Closed iramykytyn closed 2 months ago
I think the issue here is likely that you have attached images, and we are trying to be "helpful" by switching to a model that we think has vision support. See https://github.com/tarasglek/chatcraft.org/blob/fa04130e9c37c8945fda40a05594b2ff79bd0add/src/components/PromptForm/MobilePromptForm.tsx#L78 and https://github.com/tarasglek/chatcraft.org/blob/main/src/lib/ChatCraftModel.ts#L73. Our current logic is way to simple, and based on OpenAI vs. OpenRouter.
We should make this better and not force a particular model.
cc @mingming-ma
I used OpenAI models, and then I switched to OpenRouter models and selected Cloude. I was able to use Cloude for one request, but after that it switched to germini and I was not able to select Cloude anymore. https://www.loom.com/share/bb1c14802bb346f1a8a5055c33c2bafd