Can't select model from OpenaiLike

[Intiiii]

Describe the bug

I know you can select a openai like api thingy, but is there a way to configure what model you want to use for it? by default it uses the claude-3-5-sonnet, as stated in the utils/constants.ts

but i cant figure out how to change it without changing the default model setting

Link to the Bolt URL that caused the error

none

Steps to reproduce

1. configure the env.local
2. run the project

Expected behavior

you can select the model that you want to use

Screen Recording / Screenshot

it looks like it wants to display something, but there is nothing to display

Platform

• OS: Windows
• Browser: Chrome
• Version: Latest

Additional context

No response

[the-Arka]

in .env set both:
OPENAI_LIKE_API_BASE_URL=
OPENAI_LIKE_API_KEY=
if it's still does not shows up, restart the bolt (happened to me few times)

[chrismahoney]

Please see #251 to keep track of provider refactors that will improve this support.

[dctfor]

@Intiiii
This issue actually is the same as with Ollama, plz check this other related issue
I currently have a PR for that, may be not the best solution but it worked for me after several iterations for making it work

This happens because the steamtext method befor sending a POST request to provider, the model list skips the selected model because in the backend at that specific point the model is not updated, so taking the model name from the front straightforward, wont hurt since in the UI it's updated, so if it's not in the staticModels[] array, it will fail with current main version at b7d609d

[champ2050]

I’m trying to use GitHub Models as OpenAI-like in the .env.local file. I’ve entered the base URL (https://models.inference.ai.azure.com) and the token API key (i.e., my GitHub Models key). However, I don’t see any option to specify the GitHub model name in the .env.local or constants files.

Despite this, I tried to proceed, restarted the application a couple of times, but the model names don’t appear in the dropdown when I select the OpenAI-like option.

From what I understand, GitHub Models require us to pass the model name for them to function properly. Could someone check this and add a feature to pass the model names too ?

[chrismahoney]

Not familiar with Github Models yet, tagging this to be included in provider overview in #276 to make sure it's captured.