Denigma AI

Analyzes selected code snippets using machine learning models to generate natural language explanations of functionality, logic flow, and purpose. Integrates with VS Code's editor context to identify code boundaries and syntax, then sends parsed code to Denigma's backend ML service which returns human-readable explanations rendered inline or in a side panel. The system maintains language-agnostic parsing to handle multiple programming languages.

Detects the programming language of selected code using VS Code's language mode detection and syntax highlighting metadata, then routes the code to language-specific ML explanation pipelines. The backend maintains separate trained models or prompt templates optimized for each language's idioms, libraries, and common patterns, ensuring explanations reference language-specific conventions and best practices.

Registers custom keybindings in VS Code (e.g., Ctrl+Alt+E or Cmd+Shift+D) that capture the current editor selection or cursor position, extract the code context, and trigger explanation generation without requiring menu navigation or mouse interaction. The extension hooks into VS Code's command palette and keybinding system to provide instant, keyboard-driven access to explanations, improving workflow efficiency for power users.

Implements a tiered access model where free users receive a limited number of explanation requests per day/month (likely 5-20 per day), while paid subscribers unlock unlimited or higher-tier access. The extension tracks API usage client-side and enforces rate limits by disabling the explanation button or showing upgrade prompts when limits are exceeded. Backend API keys are tied to user accounts, enabling usage tracking and enforcement across devices.

Sends selected code to Denigma's cloud backend service where trained ML models (likely fine-tuned language models or transformer-based architectures) perform inference to generate explanations. The extension uses asynchronous HTTP requests (likely REST or GraphQL) to avoid blocking the editor UI while waiting for backend responses. Explanations are streamed or returned in chunks, allowing progressive display in the editor as tokens are generated.