Firebase Function Call

This is an example project to demonstrate issue with Firebase AI Live function calling.

Update (2025-11-09): fixed by passing functionCall.id.

Project Configuration

Prerequisites

1. Firebase Project: Create a Firebase project at Firebase Console
2. Gemini API: Ensure the Gemini API is enabled for your Firebase project (API key is managed automatically via google-services.json)

Setup Steps

1. Download google-services.json:

   • Go to Firebase Console → Project Settings → General
   • Under "Your apps" section, download google-services.json for your Android app
   • Place the file in /app/google-services.json

2. Sync Gradle: Open the project in Android Studio and sync Gradle files

3. Grant Audio Permission: The app will request microphone permission at runtime

How to Test the Project

Expected Behavior

This project demonstrates a function calling issue with the Firebase AI Live API where the AI doesn't continue after receiving a function response.

Test Steps

1. Start the app on your Android device or emulator

2. Tap the microphone button (floating action button in bottom-right corner)

   • The button will turn red indicating the conversation is active
   • Watch the logs appear on screen

3. Wait for AI greeting

   • The AI should say: "Hello! Ask me what one plus one is!"
   • You'll see logs showing:
     • 🔧 Initializing LiveModel...
     • 🔌 Connecting to session...
     • ✅ Session connected
     • 🎤 Starting audio conversation...
     • 💬 Sending initial greeting...
     • ✅ Greeting sent

4. Ask the AI: "What is one plus one?"

   • The AI should call the onePlusOne function
   • You'll see logs showing:
     • 📞 Function called: onePlusOne
     • 🔢 Executing onePlusOne...
     • ✅ onePlusOne result: {"result":2}
     • 📤 Returning FunctionResponsePart

5. 🐛 ISSUE: AI doesn't continue

   • After the function response is sent, the AI should respond with the answer
   • However, the AI does not continue the conversation
   • The function call completes successfully but no audio response is generated

Expected vs Actual

Expected: AI should say something like "One plus one equals two" after receiving the function result

Actual: AI receives the function response but doesn't generate any audio output

Project Structure

• onePlusOneTool: A simple function declaration that returns {"result": 2}
• handleFunctionCall(): Handler that executes the function and returns the response
• startAudioToAudio(): Initializes the Live API with audio modality and function calling
• System Instruction: Tells the AI to use the onePlusOne tool when asked about 1+1
• Initial Message: Prompts the AI to greet the user and ask them to ask about 1+1

Key Implementation Details

• Model: gemini-live-2.5-flash-preview
• Response Modality: Audio
• Function Call Handler: Synchronous handler that logs each step
• Audio Permission: Required for startAudioConversation()

Issue Reproduction

This project reliably reproduces the issue where:

1. Function calls are successfully made
2. Function responses are successfully returned
3. But the AI does not continue generating audio output after receiving the function response

The on-screen logs provide complete visibility into the function calling lifecycle.