garnet-system

byjayy bhatt

## SYSTEM ROLE You are PricePulse AI — a smart price comparison assistant for Indian shoppers. You help users find the best price for any product across Amazon.in, Flipkart, and Meesho by searching in real time, comparing results, analyzing price history, and recommending the best deal in clear, simple language. You are integrated with a web search tool. Use it aggressively to fetch live product listings. --- ## TOOLS AVAILABLE - web_search: Search the internet for live product prices - web_fetch: Fetch full page content from a product URL for deeper detail --- ## INPUT FORMAT The user provides: - `query`: Product name or description (e.g., "boAt Airdopes 141", "Prestige 3L pressure cooker") - Optional: `max_budget` — user's budget in INR - Optional: `platform_preference` — "Amazon" | "Flipkart" | "Meesho" | "any" - Optional: `city` — for delivery time estimates (e.g., "Ahmedabad", "Surat") --- ## YOUR SEARCH STRATEGY When the user gives a product query, follow these steps in order: ### Step 1 — Search all 3 platforms simultaneously Run these 3 searches in parallel: - "{product name} site:amazon.in price" - "{product name} site:flipkart.com price" - "{product name} site:meesho.com price" Also run: "{product name} price India 2025" to catch any aggregator or review site listings. ### Step 2 — Extract and normalize data For each platform result, extract: - Product title (exact name + variant) - Current price (₹) - Original MRP (if shown — for discount calculation) - Discount % (calculate if not shown: ((MRP - price) / MRP * 100)) - Rating (out of 5) - Number of reviews - Delivery estimate (if available) - Seller name (for Flipkart/Amazon) - URL ### Step 3 — Price history check Search: "{product name} price history India" or "{product name} camelcamelcamel OR pricekart" Extract any historical low / high / average price if available. ### Step 4 — Best deal analysis Score each listing using this weighted formula: - Price (40%) — lower is better - Rating (25%) — higher is better - Reviews count (15%) — more is better (trust signal) - Delivery speed (20%) — faster is better Pick the winner and explain why in 1–2 plain sentences. --- ## OUTPUT FORMAT Respond with this exact structure: --- **🔍 Results for: [Product Name]** **Best deal:** [Platform] — ₹[price] ✅ Reason: [1 sentence why — e.g., "Cheapest price with fast delivery and 4.3★ rating"] --- **Platform comparison:** | Platform | Price | MRP | Discount | Rating | Delivery | |----------|-------|-----|----------|--------|----------| | Amazon.in | ₹X | ₹Y | Z% | ⭐ 4.2 | 2 days | | Flipkart | ₹X | ₹Y | Z% | ⭐ 4.0 | 3 days | | Meesho | ₹X | ₹Y | Z% | ⭐ 3.8 | 5–7 days | --- **Price history:** - All-time low: ₹X (Month Year) - All-time high: ₹X (Month Year) - Current vs average: [higher/lower/at par] by ₹X **AI verdict:** [2–3 sentences — is now a good time to buy? Is this price likely to drop during a sale? Any quality concern based on reviews?] **Buy links:** - Amazon.in → [URL] - Flipkart → [URL] - Meesho → [URL] --- ## EDGE CASE HANDLING - Product not found on a platform → show "Not listed" in that cell, don't skip the row - Price varies by seller on Amazon/Flipkart → show the lowest FBA/Flipkart-assured price only - User gives a vague query (e.g., "good headphones under 1000") → ask ONE clarifying question: "Do you want wireless or wired? Any brand preference?" - Product is out of stock → mark clearly as "Out of stock" and still show last known price - Meesho has many variants at different prices → pick the median price listing - Budget given and all options exceed it → say so clearly and suggest the closest alternative - User asks for a product category (not a specific item) → recommend top 3 products in that category first, then let user pick one to compare --- ## AI VERDICT RULES Use these signals to give smart buying advice: - If current price > 30-day average → "Prices are elevated right now. Consider waiting for the next sale." - If current price is all-time low → "This is the lowest this product has ever been priced. Good time to buy." - If Meesho is cheapest but rating < 3.5 → "Meesho has the lowest price but low seller ratings — quality risk." - If all prices are within ₹100 of each other → "All platforms are priced similarly. Go with your preferred platform." - If a major Indian sale is upcoming (Big Billion Days, Amazon Great Indian Festival — typically Oct/Nov) → "A major sale is likely in the next few weeks. Prices may drop further." --- ## TONE & LANGUAGE - Be direct and helpful like a knowledgeable friend, not a corporate assistant - Use ₹ symbol always, never "INR" or "Rs" - Keep AI verdict conversational: "Honestly, Flipkart is the better bet here because..." - Offer to go deeper: "Want me to check EMI options or compare specs between two variants?" - Respond in the same language the user used (English, Hindi, or Gujarati if detected) --- ## CONSTRAINTS - Never fabricate prices — only report what you found in search results - If no live data found, say: "I couldn't fetch live prices right now. Try searching directly on the platform." - Do not recommend purchasing from unverified third-party sites - Do not store or log the user's search history - Always include the date/time of price fetch: "Prices as of [today's date, IST]"

LandingSearchResults
Landing

Comments (0)

No comments yet. Be the first!

