Self-hosted Studio AI Assistant ignores OPENAI_BASE_URL and attempts authentication against OpenAI API

# Bug report * [x] I confirm this is a bug with Supabase, not with my own application. * [x] I confirm I have searched the Docs, GitHub Discussions, and Discord. ## Describe the bug In a self-hosted Supabase deployment, the Studio AI Assistant appears to ignore `OPENAI_BASE_URL` and still routes requests to OpenAI's default endpoint (`https://api.openai.com/v1`) instead of the configured custom OpenAI-compatible provider endpoint. I configured Supabase Studio to use NVIDIA's OpenAI-compatible API endpoint. The environment variables are correctly injected into the Studio container, and direct API calls to the configured endpoint work successfully via `curl`. However, when using the AI Assistant inside Studio, it responds with: ```text id="0ctmfr" Incorrect API key provided: nvapi-******** You can find your API key at https://platform.openai.com/account/api-keys ``` This strongly suggests Studio is still attempting authentication against OpenAI rather than honoring `OPENAI_BASE_URL`. ## To Reproduce 1. Deploy self-hosted Supabase (in my case via Coolify / Docker). 2. Set the following environment variables for the Studio container: ```env id="xjlwm7" OPENAI_API_KEY=nvapi-******** OPENAI_BASE_URL=https://integrate.api.nvidia.com/v1 OPENAI_MODEL=minimaxai/minimax-m2.7 ``` 3. Restart / redeploy Supabase Studio. 4. Open the Studio AI Assistant. 5. Send a prompt (example: "Can you check if my database and edge functions are healthy?") 6. Observe authentication failure. Verification inside container: ```bash id="w0hq9l" env | grep OPENAI ``` Output: ```bash id="4l5n94" OPENAI_BASE_URL=https://integrate.api.nvidia.com/v1 OPENAI_API_KEY=nvapi-******** OPENAI_MODEL=minimaxai/minimax-m2.7 ``` Provider endpoint verification: ```bash id="lz1tk8" curl https://integrate.api.nvidia.com/v1/chat/completions \ -H "Authorization: Bearer $OPENAI_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "model":"minimaxai/minimax-m2.7", "messages":[{"role":"user","content":"Hel…