Intellegode

Fight vibe coding. Understand your own code.

Intellegode is a VS Code extension that leverages a local Large Language Model to quiz you on code you just wrote or AI-generated—ensuring deep comprehension before you move on. Everything runs strictly local. No cloud telemetry, no subscriptions.

How It Works

1. Highlight a block of code in your editor.
2. Press Ctrl+Alt+Q (Windows/Linux) or Cmd+Alt+Q (Mac) to trigger Intellegode.
   • Alternatively: Open the Command Palette (Ctrl+Shift+P) and run Intellegode: Quiz Me.
3. Answer the comprehension question in the sidebar.
4. Receive immediate, context-aware feedback from the AI to validate your grasp of the logic.

Prerequisites

• Node.js v18+
• Ollama installed natively or via Docker
• VS Code v1.110+

Installation & Setup

1. Install the Extension Search for Intellegode in the VS Code Extensions Marketplace and select Install.

2. Start Ollama Ensure the Ollama server is running in the background. If you installed Ollama natively:

ollama serve

3. Pull the Language Model Intellegode requires the Qwen model to operate. Run the following command in your terminal:

ollama pull qwen3.5:4b

4. Start Your First Session Highlight a snippet of code in your editor and press Ctrl+Alt+Q to begin your session.

Architecture

• VS Code Extension API (TypeScript)
• Ollama (Local LLM Runtime)
• Qwen3.5 4B (Code Comprehension Model)

Configuration

In VS Code, navigate to File > Preferences > Settings and search for Intellegode to configure:

• intellegode.ollamaUrl: Base URL for the Ollama server (Default: http://localhost:11434)
• intellegode.defaultModel: Target model mapping (Default: qwen3.5:4b)

Developer Environment Variables

When running the extension in development mode, the following flags are supported:

• INTELLEGODE_OLLAMA_FORCE_CPU=1 — Forces CPU-only execution (Beneficial for low-VRAM environments)
• INTELLEGODE_OLLAMA_REQUEST_TIMEOUT_MS=120000 — Configures the request timeout buffer in milliseconds

Troubleshooting

Extension Hangs / System Freezes

Cause: Ollama is attempting to load the language model into GPU VRAM but running out of allocation space. Solution: If your hardware has 4GB VRAM or less, force CPU mode. You can set the environment variable before launching VS Code from the terminal:

export INTELLEGODE_OLLAMA_FORCE_CPU=1

Note: CPU-only inference is inherently slower but remains entirely stable on lower-end systems.

Connection Refused to Ollama

Solution:

1. Verify Ollama is actively running.
2. Confirm Ollama is accessible by running curl http://localhost:11434/api/tags.
3. Check that your intellegode.ollamaUrl setting correctly resolves to the active Ollama host.

Model Not Found Error

Solution: The required model has not been downloaded to the Ollama runtime. Pull the required model:

ollama pull qwen3.5:4b

Request Timeout Error

Cause: The model is sustaining a cold-boot load that exceeds the timeout threshold. Solution: Wait a few seconds for the model to cache into memory and try again. Alternatively, increase the timeout limit via the INTELLEGODE_OLLAMA_REQUEST_TIMEOUT_MS variable.

Roadmap

• [ ] v1 — Active Quizzer (Highlight -> Question -> Feedback)
• [ ] v2 — Concept Debt Tracker
• [ ] v3 — Project Ownership Mapping
• [ ] v4 — Reconstruction Challenges