lookup mechanisms, ensuring seamless integration of both programmatically defined and file-based prompts, while updating core dependencies for broader Python version compatibility.

Highlights

File-based Prompt Loading: Introduced the ability to load prompts from '.prompt' files within specified directories, supporting nested structures, variants (e.g., 'prompt.variant.prompt'), and partials (files starting with '_').
Enhanced Prompt Registry and Lookup: Implemented new functions ('load_prompt', 'load_prompt_folder', 'lookup_prompt') and modified the 'Genkit' class to allow programmatic and file-based prompts to be registered and retrieved efficiently, including lazy loading of prompt metadata.
Dual Action Registration for File Prompts: File-based prompts now register two distinct actions in the Genkit registry: a 'PROMPT' action that returns a 'GenerateRequest' (for rendering) and an 'EXECUTABLE_PROMPT' action that returns 'GenerateActionOptions' (for execution).
Dependency Updates: Updated 'uv.lock' to include 'opentelemetry-exporter-gcp-monitoring' and adjusted 'grpcio' versions for Python 3.14 compatibility, along with minor 'pyproject.toml' license additions.

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point by creating a comment using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

Feature	Command	Description
Code Review	`/gemini review`	Performs a code review for the current pull request in its current state.
Pull Request Summary	`/gemini summary`	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in pull request comments and review comments.
Help	`/gemini help`	Displays a list of available commands.

Customization

To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here.

You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.

Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

MengqinShen · 2025-12-17T03:48:18Z

new changes accidentally pushed to an old branch , close it

gemini-code-assist

Code Review

This pull request introduces a significant new feature for loading prompts from directories and files, including support for lazy loading. The implementation is quite extensive, particularly in py/packages/genkit/src/genkit/blocks/prompt.py. While the overall approach is sound, I've identified a critical bug in the handling of weak references that will cause runtime errors. Additionally, there are some inconsistencies in the caching logic that could lead to memory leaks. I've also found issues in the new tests, including one that appears to be incorrect and several others with weak assertions that could be improved. My review includes suggestions to fix these issues.

gemini-code-assist · 2025-12-17T03:49:50Z

py/packages/genkit/src/genkit/blocks/prompt.py

+        if hasattr(action, '_executable_prompt') and action._executable_prompt is not None:
+            return action._executable_prompt


The _executable_prompt attribute on the action is a weakref.ref object. You are returning the weak reference itself instead of the referenced ExecutablePrompt object. You need to call the weak reference to get the object it points to. You should also handle the case where the object has been garbage collected and the weak reference returns None.

Suggested change

if hasattr(action, '_executable_prompt') and action._executable_prompt is not None:

return action._executable_prompt

if hasattr(action, '_executable_prompt') and action._executable_prompt is not None:

prompt = action._executable_prompt()

if prompt:

return prompt

gemini-code-assist · 2025-12-17T03:49:50Z

py/packages/genkit/src/genkit/blocks/prompt.py

+            if not hasattr(action, '_executable_prompt') or action._executable_prompt is None:
+                action._executable_prompt = executable_prompt


There's an inconsistency in how the _executable_prompt is cached. The create_prompt_from_file function (which is what _async_factory calls) correctly uses a weakref to avoid circular dependencies. However, here you are caching the ExecutablePrompt instance directly, not as a weak reference. This can lead to memory leaks and inconsistent behavior. You should consistently use weakref.ref() for caching.

Suggested change

if not hasattr(action, '_executable_prompt') or action._executable_prompt is None:

action._executable_prompt = executable_prompt

if not hasattr(action, '_executable_prompt') or action._executable_prompt() is None:

action._executable_prompt = weakref.ref(executable_prompt)

gemini-code-assist · 2025-12-17T03:49:50Z

py/packages/genkit/src/genkit/blocks/prompt.py

+                if not hasattr(action, '_executable_prompt') or action._executable_prompt is None:
+                    action._executable_prompt = executable_prompt


Similar to the block above, you are caching the ExecutablePrompt instance directly instead of using a weakref. This is inconsistent with how create_prompt_from_file caches the prompt and can lead to memory leaks. Please use weakref.ref() consistently.

Suggested change

if not hasattr(action, '_executable_prompt') or action._executable_prompt is None:

action._executable_prompt = executable_prompt

if not hasattr(action, '_executable_prompt') or action._executable_prompt() is None:

action._executable_prompt = weakref.ref(executable_prompt)

gemini-code-assist · 2025-12-17T03:49:50Z

py/packages/genkit/tests/genkit/blocks/prompt_test.py

+async def test_messages_with_explicit_override() -> None:
+    """Test that explicit messages in render options are included."""
+    ai, *_ = setup_test()
+
+    my_prompt = ai.define_prompt(
+        prompt='Final question',
+    )
+
+    override_messages = [
+        Message(role=Role.USER, content=[TextPart(text='First message')]),
+        Message(role=Role.MODEL, content=[TextPart(text='First response')]),
+    ]
+
+    # The override messages should be prepended to the prompt
+    rendered = await my_prompt.render(input=None, config=None)
+
+    # Check that we have the final prompt message
+    assert any('Final question' in str(msg) for msg in rendered.messages)


This test seems to be incorrect or incomplete. The override_messages variable is defined but never used. The test name suggests it's testing an override, but no override is performed as the render method is called without passing these messages.

