SwiggyMind 🧠

A Context-Aware Reasoning Engine for Hyper-Personalized Food Discovery.

Built for the Swiggy Builders Club, SwiggyMind moves beyond simple keyword search. It implements an intelligence layer that understands user intent, reasons through options, and surfaces recommendations with visible logic.

---

🚀 What Makes It Different?

SwiggyMind is not just a chat wrapper. It's a Decision Engine built on two core technical pillars that solve the "choice paralysis" of modern food apps:

1. The Deterministic Ranking Layer (Mind Engine)

Most AI apps pass a raw query to an LLM and hope for the best. SwiggyMind uses a Weighted Ranking Algorithm before the AI even sees the results.

• How it works: It parses natural language into structured signals (Diet, Budget, Speed, Spice) and scores restaurant candidates against these constraints.
• Visible Reasoning: Every card shows the exact "Win Condition": Matches Spicy · Under ₹200 · Rated 4.7★ · 18 min delivery.

2. The Mind Cache (Stateful Session Memory)

SwiggyMind remembers context across turns. It doesn't treat every message as a new search; it understands Refinement.

• The Turn Logic: Ask for "something spicy" and then follow up with "make it veg." The engine merges these constraints using stateful session memory.
• Tangible UI: When stateful reasoning is active, the app displays a Refined from last search badge, proving the engine is listening and adjusting.

---

🎯 Real-World Scenarios

The "Late Night Budget" Constraint

• User: "Hungry, something spicy but under ₹200, fast delivery please."
• Result: Spicy Paneer Wrap + 18 min delivery → Matches Spicy · Under ₹200.

The "Health-Conscious Discovery"

• User: "Show me some high-protein veg options."
• Follow-up: "make it quick."
• Result: Paneer Tikka Salad + 12 min delivery → Refined · High Protein.

---

🛠 Engineering Highlights

• Neural Intent Parser: Uses LLM to extract high-fidelity JSON (Mood, Spice Level, Occasion) from natural language.
• Robust Fallback Strategy: 3-layer execution stack (Cloud LLM → Rule-Based → Heuristic) ensures the app works offline or when API limits are reached.
• Kotlin Multiplatform (KMP): 100% of the Mind Engine logic is shared between platforms.
• Structured Discovery: Enforced JSON schemas ensure zero-hallucination results.

---

_{Built for Swiggy Builders Club · Not an official Swiggy product}