Using gemini media prompt with storage file via firebase gives cors error

Gemini API
1

I can only generate textcontent if i pass the media parts as base64, using a storrage path pointing to my firebase storage files gives a cors error. I’ve allowed cors origin all for the specific bucket.

gcloud storage buckets describe gs://my-firebase-bucket.appspot.com --format="default(cors_config)"

cors_config:
- maxAgeSeconds: 3600
  method:
  - GET
  origin:
  - '*'

the file parts im sending:

[
    {
        "fileData": {
            "mimeType": "video/mp4",
            "fileUri": "gs://my-firebase-bucket.appspot.com/videos/test.mp4"
        }
    }
]

The errors i see in the console:

Access to fetch at 'https://firebaseml.googleapis.com/v2beta/projects/my-firebase/locations/us-central1/publishers/google/models/gemini-1.5-flash:generateContent' from origin 'http://localhost:3000' has been blocked by CORS policy: No 'Access-Control-Allow-Origin' header is present on the requested resource. If an opaque response serves your needs, set the request's mode to 'no-cors' to fetch the resource with CORS disabled.
functionBeingCalled.ts:34 


POST https://firebaseml.googleapis.com/v2beta/projects/my-firebase/locations/us-central1/publishers/google/models/gemini-1.5-flash:generateContent net::ERR_FAILED 502 (Bad Gateway)

It doesnt matter if i try from a deployed firebase site or localhost. Does anyone have any idea how to solve this?

2

What library are you using and can we see the code you’re trying to use that is causing this error? Despite the error, this doesn’t seem like it is really a CORS error - but rather there’s an underlying issue that is causing this. The error suggests it has to do with calling the API.

To use a Google Cloud Storage bucket for the fileUri, you need to do two things:

  • You must be using Vertex AI. You can’t use the AI Studio API. (But I’m not sure which platform “firebaseml” is using.)
  • The bucket must be in the same project you make the calls to the Gemini API from.
3

Ah okay. I’m using the firebase vertex ai package

Sort of a quick and dirty example below

import { getVertexAI } from "firebase/vertexai-preview";

getGenerativeModel(getVertexAI(firebaseApp), {
    model: "gemini-1.5-flash",
    systemInstruction: systemPrompt
  }).generateContent([parts])

The bucket is setup using firebase, it’s all setup and using firebase so im not sure why it wouldnt work in that case

1 Like