Please either fix the test to perform the override it's meant to test, or remove the unused variable and rename the test to reflect what it actually does.

gemini-code-assist · 2025-12-17T03:49:51Z

py/packages/genkit/src/genkit/ai/_registry.py

+        Args:
+            registry: The registry to look up the prompt from.
+            name: The name of the prompt.
+            variant: Optional variant name.
+            dir: Optional directory parameter (accepted for compatibility but not used).


The docstring for the prompt method incorrectly lists registry and dir as arguments. These are not part of the method's signature and can be confusing for users of the library.

Suggested change

Args:

registry: The registry to look up the prompt from.

name: The name of the prompt.

variant: Optional variant name.

dir: Optional directory parameter (accepted for compatibility but not used).

Args:

name: The name of the prompt.

variant: Optional variant name.

gemini-code-assist · 2025-12-17T03:49:51Z

py/packages/genkit/src/genkit/blocks/prompt.py

+            executable_prompt._prompt_action = prompt_action
+            # Also store ExecutablePrompt reference on the action
+            # prompt_action._executable_prompt = executable_prompt
+            prompt_action._executable_prompt = weakref.ref(executable_prompt)


Using a weakref here is a good practice to prevent circular dependencies and potential memory leaks. However, the corresponding lookup logic in lookup_prompt seems to handle this incorrectly, which I've commented on separately. It's great that you're thinking about memory management with these complex object relationships.

gemini-code-assist · 2025-12-17T03:49:51Z

py/packages/genkit/tests/genkit/blocks/prompt_test.py

+        # Test variant prompt
+        casual_exec = await prompt(ai.registry, 'greeting', variant='casual')
+        casual_response = await casual_exec({'name': 'Bob'})
+        assert 'Hey' in casual_response.text or "what's up" in casual_response.text.lower()


The assertion here uses or, which makes the test weaker. Given that you're using an echoModel, the output should be deterministic. You should use and to ensure both parts of the expected string are present in the response. This applies to several other new tests in this file as well (e.g., lines 320, 341, 407-408).

Suggested change

assert 'Hey' in casual_response.text or "what's up" in casual_response.text.lower()

assert 'Hey' in casual_response.text and "what's up" in casual_response.text.lower()

MengqinShen-GL added 6 commits December 16, 2025 09:15

feat(py):addin directory and file prompt loading in genkit matching J…

7037cdd

…S SDK

fix: revise formatting in prompt file and add more tests in prompt_te…

51770e9

…st based on code review

fix: raise a GenkitError for consistency instead of a generic Exception

94da1cc

fix: raise based on gemini-code-assist comments

910b307

fix: bug in prompt.py

38f0c69

add prompt demo in test samples

f75eb75

github-project-automation bot added this to Genkit Backlog Dec 17, 2025

github-actions bot added python Python config labels Dec 17, 2025

MengqinShen closed this Dec 17, 2025

github-project-automation bot moved this to Done in Genkit Backlog Dec 17, 2025

gemini-code-assist bot reviewed Dec 17, 2025

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Feat/adding directory and file prompt loading in genkit #3962

Feat/adding directory and file prompt loading in genkit #3962

Uh oh!

MengqinShen commented Dec 17, 2025

gemini-code-assist bot commented Dec 17, 2025

MengqinShen commented Dec 17, 2025

gemini-code-assist bot left a comment

gemini-code-assist bot Dec 17, 2025

gemini-code-assist bot Dec 17, 2025

gemini-code-assist bot Dec 17, 2025

gemini-code-assist bot Dec 17, 2025

gemini-code-assist bot Dec 17, 2025

gemini-code-assist bot Dec 17, 2025

gemini-code-assist bot Dec 17, 2025

Labels

2 participants

		if hasattr(action, '_executable_prompt') and action._executable_prompt is not None:
		return action._executable_prompt

		if not hasattr(action, '_executable_prompt') or action._executable_prompt is None:
		action._executable_prompt = executable_prompt

	assert 'Hey' in casual_response.text or "what's up" in casual_response.text.lower()
	assert 'Hey' in casual_response.text and "what's up" in casual_response.text.lower()

Feat/adding directory and file prompt loading in genkit #3962

Feat/adding directory and file prompt loading in genkit #3962

Uh oh!

Conversation

MengqinShen commented Dec 17, 2025

gemini-code-assist bot commented Dec 17, 2025

Summary of Changes

Highlights

Footnotes

MengqinShen commented Dec 17, 2025

gemini-code-assist bot left a comment

Choose a reason for hiding this comment

Code Review

gemini-code-assist bot Dec 17, 2025

Choose a reason for hiding this comment

gemini-code-assist bot Dec 17, 2025

Choose a reason for hiding this comment

gemini-code-assist bot Dec 17, 2025

Choose a reason for hiding this comment

gemini-code-assist bot Dec 17, 2025

Choose a reason for hiding this comment

gemini-code-assist bot Dec 17, 2025

Choose a reason for hiding this comment

gemini-code-assist bot Dec 17, 2025

Choose a reason for hiding this comment

gemini-code-assist bot Dec 17, 2025

Choose a reason for hiding this comment

Labels

2 participants