AlgoX is an Explainable AI (XAI) tool that predicts the runtime time complexity of Python and Java algorithms.
Instead of just giving you a Big-O prediction, AlgoX uses SHAP (SHapley Additive exPlanations) to visually highlight the exact lines and tokens of code that influenced the model's decision, right inside your VS Code editor.
🚀 Features
Multi-Language Support: Analyzes both Python and Java algorithms.
Hybrid AI Architecture: Powered by a custom backend fusing AST-extracted static features (depth, branches, recursion) with a fine-tuned transformer model (UniXcoder).
Visual Explainability: Applies color-coded background highlights directly to your code.
🔴 Red Highlights: Tokens that increase the complexity prediction.
🔵 Blue Highlights: Tokens that decrease the complexity prediction.
Hover Insights: Hover over any highlighted token to see its exact SHAP impact score.
7 Complexity Classes: Detects CONSTANT, LINEAR, LOGN, NLOGN, QUADRATIC, CUBIC, and NP.
🛠️ Usage
Open any .py or .java file containing an algorithm.
Press Ctrl + Shift + A (or Cmd + Shift + A on Mac) to run the analysis.
Wait a moment for the AI to process the code. A notification will appear with the predicted complexity and confidence score.
Review the color-coded SHAP highlights in your editor.
To clear the highlights, press Ctrl + Shift + C (or Cmd + Shift + C on Mac).
Alternatively, you can access these commands via the VS Code Command Palette (Ctrl+Shift+P):
AlgoX: Analyze Time Complexity
AlgoX: Clear Highlights
⚙️ Requirements
This extension acts as the frontend client for the AlgoX AI model. To function, it requires the AlgoX FastAPI backend to be running.
By default, the extension points to the local development server (http://127.0.0.1:8000). If you are using the hosted Hugging Face Spaces backend, you will need to update the API URL in the extension source code and recompile.
🐛 Known Issues
Whitespace Highlighting: In rare cases, the BPE tokenizer may group newline characters with syntax, causing the highlight to slightly offset. The built-in source mapper mitigates most of this.
Cold Starts: If the backend API is hosted on a free Hugging Face Space, the first request of the day may take up to 30-60 seconds while the container wakes up.
