Project Name: LogicLens – The Deep Logic CS Tutor
Category: Education / AI Native Application
Project Description LogicLens is an AI-native educational platform designed to win the ERNIE AI Developer Challenge by leveraging Baidu's latest Deep Reasoning (ERNIE X1) and Multimodal capabilities. Instead of acting as a "code fixer," it acts as a Socratic Tutor, simulating the computer's memory to show students why their algorithms failed.
The Core Concept Current AI tools (like Copilot) are "Homework Solvers"—they fix code instantly, robbing students of learning. LogicLens is a "Thinking Partner." The Scenario: A student writes a reverse_linked_list function but forgets to save the next node before breaking the link.
The LogicLens Way: Reasoning: It mentally traces the code and finds the exact line where the pointer is lost. Socratic Question: "You just overwrote curr.next. How will you access the rest of the list now?" Visual Snapshot: It generates a diagram showing the "rest of the list" floating in red, labeled "Orphaned Memory," visually proving the error.
Why It Wins the Challenge This project is strategically engineered to check every box for the judges: Deep Reasoning: Uses ERNIE X1 to perform invisible "Mental Traces" of code execution. This proves the model can "think" through logic, not just predict text. Multimodal Tech: Uses ERNIE 4.5 to convert text-based code states into Visual JSON, creating dynamic diagrams.
Social Value: Solves a massive problem in STEM education (rote memorization vs. true understanding).
Key Features The "Dry Run" Engine: The AI creates a "Trace Table" in its mind, tracking variable values step-by-step until the logic breaks. The "Failure State Snapshot": The system generates a visual representation of the data structure at the moment of the crash (e.g., an array index pointing outside the box). No-Code Guardrails: The system is strictly prompted to never give the answer, gamifying the debugging process.
Technical Architecture AI Engine: ERNIE X1 (Reasoning) & ERNIE 4.5 (Visualization) via Baidu Qianfan API. Backend: Python (FastAPI) manages the "Meta-Prompt" and session logic. Frontend: React + React Flow renders the interactive graphs from the JSON data.
Log in or sign up for Devpost to join the conversation.