About CacheSphere
CacheSphere is a curated language reference with curated models (graduating), preview agent/company/benchmark registries, small hand-curated tools and stacks, and an experimental, language-oriented Advisor. Same MIT JSON for you and your agents — not a substitute for official docs or your own evaluation.
Who runs it
CacheSphere is built by Idaluna Labs. Core catalogue JSON is MIT-licensed; site and source are available on request. Contact: [email protected].
What each tier covers
- Reference (core): 141 languages, compare, transfer,
/api/languages.json— deepest coverage. - Hand-curated: 26 AI/dev tools and 10 stack records — small, manual, reviewable signals.
- Models (graduating): curated model registry — verify on detail pages before citing.
- Preview snapshots: agents, companies, benchmarks — real registry data, uneven depth; verify on detail pages.
- Experimental: language-oriented Advisor, Decision Brief examples, hosted
/api/agent/*when deployed — useful drafts, not authoritative.
What it is not
- Not official language documentation or a compiler spec.
- Not a course, tutorial site, or playground host.
- Not a model host, and not a substitute for your own evaluation.
How records are curated
CacheSphere adds structure between noisy discovery channels and the catalogues humans and agents use. The public pipeline is: discovery (public, ToS-respecting signals only) → verification (official docs and small experiments where sensible) → curation (relevance and minimum data) → catalogue lean records with links out for depth.
Quality and evidence
Flagship languages tend to have richer patterns and tips; niche entries may be basic reference only. Benchmark and proof claims are early evidence — see proof status and benchmark methodology before citing numbers externally.
Licence and self-host
Core catalogue JSON is MIT-licensed and free to fetch or self-host — see
Quickstart and For AI tools & agents.
Static JSON works on any host; full MCP and /v1 source on request.
FAQ · Quickstart · Home