Assistant Archival Cleanup #1011
Conversation
…all assitatns before archiving and deleteling those references as well
…assistants when the referenced assistant is going to be archived
Codecov ReportAttention: Patch coverage is ✅ All tests successful. No failed tests found.
Additional details and impacted files📢 Thoughts on this report? Let us know! |
| version_query = None | ||
| if assistant.is_working_version: | ||
| version_query = list( | ||
| map( |
There was a problem hiding this comment.
I might be wrong, but I don't think we have to map the ids to strings?
| return self.experiment_set.filter(is_archived=False) | ||
|
|
||
| def get_related_pipeline_node_queryset(self): | ||
| def get_related_pipeline_node_queryset(self, query=None): |
There was a problem hiding this comment.
Since the query param is a list of ids, I suggest that we name and type it accordingly:
def get_related_pipeline_node_queryset(self, assistant_ids: list):We could even pass in a queryset that returns the ids. We did something similar here for example. This wouldn't require parsing the query results to a list, which is nice, but up to you on which approach you'd like to take.
Also, looks like this parameter will come in useful to optimize the archive method. So instead of iterating through all assistant versions (in the case where we're archiving the working version), we can simply fetch the version ids and call this method with it. Same with get_related_experiments_queryset.
There was a problem hiding this comment.
After reviewing this and testing it locally I think we need to take a step back and look more holistically at how we trace references to objects and make that visible in the UI.
I've started a doc to make it easier to collaborate on the thoughts and ideas: https://docs.google.com/document/d/1Z09GNpO17izoVcRPGSmppNthFruyRJ0033T-RYvapoE/edit?tab=t.0
| experiment = ExperimentFactory(pipeline=pipeline) | ||
| experiment.assistant = v2_assistant |
There was a problem hiding this comment.
this isn't a valid state - and experiment should either have an assistant or a pipeline but not both
| assert assistant.is_archived is True # archiving successful | ||
|
|
||
| @patch("apps.assistants.sync.push_assistant_to_openai", Mock()) | ||
| def test_archive_versioned_assistant_with_still_exisiting_experiment_and_pipeline(self): |
There was a problem hiding this comment.
This test seems to be the same as the previous one since the presence of the pipeline has no effect because it does't reference the assistant
| pipeline = PipelineFactory() | ||
| NodeFactory(type="AssistantNode", pipeline=pipeline, params={"assistant_id": str(assistant.id)}) | ||
| exp = ExperimentFactory(pipeline=pipeline) | ||
| exp.assistant = assistant |
There was a problem hiding this comment.
you shouldn't set this property since you're testing the reference through the pipeline
| NodeFactory(type="AssistantNode", pipeline=v2_exp.pipeline, params={"assistant_id": str(v2_assistant.id)}) | ||
| NodeFactory(type="AssistantNode", pipeline=v2_exp.pipeline, params={"assistant_id": str(v2_assistant.id)}) | ||
| v2_exp.assistant = v2_assistant | ||
| v2_exp.save() |
There was a problem hiding this comment.
I don't follow what's happening here. Creating a new version should take care of creating versioned nodes as well.
| or version.get_related_pipeline_node_queryset().exists() | ||
| ): | ||
| return False | ||
| for version in self.versions.all(): |
There was a problem hiding this comment.
rather than re-iterating over the queryset you could accumulate the IDs in the previous loop
|
noting that we are moving forward with changes from the doc so those commits will be coming in later and I will re-requests reviews |
Description
990
Followup logic changes for assistant versioning archiving.
Now, when a assistant is archived that is a working version, all its versions is also archived and deleted from open AI.
Also from a UI standpoint. When it checks whether is can be archived, if its the working version it checks for all dependencies for all the versions not just the working version and displays everything that needs to be archived first. This way, the version won't be archived with the open AI deleted with something (exp or pipeline) still referencing it)
User Impact
Demo
Docs