[WAI] New Qwen models + API script fix #26510
Conversation
mchenco
requested review from
a team,
craigsdennis,
kathayl,
kodster28 and
rita3ko
as code owners
github-actions
bot
added
the
product:workers-ai
|
This pull request requires reviews from CODEOWNERS as it changes files that match the following patterns:
|
|
Preview URL: https://41758c2a.preview.developers.cloudflare.com Files with changes (up to 15)
|
mchenco
changed the title
craigsdennis
left a comment
craigsdennis
left a comment
There was a problem hiding this comment.
Need a little more info on what this is solving. The changes to the script, shouldn't be necessary. I think the API should be doing what you are attempting here probably.
Documentation on at least why would be required.
|
|
||
| // Check which models are still missing valid schemas | ||
| const modelsWithoutSchema = Array.from(accumulatedModels.values()).filter( | ||
| (model) => !hasValidSchema(model), |
There was a problem hiding this comment.
Is this because the API isn't returning the same model twice?
| /** | ||
| * Check if a model has a valid schema | ||
| */ | ||
| function hasValidSchema(model) { |
There was a problem hiding this comment.
Why would the API return an invalid schema @thatsKevinJain
| if (accumulatedModels.has(model.name)) { | ||
| // Merge with existing model data | ||
| const existing = accumulatedModels.get(model.name); | ||
| const merged = deepMerge(existing, model); |
There was a problem hiding this comment.
This is order based overrides?
| "description": "Feature extraction models transform raw data into numerical features that can be processed while preserving the information in the original dataset. These models are ideal as part of building vector search applications or Retrieval Augmented Generation workflows with Large Language Models (LLM)." | ||
| }, | ||
| "created_at": "2025-09-03 20:48:38.784", | ||
| "created_at": "2025-09-04 16:38:44.980", |
There was a problem hiding this comment.
Note there are two of these models?
| "type": "number", | ||
| "description": "Confidence that this word was transcribed correctly" | ||
| "end_of_turn_confidence": { | ||
| "type": "number", |
| @@ -0,0 +1,111 @@ | |||
| { | |||
| "id": "cb254e3b-372b-4d4e-8526-ada79940427a", | |||
There was a problem hiding this comment.
Suggest doing this in a separate commit.
5 participants
Summary