{"service":"ARA Advisor","what":"Declare a machine's hardware; get its safe local-AI memory budget and which models fit, with estimated usable context.","skill":"GET /skill.md","endpoints":["POST /assess — machine facts -> memory wall, safe budget, recommended engine","GET /fit?model=<hf-id>&ram_gb=<n>[&backend=][&vram_gb=][&quant=] — does that model fit that machine","GET /recommend?ram_gb=<n>[&backend=][&vram_gb=][&quant=][&limit=] — ranked models that fit","GET /models — the curated catalog this service ranks from"],"upstream":"https://github.com/willsarg/project-ara","honesty":"every figure is an analytic ESTIMATE from declared facts; `ara characterize` on the machine itself is what measures"}