Why is there no new Ox Alpha model when requesting models from OpenRouter?

[gitmeatarru]

Why is there no new Ox Alpha model when requesting models from OpenRouter?
stealth/ox-alpha

And why isn’t it possible to enter the model name manually???

[Smalldy]

The model list is not fetched in real-time, but rather obtained from the pi-ai upstream repository.

[gitmeatarru]

I didn’t request it in real time; I requested it when adding a new provider, and there’s a button to request the list of models, which means it should request the latest list.

[Smalldy]

The models in dsh are provided by upstream, and upstream uses different methods to obtain models from different providers. For OpenRouter, models are fetched via a network API, but for opencode go, unfortunately, it's implemented as a fixed list.

[Smalldy]

You can create a custom provider to use ox-alpha-free

[gitmeatarru]

You can create a custom provider to use ox-alpha-free

how?

я пробовал но ничего не вышло, среда просто не видит модели и провайдеры

ui-onboarding:
  welcomeNoticeVersion: 2026-08-13.1
ui-theme:
  preference: dark
permission:
  defaultPreset: danger-full-access
llm-pi-ai:
  providers:
    openrouter:
      apiKeyEnv: OPENROUTER_API_KEY
      models:
        - id: nvidia/nemotron-3-ultra-550b-a55b
          name: "NVIDIA: Nemotron 3 Ultra"
          contextWindow: 512288
          maxTokens: 4096
        - id: stealth/ox-alpha
          name: ox-alpha
          contextWindow: 200000
          maxTokens: 4096
    colibri:
      baseURL: http://127.0.0.1:8000/v1
      models:
        - id: glm-5.2-colibri
          name: glm-5.2
          maxTokens: 4096
agent-default-model:
  provider: colibri
  model: glm-5.2-colibri

[gitmeatarru]

If I set up the same thing via OpenAI, an error occurs during the requests.

This turn failedOpenAI API error (404): {"message":"Not found.","type":"invalid_request_error","param":null,"code":"not_found"}

That is, it continues to receive all the data via OpenAI, including the list of models, and doesn’t use what I specified in the base URL.

[Smalldy]

I'm not at my computer right now and can't help you immediately. I have successfully called the ox alpha model provided by opencode go before. I can provide it to you later

[gitmeatarru]

not active

[Smalldy]

This looks strange. I can confirm that I can click this button

[gitmeatarru]

This looks strange. I can confirm that I can click this button

I rolled back the changes in the config file, and the button became active; I used it to create a new provider.

but it’s not clear how to fix the error with the context in 4094 tokens

Shouldn’t the environment itself receive the context of the models?

[gitmeatarru]

And what to do with this

This turn failed400: {"message":"This model's maximum context length is 4094 tokens, however your messages resulted in at least 4095 tokens. Please shorten the conversation, or restart the server with a larger CTX.","type":"invalid_request_error","param":"messages","code":"context_length_exceeded"}

[Smalldy]

I guess you only set the model ID and model display name. In addition to that, you also need to set the model's context length and output length