Improve discoverability of up-to-date info for AI assistants

Aperta
  • Qualcosa non va?
#1,581 0 commenti 0 reazioni 2 assegnatari Vedi su GitHub

I maintainer di solito rispondono entro 1 giorno

@barhanc ci sta già lavorando.

Dal 28/10/2026.

Valutazione

Questa issue non è ancora stata valutata.

Descrizione

idea user exp
Problem

When developers ask AI assistants (Claude, ChatGPT, etc.) "what are the best tools for on-device AI in React Native?", the answers about React Native ExecuTorch tend to be outdated. In a recent test, an assistant described RNE based on the v0.8.0 blog post, while the current release is v0.11.0 with much more.

This undersells the library to people using AI for tooling research. Root causes found:

  • Search results are dominated by older content (v0.8.0 blog, Expo / third-party articles); nothing newer outranks them.
  • The landing page (executorch.swmansion.com) names model families only (Llama, etc.); newer models and capabilities aren't mentioned.
  • llms.txt exists and is well structured, but:
    • it does not state the current version: the only versions mentioned are in "Migrating from v0.9.x → v0.10.0", so models may conclude v0.10 is latest;
    • no LLM models are named (the most-searched category);
    • Some model (depth estimation, ColBERT retrieval and text embeddings, etc.) aren't named;
    • there's no one-paragraph positioning vs alternatives.
  • docs.swmansion.com/react-native-executorch/* returns a 302 to executorch.swmansion.com; some fetchers don't follow cross-host redirects.
Proposed steps
  • Add a summary header to llms.txt: current version + date, one-line positioning, and supported backends (XNNPACK / Core ML / MLX / Vulkan). Example:
  # React Native ExecuTorch
  > Latest version: v0.11.0. On-device AI for React Native, built on Meta's ExecuTorch.
  > One library for LLMs/VLMs, computer vision, speech, image generation and local RAG.
  > Backends: XNNPACK (CPU), Core ML and MLX (iOS), Vulkan (Android); multiple quantization schemas.
  • Add a "Supported models (vX.Y)" section to llms.txt with exact names per task, including LLMs/VLMs (with context length, tool calling, constrained decoding, thinking mode), depth (Depth Anything V2/V3, ZipDepth), retrieval (LFM2.5-ColBERT-350M, MiniLM, mpnet) and the rest.
  • Generate that section from the model registry at docs build time so it never goes stale between releases. Ideally also generate the version line from package.json.
  • Add a crawlable "Supported models" page (HTML table: model, task, backends, added-in version, link to benchmark) and link it from the landing page.
  • Update the landing page to list current capabilities and headline models explicitly (not just families), with a visible "Latest: v0.11 — what's new" block.
  • Publish a blog post for every minor release (not only milestones), with descriptive titles, e.g. "React Native ExecuTorch v0.11: depth estimation, inpainting, local RAG with ColBERT". This is what will outrank the v0.8.0 post.
Lingua principale
TypeScript
Stelle
1.8k
Fork
101
Merge medio
2g 5h
PR unite (30g)
57

Preparare l'ambiente

Come iniziare

  1. Leggi tutta la issue e poi la guida ai contributi del progetto.
  2. Commenta sulla issue per dire che te ne occupi tu — evita che due persone facciano lo stesso lavoro.
  3. Fai un fork del repository e lavora su un branch.
  4. Apri una pull request che faccia riferimento al numero della issue.

Altre issue di software-mansion/react-native-executorch

Tutte le issue di software-mansion/react-native-executorch

Issue simili

Altre issue su TypeScript

Ricevi le nuove issue nella tua casella

Un breve riepilogo di issue GitHub adatte ai principianti.