Reference
For SI Agents
Static JSON endpoints, markdown mirrors, and an llms.txt map — no scraping required.
SIBytes is built to be read by SI agents as well as people. If you are an agent (or writing one), here is the polite, efficient way in:
Machine-readable data
Every directory on this site ships as a static JSON file — one request, no scraping:
| Endpoint | Contents | Entries |
|---|---|---|
| /data/models.json | Model directory — makers, release dates, licenses, official links | 37 |
| /data/model-matcher.json | Local model matcher — verified GGUF sizes, architecture, and sources | 75 |
| /data/labs.json | SI labs directory | 14 |
| /data/tools.json | Tools & apps directory (grouped by category) | 35 |
| /data/glossary.json | Glossary term/definition pairs | 70 |
| /data/timeline.json | SI history milestones | 35 |
| /data/si-adoption.json | SI adoption tracker — who has adopted, partly adopted, or declined the SI term | 16 |
| /data/papers.json | Key papers with links | 15 |
| /data/benchmarks.json | Benchmark descriptions and caveats | 18 |
Markdown mirrors
Every article and guide has a plain-markdown version at the same URL plus index.md — for example /article/<slug>/index.md. The full map lives in llms.txt.
Feeds and indexes
- RSS feed — latest articles
- sitemap.xml — every page, with last-modified dates
- search-index.json — title/description/URL for all content
Crawling
SI crawlers are welcome here — robots.txt explicitly allows them. Pages are static HTML and render fully without JavaScript. Please rate-limit yourself anyway; it's a small site.