In a support thread titled ‘Invalid request error in Google Flash,’ BrianC reports an error when enabling image (vision) support with Google Flash 2.0 (Gemini 2.0 Flash) via the OpenAI-compatible endpoint, receiving an ‘invalid_request_error’ referencing OpenAI’s API docs. sam suggests switching to the native Gemini endpoint (https://generativelanguage.googleapis.com/v1beta/models/gemini-2.0-flash) and using the Gemini provider, noting the OpenAI compatibility layer has limitations. BrianC confirms the correct URL is set and shares screenshots showing the error occurs after uploading a JPG with vision enabled, despite valid configuration. sam notes it works in their setup and requests the raw request/response logs via gist for deeper troubleshooting, sharing their own config screenshots. BrianC asks where to find the debug screen, and sam clarifies it’s accessible via the ‘i’ button at the bottom of the interface. The issue appears tied to endpoint compatibility rather than configuration, with vision support failing under OpenAI emulation despite correct setup.
I’m currently using Google Flash 2.0 (Gemini 2.0 Flash) with a direct API key from Google and have configured the system to use the Google Tokenizer. The chat functionality is working perfectly; however, when I try to enable image support, I encounter the following error:
{ "error": { "message": "'$.input' is invalid. Please check the API reference: https://platform.openai.com/docs/api-reference.", "type": "invalid_request_error", "param": null, "code": null } }
Has anyone else seen this? Is this error due to Gemini 2.0 not yet supporting vision features in Discourse AI? Any insights or workarounds would be appreciated.
Thanks Sam, that is the URL for the setting. The error message occurs when using a persona with that LLM and also set for Vision Enabled and you upload a picture. It fails after hitting upload for jpg. I try to validate as best possible because there are so many settings it’s easy to miss something.
