Fix missing pipeline implementations. #1514
Conversation
| @@ -91,6 +99,20 @@ def test_default_fallback_playbook_explanation(self): | |||
| log, | |||
| ) | |||
|
|
|||
| # 'dummy' does not have a ChatBot pipeline | |||
| @override_settings(ANSIBLE_AI_MODEL_MESH_CONFIG=json.dumps(CHATBOT)) | |||
There was a problem hiding this comment.
We can do without the json.dumps call here.
There was a problem hiding this comment.
ANSIBLE_AI_MODEL_MESH_CONFIG must be a string containing either JSON or YAML.
Using a dict is not supported (remember, the value is loaded from an environment variable).
There was a problem hiding this comment.
I mean, CHATBOT can just be a plain YAML string :-)
|
|
||
| return {k: v_or_default(k, v) for k, v in REGISTRY[pipeline_provider].items()} | ||
|
|
||
|
|
There was a problem hiding this comment.
Since this runs during the module loading, it will be a bit tricky to properly unit-test it. I feel like it would be better to isolate this in a dedicated function.
There was a problem hiding this comment.
Looks like you're right (I've added a test but it's not being detected as covering this code).
I'll move it out.
manstis
force-pushed
the
fix_missing_pipeline_implementations
branch
4 times, most recently
from
2e1700d to
edd5623
Compare
manstis
force-pushed
the
fix_missing_pipeline_implementations
branch
from
edd5623 to
189e0da
Compare
|
This PR does not need a corresponding Jira item.
Description
It was possible for a User to configure a pipeline to use a provider that did not support a given feature.
In such scenarios
get_model_pipeline(<feature>)returnedNoneleading to RTEs.This PR ensures a
NOPimplementation is returned if an explicit implementation does not exist.Testing
See "Steps to test"
Steps to test
ANSIBLE_AI_MODEL_MESH_CONFIGto include:Scenarios tested
See "Steps to test"
Production deployment