For LLM coding assistants

Weaverlet ships first-class context files for AI coding assistants. If you're writing Weaverlet code with help from Claude, GPT, Codex, Cursor, or Aider, point your assistant at the resources below and it will produce idiomatic Weaverlet code instead of guessing.

What's available

File	Purpose	Convention
ReadMe.LLM.md	Full LLM-targeted context: framework rules, mental model, 14 worked examples, API signatures, common patterns.	ReadMe.LLM
llms.txt	Short index that points at ReadMe.LLM.md plus the canonical docs. Designed for auto-discovery.	llmstxt.org

Both live at the root of the observatoriogeo/weaverlet repo. The docs you're reading now are the human-targeted long form of the same material; the LLM files are a compressed version optimized for being pasted (or fetched) into an assistant's prompt.

CDN URLs for ReadMe.LLM.md

The repo file works directly, but the GitHub raw URL has rate limits and a CORS profile that some assistants choke on. The jsDelivr mirror is the recommended fetch target:

# Latest (tracks main, updated as Weaverlet evolves):
https://cdn.jsdelivr.net/gh/observatoriogeo/weaverlet@main/ReadMe.LLM.md

# Pinned to v0.3.0 (stable, won't shift under your assistant):
https://cdn.jsdelivr.net/gh/observatoriogeo/weaverlet@v0.3.0/ReadMe.LLM.md

Use the pinned URL when you want reproducible assistant behavior across runs. Use the @main URL when you want the assistant to track the project as it ships new features.

Drop-in prompts

Paste one of these at the start of your conversation, before describing the task you want help with.

Claude (claude.ai, Claude Code, API)

Before answering, fetch and read:
https://cdn.jsdelivr.net/gh/observatoriogeo/weaverlet@main/ReadMe.LLM.md

That document is the canonical reference for Weaverlet. Follow its rules
and patterns when writing code.

Claude Code can also auto-discover llms.txt if you point it at the repo or include the URL in CLAUDE.md.

GPT / ChatGPT / Codex

For this conversation, the authoritative reference for Weaverlet is:
https://cdn.jsdelivr.net/gh/observatoriogeo/weaverlet@main/ReadMe.LLM.md

Fetch it (if you have browsing) or assume its content when writing
Weaverlet code. Follow the rules and patterns it documents.

Cursor / Continue / other IDE assistants

Add the CDN URL to your project's docs configuration so the assistant fetches it on each session. Cursor's @Docs feature, for example, accepts the URL directly.

llms.txt auto-discovery

Tools that follow the llmstxt.org convention look for /llms.txt at the root of a project to discover its LLM-targeted resources. Weaverlet's llms.txt lists ReadMe.LLM.md plus the canonical docs entry points so any conforming assistant gets the right context with zero configuration.

If you maintain a project that depends on Weaverlet, you can do the same: drop your own llms.txt at your project root that links back to ours.

What's in ReadMe.LLM.md

The file is organized for an assistant to skim once and then refer back to as needed:

• Rules: hard constraints (e.g. never use string IDs, always use Identifier() descriptors; callbacks use dash_extensions.enrich, not dash.dependencies).
• Mental model: the DAG-of-components view, the wiring lifecycle, and how WeaverletApp walks the tree.
• 14 worked examples: the same gallery as the Examples section, but condensed into copy-pasteable scripts with framing comments aimed at an LLM.
• API signatures: every public class and helper with its constructor signature and the most-common usage shape.
• Patterns: idioms the framework expects (constructor-injected state, signal-based events, keep_mounted for WebGL preservation, AuthRouterComponent for session-gated routes).

Together that's enough context for an assistant to produce code that compiles against current Weaverlet without hallucinating the API surface.