PearAI is an open-source AI code editor built as a fork of VS Code, combining AI-powered coding assistants like Continue and Roo Code into a unified development environment. It’s designed for developers who want contextual AI assistance — such as code generation, bug fixing, and project creation — without juggling multiple AI tools or subscriptions. PearAI brings together LLM-powered editing, project scaffolding, and automated coding agents directly inside the editor.
Technically, PearAI is built on Electron.js and TypeScript for the frontend, with a Python FastAPI backend and Supabase for authentication and data storage. The AI capabilities are powered by Continue (chat) and Roo Code (agent/creator), while the PearAI Router dynamically selects the top-performing LLMs (GPT-4o, Claude 3 Opus, Llama 3.1) based on task performance. It supports hot module reload for React components and requires a specific toolchain including Node.js 20.18.0, Rust, and a C/C++ compiler.
What You Get
- PearAI Router - Automatically selects the highest-performing LLM (GPT-4o, Claude 3 Opus, Llama 3.1) for your coding task based on real-time performance metrics, eliminating the need to manually switch between AI providers.
- PearAI Chat - Powered by Continue, enables contextual code edits and conversational AI assistance directly within your codebase, understanding file context and project structure.
- PearAI Agent - An AI coding agent powered by Roo Code/Cline that can automatically generate features and fix bugs without manual prompting.
- PearAI Creator - (Coming Soon) Scaffolds new projects from scratch using expert templates and modern best practices, ensuring production-ready code from the start.
- Integrated VS Code Fork - Full VS Code compatibility with enhanced AI features, including syntax highlighting, debugging, and extensions — all with AI augmentation built-in.
- Supabase Auth & FastAPI Backend - Secure, scalable authentication and server-side logic for user accounts, telemetry, and optional cloud-based AI processing without requiring personal API keys.
Common Use Cases
- Switching from multiple AI subscriptions - A developer cancels separate ChatGPT, Claude, and GitHub Copilot subscriptions and uses PearAI’s Router to access top LLMs through one unified interface, saving $30+/month.
- Learning a new programming language - A developer with no prior Swift experience uses PearAI’s contextual code suggestions and agent to build an iOS app in under a month, accelerating their learning curve.
- Rapid prototyping of web apps - A solo founder uses PearAI Creator (beta) to generate a full-stack web application with React, Node.js, and Supabase from a single prompt, reducing setup time from hours to minutes.
- Debugging complex codebases - A Big Tech engineer uses PearAI Agent to automatically identify and fix bugs in legacy Python and Hack codebases without needing deep domain expertise.
Under The Hood
Architecture
- Clear separation of concerns with distinct modules for business logic, API endpoints, and domain services, following a layered architecture
- Dependency injection implemented via a custom container with constructor injection and named bindings
- Strategy and factory patterns used to dynamically select processing algorithms based on input type
- Event-driven communication through a centralized dispatcher enabling loose coupling between components
- Modular plugin system with interfaces and loaders supporting runtime extensibility without core modifications
- Repository pattern applied to abstract data access layers for multiple storage backends
Tech Stack
- Python 3.9+ backend with FastAPI and dependency injection framework
- React 18 frontend with TypeScript, Redux Toolkit, and Vite for responsive UI development
- PostgreSQL with SQLAlchemy ORM and Alembic for schema management
- Docker Compose orchestrating Redis caching and Celery for asynchronous processing
- Pytest with factory and mocking utilities for testing
- GitHub Actions automating testing and containerized deployment pipelines
Code Quality
- Limited testing with minimal test functions and absent assertions, resulting in poor behavior validation
- Inadequate error handling with no custom exceptions or structured recovery mechanisms
- Inconsistent naming conventions across modules without clear style guidelines
- Absence of type annotations and static type checking, increasing runtime error risk
- No linting or code quality enforcement tools, leading to unstandardized and hard-to-maintain code
- Fragmented code organization with blurred module boundaries and weak encapsulation
What Makes It Unique
- Hybrid reasoning engine that intelligently switches between symbolic logic and neural inference based on query complexity
- Self-optimizing prompt pipeline that auto-refines instruction templates using user feedback loops
- Embedded knowledge graph builder that extracts semantic relationships from unstructured data with confidence-weighted edges
- Decentralized model versioning using IPFS with cryptographic integrity for trustless model distribution
- Real-time collaborative annotation with semantic intent-based conflict resolution
- Native multimodal input parser that unifies text, audio, and visual data into a single reasoning graph