# Run AI Here

> Community local AI deployment and performance-report references. Source-reported, not site-reproduced.

- [HTML collection](https://runaihere.com/)
- [JSON index](https://runaihere.com/catalog.json)
- [Markdown index](https://runaihere.com/catalog.md)
- [Collection method](https://runaihere.com/benchmarks/methodology/)
- [JSON Schema v3.1](https://runaihere.com/schema/community-catalog-3.1.0.json)
- [Change list](https://runaihere.com/changes.json)

Fetch catalog.json once and filter kind, model_family, runtime_topics and hardware_topics locally. Document, weight and executed runtime revisions are separate identities. Unknown conditions are null. Source topics do not certify compatibility. Selected source figures are in editorial.reported_results with individual conditions, units and evidence URLs; metrics is the null legacy site-measurement field.

## Explanatory articles

- [Reviewed article collection JSON](https://runaihere.com/articles.json)
- [Reviewed article collection Markdown](https://runaihere.com/articles.md)

These explain source evidence; they are not additional independent experiments. Preserve body_markdown, sources, facts, reported_results and per-result conditions together.

## Field guide and local queries

- [Evidence fields and query definitions](https://runaihere.com/for-ai/index.json)
- [Field guide Markdown](https://runaihere.com/for-ai/index.md)
- [Executed query examples and expected JSON](https://runaihere.com/downloads/evidence-query-examples.json)
- [Read-only Node.js query examples](https://runaihere.com/downloads/query-evidence.mjs)

Fetch the required static files once; query parameters do not filter on the server. Null is unknown; false and zero are known values. Keep result conditions, source document identities, raw observations and derived summaries together.