Firestore Real-time Poll and Message Processor with Google AI

This Go application listens to new user messages and poll updates in Firestore, processes them in real-time, and responds with AI-generated messages using the Gemini model from Google AI. It also periodically sends status updates to a chat platform based on the current state of polls and conversations.

Features

• Real-time Firestore Integration: Listens for new user messages and processes them immediately.
• AI-Powered Responses: Uses Google AI's Gemini model to generate responses based on the conversation summary.
• Poll Monitoring: Fetches poll status from Firestore and updates the conversation summary.
• Concurrency: Utilizes Go's sync.WaitGroup and sync.Mutex to ensure concurrent processes run safely.

Prerequisites

• Go (1.16+)
• Firestore database with proper collections set up.
• Service account credentials with access to your Firestore project.
• .env file with the necessary configuration.
• Firebase Admin SDK and Google AI Genkit SDK.

Setup

Firestore Collections

1. User Messages: This collection (gccdpune-user) stores incoming user messages.
2. Poll Collection: This collection (gccdpune-poll) contains poll questions and options.
3. Ping Collection: This collection (gccdpune-go-pings) stores the AI-generated responses.

Environment Variables

Create a .env file in the root of your project:

# .env
SERVICE_ACCOUNT_PATH=".keys/serviceAccountKey.json"

Firestore Document Schema

User Messages Collection (gccdpune-user):

• id: string (unique identifier)
• message: string (user's message)
• timestamp: timestamp (message creation time)
• processed: boolean (whether the message has been processed)

Poll Collection (gccdpune-poll):

• question: string (the poll question)
• options: map (keyed by option label, containing poll options with their text and voters)

Installation

1. Clone this repository:

git clone https://github.com/your-username/firestore-ai-poll.git
cd firestore-ai-poll

2. Install the necessary Go modules:

go mod tidy

3. Run the application:

go run main.go

How It Works

1. Mark Existing Messages as Processed: The program first scans and marks all existing unprocessed messages in the gccdpune-user collection as processed, so that only new messages are handled.

2. Listen for New Messages: The program listens for any new user messages and processes them by generating a response using the Gemini AI model.

3. Poll Monitoring: The app periodically checks the status of a poll in Firestore and generates a summary, which is then used to update the conversation summary.

4. AI-Generated Responses: When a new message arrives, the Gemini AI model generates a response, and it is stored in Firestore for display in the chat.

Contributing

Feel free to fork this repository, create a new branch, and submit pull requests for any improvements or features you'd like to add.

License

This project is licensed under the MIT License.