Overview
DeepCode is an open-source multi-agent code generation platform developed by the HKU Data Intelligence Lab. It automates the conversion of research papers, textual requirements, and documents into runnable, well-structured code, accelerating research reproduction and prototype development.
Key Features
- Paper2Code: automated implementation generation from academic papers.
- Text2Web / Text2Backend: rapid frontend and backend prototyping from text.
- Integrated QA: automated testing and static analysis to improve code reliability.
Use Cases
- Research reproduction: quickly obtain runnable implementations of algorithms.
- Rapid prototyping: scaffold full-stack prototypes from natural language.
- Education: a platform for teaching code generation and reproducibility.
Technical Highlights
- Orchestrated multi-agent engine with dynamic planning and task allocation.
- CodeRAG integration for retrieval-augmented code generation.
- Multiple interfaces: CLI, web UI, and Python package (
deepcode-hku).