Model-specific documentation for Dioxus, optimized for AI coding assistants.
AI models have different knowledge cutoffs and training data. This repository provides pruned documentation tailored to each model's knowledge gaps—only what's genuinely new or Dioxus-specific.
Benefits:
- Reduced token usage - Don't waste context on concepts the model already knows
- Focused content - Only novel Dioxus-specific patterns and APIs
- Model-optimized - Each folder is curated for a specific model's blind spots
dioxus-ai-docs/
├── PROMPT.md # Task prompt for new AI models
├── CONTRIBUTING.md # Guidelines for adding models
├── complete/ # Full unedited documentation
│ ├── AGENTS.md # Entry point / overview
│ └── architecture/ # Detailed architecture docs
└── <model-name>/ # Model-specific pruned docs
# Point Claude to your model's folder
claude --add-dir ./claude-opus-4.5
# Or use the complete docs if your model isn't listed
claude --add-dir ./completeAdd the appropriate folder to your project context:
- Copy the model folder into your project, or
- Add it to your IDE's context/reference files setting
Copy-paste the contents of 00-QUICK-REFERENCE.md plus any specific files relevant to your task.
| Model | Tokens | Knowledge Cutoff | Notes |
|---|---|---|---|
| claude-opus-4.5 | 5,555 | May 2025 | Subsecond, WASM-split, Manganis focus |
| complete | ~25,000 | — | Full documentation, no pruning |
To add documentation for a new model:
- Give the AI access to this repo
- Paste
PROMPT.mdas instructions - The model will generate its own pruned folder
See CONTRIBUTING.md for detailed guidelines.
A model's self-assessment of its knowledge may not be perfectly accurate. To help verify:
- Check the metadata - Each model folder lists what was pruned
- Test with real tasks - If the model struggles, it may need more docs
- Compare outputs - Run the same prompt with
complete/vs pruned to check
Documentation sourced from the official Dioxus repository: dioxus/notes/architecture/
MIT