Skip to content

GEMINI_SETUP.md

Latest commit

 

History

History
238 lines (187 loc) · 7.11 KB

GEMINI_SETUP.md

File metadata and controls

238 lines (187 loc) · 7.11 KB

Gemini AI Integration Setup Guide

🚀 Overview

The JobAppsMailTracker now includes Gemini AI integration for intelligent email parsing! This replaces the basic keyword-based parsing with AI-powered understanding that can extract much more detailed and accurate information from job application emails.

🎯 Benefits of Gemini Integration

Before (Basic Parsing):

  • Simple keyword matching
  • Limited context understanding
  • Basic company/position extraction
  • No confidence scoring
  • Misses nuanced information

After (Gemini AI):

  • Intelligent context understanding
  • Accurate company extraction from email domains and content
  • Detailed position information with seniority levels
  • Location detection (remote, onsite, hybrid)
  • Job type classification (full-time, part-time, contract, intern)
  • Department identification (engineering, marketing, sales, etc.)
  • Experience level detection (entry, junior, mid, senior, lead)
  • Confidence scoring for extraction quality
  • Graceful fallback to basic parsing if AI fails

🔧 Setup Instructions

Step 1: Get Gemini API Key

  1. Visit Google AI Studio:

  2. Create API Key:

    • Click on "Get API key" in the top right
    • Choose "Create API key in new project" or use existing project
    • Copy the generated API key

  3. API Key Format:

    AIzaSyC... (long string of characters)

Step 2: Set Environment Variable

On macOS/Linux:

export GEMINI_API_KEY="your_api_key_here"

On Windows:

set GEMINI_API_KEY=your_api_key_here

Permanent Setup (Recommended):

macOS/Linux:

# Add to your shell profile (~/.zshrc, ~/.bashrc, etc.)
echo 'export GEMINI_API_KEY="your_api_key_here"' >> ~/.zshrc
source ~/.zshrc

Windows:

  • Add to System Environment Variables
  • Or create a .env file in the project directory

Step 3: Test the Integration

Run the test script to verify Gemini is working:

python test_gemini_integration.py

You should see output like:

✅ Gemini AI available - testing intelligent parsing...
🤖 Using Gemini AI for intelligent email parsing...

📊 Example Comparison

Sample Email:

From: recruiter@google.com
Subject: Application Confirmation - Senior Backend Software Engineer, Infrastructure Team
Body: Thank you for your application to Google for the Senior Backend Software Engineer 
position on our Infrastructure team. This is a full-time position based in Mountain View, 
CA with a competitive salary range of $150,000 - $200,000.

Basic Parsing Results:

Company: Google
Position: Software Engineer
Status: Applied
Notes: Subject: Application Confirmation - Senior Backend Software Engineer, Infrastructure Team | Basic parsing used

Gemini AI Results:

Company: Google
Position: Senior Backend Software Engineer
Status: Applied
Location: Mountain View, CA
Job Type: full-time
Experience Level: senior
Department: Infrastructure
Confidence: 0.95
AI Notes: Successfully extracted detailed position information including seniority level, location, and department

🔄 How It Works

1. Automatic Detection:

The job tracker automatically detects if Gemini is available:

  • If GEMINI_API_KEY is set → Uses Gemini AI parsing
  • If not set or fails → Falls back to basic parsing

2. Intelligent Parsing:

Gemini analyzes the email content and extracts:

  • Company name (from domain, subject, or body)
  • Job position (with full title and context)
  • Application status (based on email content)
  • Location (city, state, remote, etc.)
  • Job type (full-time, part-time, contract, intern)
  • Experience level (entry, junior, mid, senior, lead)
  • Department (engineering, marketing, sales, etc.)
  • Confidence score (how certain the AI is)

3. Enhanced Notes:

The system creates rich notes with all extracted information:

Subject: Application Confirmation - Senior Backend Software Engineer, Infrastructure Team | 
Location: Mountain View, CA | Type: full-time | Level: senior | Dept: Infrastructure | 
AI Confidence: 0.95 | AI Notes: Successfully extracted detailed position information

🛠️ Configuration

Environment Variables:

GEMINI_API_KEY=your_api_key_here

Optional Settings:

You can customize the Gemini model in gemini_parser.py:

# Change model if needed
self.model = genai.GenerativeModel('gemini-1.5-flash')  # or 'gemini-1.5-pro'

🧪 Testing

Test Basic vs AI Parsing:

python test_gemini_integration.py

Test with Real Emails:

python job_tracker.py

Look for these indicators in the output:

  • 🤖 Using Gemini AI for intelligent email parsing... = AI active
  • ⚠️ Gemini not available, using basic parsing = Fallback mode
  • Source: Gmail (Gemini AI) = AI was used
  • Source: Gmail (Basic) = Basic parsing was used

💰 Cost Considerations

Gemini API Pricing:

  • Free Tier: 15 requests per minute, 1,500 requests per day
  • Paid Tier: $0.00025 per 1K characters input, $0.0005 per 1K characters output

Typical Usage:

  • Job application email: ~500-1000 characters
  • Cost per email: ~$0.0001-0.0002
  • 100 emails per month: ~$0.01-0.02

Cost Optimization:

  • The system only uses Gemini for new emails
  • Processed emails are skipped
  • Fallback to basic parsing if API limits are reached

🚨 Troubleshooting

Common Issues:

  1. "Gemini API key required"

    • Set the GEMINI_API_KEY environment variable
    • Restart your terminal after setting

  2. "API quota exceeded"

    • Wait for quota reset (usually daily)
    • System will automatically fall back to basic parsing

  3. "Invalid API key"

    • Check your API key format
    • Ensure it starts with "AIzaSy"

  4. "Network error"

    • Check your internet connection
    • System will fall back to basic parsing

Fallback Behavior:

The system is designed to be robust:

  • Gemini fails → Uses basic parsing
  • API quota exceeded → Uses basic parsing
  • Network issues → Uses basic parsing
  • Invalid key → Uses basic parsing

🎉 Success Indicators

When Gemini is working correctly, you'll see:

  • 🤖 Using Gemini AI for intelligent email parsing...
  • Rich notes with location, job type, experience level
  • High confidence scores (0.8-1.0)
  • Source: Gmail (Gemini AI) in the output

📈 Performance Comparison

Feature Basic Parsing Gemini AI
Company Extraction Domain only Domain + context
Position Details Basic keywords Full title + context
Location Detection ❌ No ✅ Yes
Job Type ❌ No ✅ Yes
Experience Level ❌ No ✅ Yes
Department ❌ No ✅ Yes
Confidence Scoring ❌ No ✅ Yes
Context Understanding ❌ Limited ✅ Advanced

The Gemini integration transforms your job tracker from a basic email processor into an intelligent job application analyzer! 🚀