Skip to content

Introduction

Decodie generates structured learning entries as a byproduct of AI-assisted coding sessions, helping developers understand what was built, how, and why.

When you use an AI coding assistant to write or modify code, Decodie analyses the results and produces learning entries that capture:

  • Code patterns with focused snippets
  • Explanations of why a particular approach was chosen
  • Alternatives considered and their trade-offs
  • Key concepts that underpin the code

These entries are stored in a .decodie/ directory alongside your code in a structured JSON format.

  • VS Code Extension — analyse and explain code selections, ask follow-up questions, all within your editor
  • Local UI — a web dashboard that reads your .decodie/ directory and presents entries with filtering, lessons, and Q&A
  • GitHub Action — automatic analysis on every pull request
  • GitHub Bot — on-demand analysis via @decodie-bot in PR comments

Decodie works with any major LLM provider. You bring your own API key from whichever provider you prefer:

ProviderModels
AnthropicClaude Sonnet, Claude Opus, Claude Haiku
OpenAIGPT-4o, GPT-4o mini, o1
GoogleGemini 2.0 Flash, Gemini 2.5 Pro
MistralMistral Large, Mistral Medium
xAIGrok 3, Grok 3 mini
DeepSeekDeepSeek Chat, DeepSeek Coder

See Provider Setup for details on getting an API key and configuring it.