Spark V2 - Smart Query Assistant (CONTEXT-AWARE)

Intelligent query validation with context-aware pattern detection for Intent and B2B search.

🎯 Features

• Context-Aware Detection: Intelligent pattern matching instead of keyword lists
• Smart Mode Separation: Understands context, not just keywords
• 7-Rule Validation System: Comprehensive query quality analysis
• Real-Time Feedback: Instant validation as you type
• No False Positives: Distinguishes between topics and filters
• Advanced Options: Context phrases, lens, and granularity controls

🧠 Context-Aware Intelligence

How It Works

Instead of hard-coded keyword lists, Spark V2 uses intelligent pattern detection:

Location Detection (Context-Aware)

• ✅ Allows: "interested in San Francisco culture" (topic of interest)
• ❌ Blocks: "people in San Francisco" (location filter)
• Logic: Only flags when location is used to filter WHO people are

Persona Detection (Pattern-Based)

• ✅ Detects: "software engineers", "product managers", "CEOs"
• ✅ Handles: Plurals, compound titles, variations
• Logic: Regex patterns match job title structures

Intent Detection (Behavioral)

• ✅ Detects: "interested in", "passion for", "care about"
• ✅ Context: Distinguishes intent signals from forbidden terms
• Logic: Pattern-based behavioral signal detection

---

🔒 Validation Rules

Intent Mode

FORBIDDEN (context-aware detection):

• ❌ Personas: Job titles, roles (e.g., "software engineers", "CEOs")
• ❌ Demographics: Age, gender, income, education
• ❌ Location Filters: "people in [Location]", "based in [Location]"

ALLOWED:

• ✅ Location Topics: "interested in San Francisco culture"
• ✅ Behavioral Intent: "interested in", "passion for", "care about"

REQUIRED:

• ✅ Intent Signals: Must include behavioral language

Example Good Query:

people interested in sustainable living and environmental conservation

Example Bad Query:

software engineers in San Francisco interested in AI

Fails: persona "software engineers" + location filter "in San Francisco"

---

B2B Mode

FORBIDDEN (pattern detection):

• ❌ Intent Language: "interested in", "looking for", "want to"
• ❌ Behavioral Signals: "recently", "about to", "planning to"

REQUIRED:

• ✅ Company Attributes: "companies", "revenue", "employees", "industry"

Example Good Query:

SaaS companies with 50-200 employees generating $5M-$20M annual revenue

Example Bad Query:

companies interested in adopting AI technologies

Fails: intent language "interested in"

---

📊 Validation Rules (Weighted)

1. Length Check (10% weight) - 10-500 characters optimal
2. No Marketing Language (15% weight) - Forbid vague terms
3. No Questions (15% weight) - Keywords only
4. Mode Purity (30% weight) - MOST IMPORTANT - Context-aware separation
5. Specificity (15% weight) - Concrete details
6. No Keyword Stuffing (10% weight) - Avoid repetition
7. Actionability (5% weight) - Clear and actionable

⚖️ Penalty System

Mode Purity (30% weight) has the highest impact:

• 1 violation → 40/100 score (~18-point penalty)
• 2 violations → 20/100 score (~24-point penalty)
• 3+ violations → 0/100 score (~30-point penalty)

🧪 Test Cases

Intent Mode

✅ Pass (97/100):

people interested in artificial intelligence and machine learning technologies

✅ Pass (97/100):

people interested in San Francisco culture and Bay Area lifestyle

Topic, not filter - correctly allowed

❌ Fail (79/100):

software engineers interested in AI

Persona detected: "software engineers"

❌ Fail (73/100):

people in San Francisco interested in technology

Location filter detected: "people in San Francisco"

B2B Mode

✅ Pass (100/100):

SaaS companies with 50-200 employees generating $5M-$20M annual revenue

❌ Fail (79/100):

companies interested in cloud computing technologies

Intent language detected: "interested in"

---

🚀 Quick Start

# Install dependencies
pnpm install

# Start development server
pnpm run dev

# Build for production
pnpm run build

# Preview production build
pnpm run preview

📦 Bundle Size

• Total (gzipped): 147.57 KB
• JavaScript: 143.23 KB
• CSS: 4.04 KB
• HTML: 0.34 KB
• Status: ✅ Under 150KB target

🛠️ Tech Stack

• React 19.2.3 - UI framework
• TypeScript 5.9.3 - Type safety
• Vite 7.2.7 - Build tool
• Tailwind CSS 4.1.18 - Styling
• Radix UI - Accessible components
• Lucide React - Icons
• Wouter - Lightweight routing

📁 Project Structure

src/
├── components/              # UI components
│   ├── ModeSelector.tsx     # Intent vs B2B toggle
│   ├── QueryInput.tsx       # Query input with counts
│   ├── ValidationDisplay.tsx # Score & rules display
│   └── AdvancedOptions.tsx  # Collapsible options
├── lib/
│   ├── queryValidation.ts   # Context-aware validation
│   └── utils.ts             # Utility functions
├── pages/
│   └── SparkPage.tsx        # Main page
├── App.tsx                  # Routing
├── main.tsx                 # Entry point
└── index.css                # Global styles

🎨 Design Principles

• Context-Aware: Intelligent pattern detection
• Minimal Dependencies: Only 14 packages
• Simple Components: 4 focused components
• Performance First: Bundle under 150KB
• Accessible: Keyboard navigation, ARIA labels
• Responsive: Mobile-first design

📝 License

MIT

🙏 Acknowledgments

Built with context-aware intelligence for maximum accuracy and minimal false positives.