Add more eval examples + filtering examples by language + fix git concurrent usage (#28719)
Release Notes: - N/A --------- Co-authored-by: michael <michael@zed.dev> Co-authored-by: agus <agus@zed.dev>
This commit is contained in:
parent
a8b1ef3531
commit
d74f0735c2
76 changed files with 365 additions and 8 deletions
|
@ -0,0 +1,3 @@
|
|||
url = "https://github.com/vercel/ai.git"
|
||||
revision = "1766edec300deb05c84bb7fefc034af4c2bc1165"
|
||||
language_extension = "ts"
|
|
@ -0,0 +1,3 @@
|
|||
1. Introduces a new changeset file that documents a patch for the '@ai-sdk/provider' package. The changeset indicates a chore task where 'LanguageModelV2File' is being extracted, suggesting a refactoring effort to modularize the codebase by separating file-related types into their own module.
|
||||
2. Modifications to the language model v2 index file where a new export statement for 'language-model-v2-file' has been added. This change reflects the extraction mentioned in the changeset and makes the new file type available to other parts of the application. Additionally, there are significant changes to the language model v2 implementation file where the inline file type definition has been replaced with the newly extracted 'LanguageModelV2File' type, both in the main model interface and in the stream part union type, demonstrating the consolidation of file-related types into a single, reusable definition.
|
||||
3. Present the newly created 'language-model-v2-file.ts' file which defines the 'LanguageModelV2File' type with comprehensive documentation. The type includes two properties: 'mediaType' which specifies the IANA media type of the file with a reference to the official media types registry, and 'data' which can be either a base64 encoded string or binary data, with clear documentation about maintaining the original format from the API without unnecessary conversion. This new file represents the extracted type that is now being used throughout the codebase.
|
|
@ -0,0 +1 @@
|
|||
We need to improve how our language model handles file attachments by making the file type definitions more modular and reusable. Currently, file-related properties are defined inline within the model’s response and stream types, which makes maintenance harder and duplicates documentation. The goal is to extract these definitions into a dedicated type that can be shared consistently across both static responses and streaming payloads. The new type should include clear documentation about media types (referencing IANA standards) and support both base64 and binary data formats without unnecessary conversions. This change should maintain backward compatibility while centralizing the file structure definition for better type safety and readability. Focus on clean separation of concerns, and ensure the extracted type is properly exported and imported where needed.
|
Loading…
Add table
Add a link
Reference in a new issue