chore: fix openai vision models

[louis-jan]

Describe Your Changes

This press release aims to enable the OpenAI GPT-4o model’s vision capabilities.

Fixes Issues

• bug: GPT-4 doesn't support vision & file search in Jan #3520

Changes made

The code changes in this Git diff involve three primary modifications across different files:

1. package.json:

   • The version of the @janhq/inference-openai-extension package is updated from 1.0.4 to 1.0.5.

2. models.json:

   • In the JSON configuration for a model with the "version" set to "1.1", a new entry "vision_model": true is added within the "settings" object.

3. index.ts:

   • A new check is introduced in the transformPayload method to remove the stop property from the payload if it is an empty array. If payload.stop is an empty array, it uses destructuring to omit stop and reassigns the remaining parameters back to payload. This is done before the existing transformation logic for preview models.

These changes indicate a minor version update with an addition to the model settings to support a vision model and a code enhancement to handle empty stop word arrays in payload transformations.

[github-actions]

Barecheck - Code coverage report

Total: 68.82%

Your code coverage diff: 0.00% ▴

✅ All code changes are covered

[github-actions]

This is the build for this pull request. You can download it from the Artifacts section here: Build URL.