OpenRouter errors

[TomLucidor]

400 DeepSeek-R1 is not a valid ID since I have:

• deepseek/deepseek-r1-0528 as ID and
• DeepSeek-R1 as the name seen in the 400 error
• https://openrouter.ai/api/v1 as the base URL
• (also I have a working API key added)

I also want to use OpenRouter as the single source, since it is easier to pick up more models. For embedding we can use Ollama sometimes, but that also needs documentation.

[Lapis0x0]

I remember that the id seems to be just the model name displayed on the front end, while the name is the actual model name called by the upstream api?

Yesterday, when I tested it, I was able to call the gpt oss of openrouter normally, but to be honest, this is not normal in the front-end display logic. I might push a new update tonight with a better expression.

[TomLucidor]

But then the UX would make no sense since deepseek/deepseek-r1-0528 is what the API should see (DeepSeek-R1 would be for display), yet in OpenAI openai/gpt-5 is the ID while gpt-5 is the model name!? I don't use OpenAI at all so I naturally assumed the LiteLLM format (if it has a slash, it is an ID not a display name). e.g. https://docs.litellm.ai/docs/providers/openrouter https://docs.litellm.ai/docs/providers/gemini https://docs.litellm.ai/docs/providers/openai

Side note: Ollama would get odd since some of the models do also have slashes and ollama/user/model exists as a format for HF-compliant local models https://docs.litellm.ai/docs/providers/ollama

[Lapis0x0]

In theory, I haven’t made major changes to the underlying embedding logic from Smart Composer, so anything supported by Smart Composer should also be supported by YOLO.

Unfortunately, my current MacBook isn’t well-suited for running Ollama, so if there are related bugs, I’ll do my best to fix them, but I can’t guarantee it will go very smoothly.

[Lapis0x0]

But then the UX would make no sense since deepseek/deepseek-r1-0528 is what the API should see (DeepSeek-R1 would be for display), yet in OpenAI openai/gpt-5 is the ID while gpt-5 is the model name!? I don't use OpenAI at all so I naturally assumed the LiteLLM format (if it has a slash, it is an ID not a display name). e.g. https://docs.litellm.ai/docs/providers/openrouter https://docs.litellm.ai/docs/providers/gemini https://docs.litellm.ai/docs/providers/openai

Side note: Ollama would get odd since some of the models do also have slashes and ollama/user/model exists as a format for HF-compliant local models https://docs.litellm.ai/docs/providers/ollama

Haha, yes, that’s exactly how Smart Composer’s logic works—I just forgot to change it! It does feel quite odd this way, I’ll push an update tonight.

[TomLucidor]

Let it try it "their way" first, hopefully it would function

[TomLucidor]

Okay this is just a naming thing, I can get the chat to work, other issues are popping up tho #6