React Code Instructions

React Code Instructions is an open-source dataset containing over 74,000 pairs of instructions generated by LLM models, to create complete components or applications in React. The examples cover concrete cases such as LinkedIn clones, iPhone calculators or waiting simulators (Chipotle). Ideal for training or refining front-end code generation models.

Download dataset

Size

74,428 entries, text formats + Parquet (219 MB)

Licence

MIT

Description

‍

React Code Instructions Is a dataset designed to train models to generate React code based on text instructions. The data was generated via powerful models such as LLama 3.1 70B, 405B, and Deepseek Chat V3. Each entry contains a descriptive text instruction and generated source code to produce interactive components and complete applications.

‍

What is this dataset for?

‍

Fine-tuning generative models for front-end React code generation
Creation of AI assistants capable of generating web interfaces on the fly
Rapid prototyping of interfaces based on natural instructions

‍

Can it be enriched or improved?

‍

Yes, this dataset can be completed with TypeScript examples, automated tests, or components with enhanced accessibility. It is also possible to classify the examples according to their complexity or type (forms, dashboards, e-commerce UI...).

‍

🔎 In summary

Criterion	Evaluation
🧩Ease of use	⭐⭐⭐⭐☆ (clear format and immediately usable)
🧼Cleaning required	⭐⭐⭐⭐☆ (low – examples are coherent and React-focused)
🏷️Annotation richness	⭐⭐⭐☆☆ (medium – only instructions and code, no complex metadata)
📜Commercial license	✅ Yes (MIT)
👨‍💻Ideal for beginners	👍 Yes – excellent learning material for junior React developers
🔁Reusable for fine-tuning	🔥 Perfect for code generation models
🌍Cultural diversity	⚠️ Low – standardized technical content in English