Add support for proxy for openai/gemini modes

[fireattack]

As title.

These services are geo-locked to only some countries and many people need to use proxy to visit them, including API endpoints.

[0xdevalias]

Related:

• https://github.com/openai/openai-node/issues/85

  • Our upcoming v4 also has support for proxies, as documented here:

    • https://github.com/openai/openai-node/tree/v4#configuring-an-https-agent-eg-for-proxies

    Originally posted by @rattrayalex in https://github.com/openai/openai-node/issues/85#issuecomment-1627939433

    • https://github.com/openai/openai-node#configuring-an-https-agent-eg-for-proxies
• https://github.com/google-gemini/generative-ai-js/issues/16
• https://github.com/google-gemini/generative-ai-js/pull/38

[0xdevalias]

For openai proxy support, see also this thread:

• https://github.com/jehna/humanify/issues/265

Of which this most recent update I added is most relevant:

Looks like openai v5 is out now, which uses the native fetch, and has better support for proxies:

Here's some updates on the timeline as to when node-fetch will no longer be included in it, which I believe will resolve this error:

seems that openai is still using node-fetch, so it might not work

• https://github.com/openai/openai-node/blob/212710db8c8c139392c6532c0eccfd13558ef2d4/README.md?plain=1#L1-L5

  • We're actively working on a new alpha version that migrates from node-fetch to builtin fetch.

    Please try it out and let us know if you run into any issues!

    • https://community.openai.com/t/your-feedback-requested-node-js-sdk-5-0-0-alpha/1063774

    • We’re pleased to share that a new alpha of the 5.0.0 version of the JavaScript / TypeScript SDK is available on npm and in the alpha branch of the GitHub repository. This version of the SDK represents a breaking change to the existing 4.x SDK, as we have migrated completely to using built-in fetch for HTTP requests across all environments (Node.js, Deno, etc). However, we do expect that most of your code will “just work” with minor adjustments in types.

Looks like openai v5 is officially out, which moves to the built-in fetch:

• https://github.com/jehna/humanify/pull/459
• https://github.com/openai/openai-node/releases/tag/v5.0.0
  • https://github.com/openai/openai-node/blob/master/CHANGELOG.md#500-2025-05-29
• https://github.com/openai/openai-node/releases/tag/v5.0.1
  • https://github.com/openai/openai-node/blob/master/CHANGELOG.md#501-2025-05-29

We can see the migration guide here:

• https://github.com/openai/openai-node/blob/master/MIGRATION.md

  • The main changes are that the SDK now relies on the builtin Web fetch API instead of node-fetch and has zero dependencies.

We can see that it now supports a new fetchOptions property for customizing the underlying fetch:

• https://github.com/openai/openai-node/blob/master/MIGRATION.md#removed-httpagent-in-favor-of-fetchoptions

  • Removed httpAgent in favor of fetchOptions

    The httpAgent client option has been removed in favor of a platform-specific fetchOptions property. This change was made as httpAgent relied on node:http agents which are not supported by any runtime's builtin fetch implementation.

    If you were using httpAgent for proxy support, check out the new proxy documentation.

• https://github.com/openai/openai-node#fetch-options

  • Fetch options If you want to set custom fetch options without overriding the fetch function, you can provide a fetchOptions object when instantiating the client or making a request. (Request-specific options override client options.)

fetchOptions is also the basis for how to configure proxies now:

• https://github.com/openai/openai-node#configuring-proxies

  • Configuring proxies

    To modify proxy behavior, you can provide custom fetchOptions that add runtime-specific proxy options to requests:

Originally posted by @0xdevalias in https://github.com/jehna/humanify/issues/52#issuecomment-2921148059

Originally posted by @0xdevalias in https://github.com/jehna/humanify/issues/265#issuecomment-2921150326