All committers have signed the CLA.

CLA should be signed, and I've confirmed that it's working for me with Qwen3-4B with 32k context at ~100 tokens/sec :) doesn't always decide to create tags but that's all down to the model. Might be worth some customization of the prompt in the future.

Set to the proper branch and merged from dev to clear the conflicts.

Hey thanks for the PR! Noticed there's an process.env.OPENAI_NAME environment variable, is that required to be changed for the third-party providers?

I'll be making some changes to make the config.compatibility dynamic based on the config.baseURL and merge this afterwards :)

Also wouldn't it be better if we remove the OPENAI_PROJECT and OPENAI_ORGANIZATION environment variables for minimal complexity to set it up? Will that cause any potential issues?

@daniel31x13

As far as I could tell what I added for the OPENAI_NAME variable seems to be seems to be something related to the library itself related to compre atibility. Not entirely sure what it does in the end because the docs aren't all that helpful there.

https://ai-sdk.dev/providers/ai-sdk-providers/openai

Likely setting the compatibility setting dynamically based on if the base is changed is likely to work fine I would think given what it does have there.

The project and organization parameters are entirely optional and I added them while I was there because I know they can be useful for even the official API since they can get used for logging more fine grained billing and other stats for usage if desired.

I already used the @ai-sdk/openai-compatible package instead which now directly handles everything. Thanks for your help!