Skip to content

iyawnnn/KodaSync

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

34 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

KodaSync | Neural Code & Knowledge Engine

KodaSync CI Docker Ready License

KodaSync is a professional intelligence hub designed to centralize technical knowledge for software engineers. It combines a high-performance code editor with contextual AI agents to transform scattered code snippets into a searchable, neural knowledge base.

Technical Solution

The platform solves the "lost context" problem in modern development by creating a semantic layer over every saved note. By utilizing high-dimensional vector embeddings, KodaSync understands the underlying intent and logic of your code rather than relying on traditional keyword matching.

Key Functional Areas

  • Neural Chat Interface: An AI-powered workspace using Retrieval Augmented Generation (RAG) to provide accurate, project-specific answers by referencing your personal library.
  • The Studio: A full-featured code editor built with the Monaco Editor engine (the core of VS Code) for a native development experience.
  • Semantic Note Library: A centralized repository where notes are processed in the background to generate technical tags and embeddings automatically.
  • Project Contextualization: Organizes technical data into projects, allowing users to restrict AI agents to specific architectural rules and requirements.
  • Knowledge Scraper: A specialized utility that imports and cleans data directly from technical documentation URLs to expand the neural brain.

Tech Stack

The architecture is built for low latency and high concurrency using an async-first approach.

Frontend

  • Next.js 15 (App Router): High-performance React framework.
  • Tailwind CSS & Shadcn/ui: Modern, responsive design system.
  • Monaco Editor: Professional-grade code editing and syntax highlighting.
  • Axios: Robust HTTP client for API communication.
  • Lucide React & Sonner: Streamlined technical iconography and toast notifications.

Backend

  • FastAPI: A modern, high-performance Python framework used for building the primary API with asynchronous support.
  • SQLModel: Elegant database interaction combining SQLAlchemy and Pydantic.
  • Groq SDK (Llama 3.1/3.3): Powers the AI core with ultra-fast inference for technical reasoning.
  • SlowAPI: Token-bucket rate limiting to ensure system stability.
  • Jose (JWT): Secure, industry-standard authentication and session management.

Database & Cache

  • PostgreSQL with pgvector (Supabase): Relational storage with specialized vector support for semantic retrieval.
  • Redis (Upstash): High-speed caching for AI responses and rate-limiting state.

Core Features

  • RAG Chat System: Automatically retrieves relevant notes from your library to provide context-aware answers.
  • AI Code Actions: Integrated tools to Fix, Document, Scan Security, and Generate Tests instantly from the editor.
  • Semantic Search: Natural language search powered by FastEmbed 384-dimension vectors.
  • URL Scraper: Seamlessly import technical snippets and full documentation pages from the web.
  • Background AI Tasking: Non-blocking tag generation and embedding calculations to keep the UI responsive.
  • Hybrid Authentication: Support for traditional credentials and secure GitHub OAuth integration.
  • Project Isolation: Scope your searches and AI context to specific project boundaries.
  • Real-time Event Bus: Custom UI events ensure the sidebar and library update instantly during modifications.

Deployment & Infrastructure

Designed for consistency across development and production environments.

  • Docker & Docker Compose: Full containerization to ensure environment parity and simplified local setup.
  • GitHub Actions: A robust CI/CD pipeline that executes the full integration test suite on every push.
  • Pytest: Comprehensive testing suite covering API endpoints, authentication flows, and RAG logic.
  • Managed Hosting: Backend deployed on Render (Python 3 runtime) and Frontend on Vercel (Edge network).

About

Neural code knowledge engine powered by AI context search.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published