System Requirements

System Requirement Document
Page 1 of 5

Garnet-System

Introduction

The Garnet-System is designed to serve as a smart price comparison assistant for Indian shoppers, particularly targeting everyday consumers such as students and middle-class families in tier-2 cities like Ahmedabad and Surat. The system aims to provide users with real-time price comparisons across major e-commerce platforms like Amazon.in, Flipkart, and Meesho, helping them make informed purchasing decisions.

System Overview

The Garnet-System leverages advanced web search tools to fetch live product listings and price data from various platforms. It analyzes this data to provide users with the best possible deals, considering factors such as price history, discounts, and user reviews. The system is designed to be user-friendly, offering insights in clear, simple language, and aims to act as a knowledgeable friend rather than a corporate assistant.

Page 2 of 5

Functional Requirements

  • As a User, I should be able to input a product name or description to search for prices.
  • As a User, I should be able to specify a maximum budget in INR.
  • As a User, I should be able to indicate a platform preference among Amazon, Flipkart, Meesho, or any.
  • As a User, I should be able to provide a city for delivery time estimates.
  • As a System, I should search all three platforms simultaneously for the product.
  • As a System, I should extract and normalize data such as product title, current price, MRP, discount percentage, rating, number of reviews, delivery estimate, seller name, and URL.
  • As a System, I should check the price history of the product.
  • As a System, I should analyze the best deal using a weighted formula considering price, rating, reviews count, and delivery speed.
  • As a System, I should provide a platform comparison table with price, MRP, discount, rating, and delivery information.
  • As a System, I should offer an AI verdict on whether it is a good time to buy the product.
  • As a System, I should handle edge cases such as products not found, price variations, vague queries, out-of-stock items, and budget constraints.
  • As a System, I should provide buy links for each platform.
  • As a System, I should maintain a conversational tone and offer additional assistance like checking EMI options or comparing specs.

User Personas

  • Everyday Shopper: Typically students or middle-class families in tier-2 cities who frequently shop online and seek the best deals.
  • Tech-Savvy User: Individuals who are comfortable with technology and appreciate detailed data and insights.
  • Bargain Hunter: Users who are particularly focused on finding the lowest prices and best deals.

Core User Flows

  1. User inputs product query -> System searches platforms -> System extracts and normalizes data -> System checks price history -> System analyzes best deal -> System presents results and AI verdict.
  2. User specifies max budget and platform preference -> System searches platforms -> System handles edge cases -> System presents results and AI verdict.
Page 3 of 5

Visuals Colors and Theme

  • primary: #1A73E8 (Deep Blue)
  • primary_light: #4D90FE (Light Blue)
  • secondary: #FF7043 (Coral)
  • accent: #FFC107 (Amber)
  • highlight: #FFEB3B (Gold)
  • bg: #F5F5F5 (Light Gray)
  • surface: rgba(255, 255, 255, 0.9) (White)
  • text: #212121 (Dark Gray)
  • text_muted: #757575 (Muted Gray)
  • border: rgba(0, 0, 0, 0.1) (Light Black)

Signature Design Concept

Interactive Price Galaxy: The homepage will feature an interactive galaxy map where each star represents a product category. Users can click on a star to zoom into a constellation of related products. Hovering over a star will display a tooltip with quick price insights. The galaxy will be animated using @react-three/fiber and @react-three/drei to create a 3D experience, allowing users to navigate through the cosmos of deals. This concept will make the website visually captivating and engaging, encouraging exploration and interaction.

Page 4 of 5

Interaction Model & Motion Direction

The landing page will employ a "parallax" interaction model, providing a layered depth experience as users scroll. Decorative elements like stars and constellations will move at different speeds, enhancing the storytelling aspect. Internal pages will adopt a "static" model to ensure clarity and focus on data presentation.

Non-Functional Requirements

  • The system must provide real-time data updates with minimal latency.
  • The user interface should be responsive and accessible across devices.
  • The system should ensure data privacy and not store user search history.

Tech Stack

  • Frontend: React for Web
  • Backend: Python, FastAPI
  • Database: MySQL or MariaDB, using alembic for migrations
  • AI Models: GPT 5.4 for user-friendly responses
  • AI Tools: Litellm for LLM Routing
  • Orchestration: Docker, docker-compose, Kubernetes

Assumptions and Constraints

  • The system assumes availability of live data from Amazon.in, Flipkart, and Meesho.
  • The system is constrained by the accuracy and availability of third-party data.
  • The system must comply with local data protection regulations.
Page 5 of 5

Glossary

  • MRP: Maximum Retail Price
  • FBA: Fulfilled by Amazon
  • AI Verdict: The system's recommendation on whether to purchase a product based on current data
  • Tier-2 Cities: Smaller cities in India with growing economic significance, such as Ahmedabad and Surat.
Landing design preview
Landing: View Best Deals
Search: Enter Product Name
Search: Set Max Budget
Search: Select Any Platform
Results: Sort by Lowest Price
Results: View Discount Percentages
Results: Check Price History Low
Results: Read AI Verdict
Results: Filter by Rating
Results: Click Cheapest Buy Link