GenAI Fireside Chat application

Workshop

This project is a starting point for a Flutter application.

Prerequisites

• Ensure you have Flutter installed on your development machine. If not, visit the Flutter installation guide.
• An IDE that supports Flutter (e.g., Android Studio, VS Code) is recommended.
• A Google account for Firebase and Google Cloud setup.

Run a Flutter Project Template

• Execute the current app using terminal:

  flutter pub get
  flutter run
  

• Alternatively, use your IDE's run button or shortcuts to launch the app.

ChatGPT

Add the Chat GPT API key

• Follow the link to OpenAI developer platform and create new API key
• Add this key value to config/config.json as chatGptApiKey Call the Chat GPT API using REST API
• Remove placeholder code from chat_gpt_service.dart
• Uncomment REST API call

Gemini

Add the Gemini AI API key

• Follow the link to Google AI Studio platform and create new API key
• Add this key value to config/config.json as geminiApiKey Add google_generative_ai dependency
• Uncomment google_generative_ai in pubspec.yaml
• Execute flutter pub get using terminal Call the Gemini API using plugin
• Remove placeholder code from gemini_service.dart
• Uncomment generative AI SDK call

Gemma locally

Add flutter_gemma dependency

• Uncomment flutter_gemma in pubspec.yaml
• Execute flutter pub get using terminal Upload model to device and setup
• Follow plugin setup manual to upload Gemma to device from here Call the Gemma API using plugin
• Remove placeholder code from gemma_local_service.dart
• Uncomment flutter_gemma SDK call

Execute

• Execute the current app using terminal:

  flutter run --dart-define-from-file=config/config.json
  

• Alternatively, use your IDE's run button or shortcuts to launch the app with environment parameters from config/config.json

Gemini with Vertex AI API

Create a New Project in Firebase (you also can use already existing account)

• Follow the instruction (Steps 1 and 2) here Enable Vertex API for this project
• To access the Vertex AI service, you need enable it in the GCP console. The instructions can be found here Get the Google Cloud API key and URL
• Install Google Cloud Command Line Interface
• Log in to Google Cloud using CLI

  gcloud auth login
  

• Get access token

  gcloud auth print-access-token
  

• Construct your api Url using template from here
• Add these token and url values to config/config.json as googleCloudApiUrl and googleCloudApiKey

IMPORTANT NOTICE: The token is valid during one hour, then you have to request it again

Call the Gemini API using plugin

• Replace GeminiService by GeminiVertexService in service_map.dart

Gemini with Vertex AI Firebase API

Connect Firebase project

• Add the application to your Firebase project, the instruction is here Call the Gemini API using Vertex AI Firebase API
• Remove placeholder code from gemini_fireabase_service.dart
• Uncomment firebase_core and firebase_vertexai in pubspec.yaml
• Execute flutter pub get using terminal
• Uncomment Firebase initialization in main()
• Replace GeminiVertexService by GeminiFirebaseService in service_map.dart Call the Gemini API using plugin
• Replace GeminiService by GeminiVertexService in service_map.dart

Gemini with Vertex AI API and Firebase Cloud Functions (in order not to request new token every hour)

Deploy Cloud Functions

• Take a look to code in functions/src/index.ts, there is the cloud function
• Deploy the cloud function to your firebase account, there are instructions Call the Gemini API using Cloud Functions
• Remove placeholder code from gemini_cloud_service.dart
• Uncomment cloud_functions in pubspec.yaml
• Execute flutter pub get using terminal
• Uncomment Firebase initialization in main()
• Replace GeminiFirebaseService by GeminiCloudService in service_map.dart

Call Gemma from Google CLoud using REST API

Deploy Gemma to your Google Cloud project

• Deploy Gemma to your Cloud from Models Garden Get deployed Gemma URL
• Get apiUrl from Endpoints
• Add this URL to config/config.json as gemmaCloudApiUrl and googleCloudApiKey Call the Gemma using REST API
• Uncomment code in gemma_service.dart
• Replace GemmaLocalService by GemmaService in service_map.dart

For help getting started with Flutter development, view the online documentation, which offers tutorials, samples, guidance on mobile development, and a full API reference.