🤖 Local AI-Powered Intelligence

Code-y integrates powerful AI capabilities directly into your local development environment, enhancing your understanding and documentation process without compromising privacy.

Flexible AI Backend Options

You have control over how AI processing is handled:

  • Ollama Integration (Recommended): This is the cornerstone of Code-y's local-first AI.
    • 100% Local Processing: All AI tasks run on your machine using Ollama.
    • Multiple Model Support: Compatible with a wide range of open-source models available through Ollama, such as Gemma, Llama, CodeLlama, Mistral, and more.
    • Automatic Model Management: Code-y can assist with managing and ensuring the correct Ollama models are available.
    • Offline Capability: Once Ollama and models are set up, no internet connection is required for AI features.
  • OpenAI Integration (Optional): If you prefer, or for specific models, you can configure Code-y to use the OpenAI API.
    • Requires your own OpenAI API key.
    • Processing occurs on OpenAI's servers.
  • LangFlow Workflows (Experimental): For advanced users, Code-y offers integration with LangFlow, allowing you to design and implement custom AI pipelines and workflows for documentation and analysis tasks.

Advanced Vector Similarity System

Code-y builds a semantic understanding of your code using vector embeddings, enabling powerful search and discovery features:

  • Semantic Code Search: Find code snippets based on their meaning and functionality, not just exact keyword matches. For example, search for "function to sort users by name."
  • Local FAISS Vector Database: Utilizes FAISS, a high-performance library for efficient similarity search, to store and query code embeddings locally. Embeddings are typically 768-dimensional.
  • Nomic Embed Text Model: By default, Code-y uses the `nomic-embed-text` model (via Ollama if local), which is specialized for generating high-quality embeddings for code and technical text.
  • Cosine Similarity Scoring: Measures the similarity between code snippets with precision. Thresholds can be configured to fine-tune similarity detection.
  • Real-time Vector Generation: Embeddings are automatically generated or updated during the code analysis process for new or modified files.

AI-Generated Documentation

Let AI assist in the often tedious task of writing documentation:

  • Intelligent Descriptions: The AI generates context-aware, natural language explanations for your React components, functions, and class methods.
  • Prop Documentation: For React components, Code-y's AI can automatically generate descriptions for each prop, explaining its purpose and expected type.
  • Code Explanation: Get plain English explanations of complex or obscure code logic, helping both new and experienced developers understand the codebase faster.
  • Smart Caching: AI-generated content is cached. Descriptions are only regenerated for code sections that have actually changed, saving time and processing resources.