[Platform] Please add support for OpenAI-compatible endpoints

[singleparadox]

A lot of platforms support OpenAI-like endpoints, such as Ollama, Llama-server, Gemini, LM Studio and more.

Supporting this feature will make it easier to use most of the platforms that support it.

Maybe this could be a new platform bridge?

[singleparadox]

https://ai.google.dev/gemini-api/docs/openai
https://github.com/ggml-org/llama.cpp/blob/master/tools/server/README.md
https://ollama.com/blog/openai-compatibility
https://lmstudio.ai/docs/app/api/endpoints/openai

Here are some references.

[chr-hertel]

Hey @singleparadox, for Ollama, LM Studio, and Gemini WE habe dedicated Bridges - please have a look at the examples directory :)

[chr-hertel]

However, i think we need to enable this, or document, if already possible.

[singleparadox]

Hey @singleparadox, for Ollama, LM Studio, and Gemini WE habe dedicated Bridges - please have a look at the examples directory :)

Yes, I’m aware, but I still think it would be beneficial to support OpenAI-compatible endpoints since they provide standardized routes. Many projects already use them, so it would be very useful for the Symfony AI component to support this as well.

[ineersa]

Hey @singleparadox, for Ollama, LM Studio, and Gemini WE habe dedicated Bridges - please have a look at the examples directory :)

For local hosting a lot of people usually run llama.cpp / vLLM which provide OpenAI API compatible servers, at least somewhat compatible.

ai/src/platform/src/Bridge/OpenAi/AbstractModelClient.php

Line 21 in 2f05b00

protected static function getBaseUrl(?string $region): string

URLs are hardcoded, those should be manageable via env, or configurable.