Skip to content

SETUP_GEMINI.md

Latest commit

 

History

History
89 lines (68 loc) · 2.75 KB

SETUP_GEMINI.md

File metadata and controls

89 lines (68 loc) · 2.75 KB

Setting Up Gemini API for Emotion Analysis

Overview

This project now uses Gemini 2.5 Flash to analyze emotions in AI responses, providing much more accurate and context-aware emotion detection compared to simple keyword matching.

Your Gemini API Key

AIzaSyALKf9SnATD_4OEh_Atx7AFXDSj6hhpfho

Setup Instructions

Option 1: Using Supabase CLI (Recommended)

  1. Install Supabase CLI (if not already installed):

    npm install -g supabase

  2. Login to Supabase:

    supabase login

  3. Link your project:

    supabase link --project-ref your-project-ref

  4. Set the Gemini API key as a secret:

    supabase secrets set GEMINI_API_KEY=AIzaSyALKf9SnATD_4OEh_Atx7AFXDSj6hhpfho

  5. Deploy the updated function:

    supabase functions deploy gemini-chat

Option 2: Using Supabase Dashboard

  1. Go to your Supabase project dashboard
  2. Navigate to Edge Functions → Manage secrets
  3. Add a new secret:
    • Name: GEMINI_API_KEY
    • Value: AIzaSyALKf9SnATD_4OEh_Atx7AFXDSj6hhpfho
  4. Click Save
  5. Redeploy your gemini-chat function

How It Works

The updated system now:

  1. Sends user message to Gemini 2.5 Flash for response generation
  2. Analyzes the AI response using a second Gemini call that:
    • Considers the conversation context
    • Analyzes tone, word choice, and emotional undertones
    • Provides accurate emotion classification (joy, curiosity, calm, anger, confusion, confidence, surprise, sadness)
    • Returns confidence and sentiment scores
  3. Displays the emotion with appropriate colors and animations

Benefits

✅ Context-aware: Understands the conversation flow, not just keywords
✅ Accurate: Uses AI to detect subtle emotional nuances
✅ Diverse emotions: No more repetitive "curiosity" - each response gets properly analyzed
✅ Fallback support: If Gemini API fails, falls back to keyword-based analysis

Testing

After setup, try different types of messages:

  • Ask happy questions → Should show joy
  • Ask for help → Should show curiosity or confidence
  • Make statements → Should show calm
  • Express frustration → Should show anger or sadness
  • Ask confusing questions → Should show confusion

Troubleshooting

Still showing same emotion?

  • Check that the Gemini API key is set in Supabase secrets
  • Verify the edge function is deployed
  • Check browser console for API errors
  • Ensure you're using the correct Supabase URL in your .env file

API errors?

  • Verify the API key is valid
  • Check Gemini API quota/billing
  • Review Supabase function logs