Vertex AI Error: "Publisher Model not found"

Google Cloud Build with AI AI Solutions
1

Hello,

I am experiencing a persistent error when calling the gemini-pro-vision model from a Firebase Function in project ec-dashboard-app.

The detailed error is: GoogleApiError: Publisher Model 'projects/ec-dashboard-app/locations/asia-southeast1/publishers/google/models/gemini-pro-vision' was not found or your project does not have access to it.

  1. IAM: My user-facing service accounts have the Vertex AI User role.

  2. API: The Vertex AI API is enabled.

  3. Billing: My project is linked to an active billing account.

  4. Code: My Node.js code is correct and uses the latest stable methods.

  5. SDK: The @google-cloud/vertexai package is updated to the @latest version.

  6. Vertex AI Service Agent: I have confirmed that the service agent (service-451980415127@gcp-sa-aiplatform.iam.gserviceaccount.com) exists in my project and it has the correct Vertex AI Service Agent role.

Despite this perfect configuration, the error still occurs on every API call. This appears to be a backend provisioning issue with my project that is beyond my control.

Since I am on the Basic support plan, I cannot open a technical case. Can anyone from Google or the community advise on how to escalate an issue like this?

Thank you.

2

Hello @yokesanhoo,

I think that you’re referring to an old Model of Gemini Pro Vision. In your case, you would have needed to use gemini-1.0-pro-vision as Model ID instead of gemini-pro-vision. See the Vertex AI Doc for Pro Vision.

The documentation page about Model versions and lifecycle seems up to date only in english. It seems that that the model you’re trying to use is deprecated.

I invite you to check out the latest Image generation with Gemini Documentation. Alternatively, you can check out Imagen too.

If you’re looking to learn more about Model capabilities, you can read the Google models page.

3

Hi yokesanhoo,

In addition to LeoK’s insight, the “Publisher Model not found” error usually suggests that the model is either not available in your endpoint location or that you’re using a retired or legacy model. Legacy models are restricted from new projects, and API requests to retired models will fail.

In your case, the gemini-pro-vision model, with version gemini-1.0-pro-vision-001, was retired on April 21, 2025. You may want to consider upgrading to one of the latest stable models, such as gemini-2.0-flash, which is the recommended upgrade for gemini-1.0-pro-vision-001, as stated in this documentation under retired models. However, please note that when deploying models, you should also consider your deployments and endpoints, as gemini-2.0-flash is not available in the asia-southeast1 region. Alternatively, you can select from other models available in your location that meet your requirements.

4

Thank @marckevin and @LeoK ,

just an update on the code i made

// — Vertex AI Logic —

  try {

    // Initialize Vertex AI in the correct region for Gemini

    const vertex_ai = new VertexAI({ project: "ec-xxxx-app", location: "us-central1" });



    // Use correct model path

    //const model = "projects/xxxxx/locations/us-central1/publishers/google/models/gemini-1.5-pro";

    // const model = "models/gemini-1.5-pro";

    // const model = "models/gemini-2.0-flash";

      const model = "gemini-2.0-flash";

    // Instantiate the model

    // const generativeModel = vertex_ai.getGenerativeModel({ model });

    // const generativeModel = vertex_ai.getGenerativeModel("gemini-2.0-flash");

     // Get an instance of the GenerativeModel class

const generativeModel = vertex_ai.getGenerativeModel({

model: model,

// Optional: Configure generation parameters if needed

// generationConfig: {

//   maxOutputTokens: 2048,

//   temperature: 0.9,

//   topP: 1,

//   topK: 1,

// },

});

thank you very much !