Firebase Studio API Issues

[wjggoom]

Body

I'm a coding beginner. I'm building a simple web app with Firebase Studio, and a few days ago, it was working fine, but suddenly I'm getting an API call error. I was using Gemini Flash 2.0, and it seems like something went wrong when I upgraded to version 2.5. I've looked into it, but I haven't found a solution, so I've come to this point. The web app is a simple app that uses AI to analyze chat logs and organize them into a sheet based on rules. Every time I try to analyze the AI, the console repeatedly displays the following error and fails.

Error: AI 분석 실패: INVALID_ARGUMENT: Must supply a model to generate() calls.

Current Architecture:

1. src/app/actions.ts: Calls the extractSubmissionsFromLog function from the runExtractionFlow server action.
2. src/ai/flows/extract-submissions-from-log.ts: Defines the Genkit Flow (extractSubmissionsFromLogFlow) containing the actual AI prompts and logic.
3. src/ai/genkit.ts: Creates and configures the Genkit AI object as a singleton using the getAi() function.

Solutions Tried: I tried the following solutions to resolve the error, but they did not resolve the issue.

1. model configuration: DefaultModel: 'gemini-2.0-flash' was specified in the genkit() configuration object in src/ai/genkit.ts.
2. Direct model specification: Explicitly added the model: 'gemini-2.0-flash' property within ai.definePrompt() in src/ai/flows/extract-submissions-from-log.ts.
3. Model name change: I tried changing the model name to various names, such as gemini-1.5-flash, gemini-pro, and models/gemini-1.5-flash, but the same error occurred.
4. Genkit version synchronization: Unified the versions of genkit and @genkit-ai/google-genai to ^1.14.1 in package.json.
5. Import method change: Instead of directly importing the ai object, I imported the getAi() function and called it with const ai = getAi();. Fixed variable declaration: Fixed reassignment error by declaring ai variable as let in src/ai/genkit.ts.

Guidelines

• I have read and understood this category's guidelines before making this post.

[ashmeet07]

This issue, INVALID_ARGUMENT: Must supply a model to generate() calls in Genkit, especially after an upgrade, indicates that the model reference is being dropped or misidentified exactly at the point of the API call, even if the default model is configured. This often happens when the model identifier is not passed in the format Genkit expects for that specific version/plugin.

Since you are using a singleton (getAi()), the simplest and most robust solution is to ensure you pass the complete, imported model object reference directly to the generate() function inside your flow.

---

Recommended Solution: Explicit Model Object

The most reliable fix is to explicitly import the model object reference from the Genkit Google AI plugin and pass it directly within your flow definition.

1. Update Imports and generate() Call

In your flow file (src/ai/flows/extract-submissions-from-log.ts), modify the imports and the call to generate():

```typescript
// src/ai/flows/extract-submissions-from-log.ts

// 1. Import the specific model object you want to use
// Using gemini-2.5-flash as the best current option for speed/cost:
import { gemini25Flash } from '@genkit-ai/google-genai';
import { defineFlow, generate, z } from 'genkit';

// ... other flow code ...

export const extractSubmissionsFromLogFlow = defineFlow(
// ... flow setup
async (/* input */) => {
// ... logic

// 2. Pass the imported model object directly to the generate call
const response = await generate({
  model: gemini25Flash, // <-- USE THIS IMPORTED OBJECT
  prompt: 'Your extraction prompt here...',
  // ... other config (e.g., config, output schema)
});

// ... return

}
);
```

2. Verify Genkit Initialization (Optional Check)

While the fix above should solve the issue, ensure your `src/ai/genkit.ts` is initializing the plugin correctly, potentially using the recommended model reference format:

```typescript
// src/ai/genkit.ts

import { genkit } from 'genkit';
import { googleAI, gemini25Flash } from '@genkit-ai/google-genai';

// ... other code

let ai: ReturnType;

export function getAi() {
if (!ai) {
ai = genkit({
plugins: [
googleAI({
// Your API key handling logic here
})
],
// Set the default model using the reference object format
// This is a good fallback, but the explicit call in step 1 is key.
defaultModel: gemini25Flash,
});
}
return ai;
}
```

The error message is extremely specific to the missing model parameter on the `generate()` call. Forcing the use of the imported model object reference (`gemini25Flash`) inside the flow is the most reliable way to override any issue where the default model name/string is failing validation during the API request.

[Sukhtumur]

Error cause might be missing model binding after Genkit upgrade.
Maybe you should:

src/ai/genkit.ts:
import { genkit } from "@genkit-ai/core";
import { googleAI } from "@genkit-ai/google-genai";

export const getAi = () =>
genkit({
plugins: [
googleAI({
apiKey: process.env.GOOGLE_API_KEY,
defaultModel: "models/gemini-2.5-flash",
}),
],
});

extract-submissions-from-log.ts:
const ai = getAi();
ai.definePrompt({
model: "models/gemini-2.5-flash",
prompt: "...",
});

Reinstall packages:
rm -rf node_modules package-lock.json
npm install

Ensure GOOGLE_API_KEY is set and redeploy.