image-cache — give it a prompt, get an image

GET  /healthz
GET  /v1/image?prompt=&judge=1&generate=1&model=   image bytes + X-Cache-Status/Layer/Image-Id/Prompt-Key/Latency-Ms/Cost-USD
GET  /v1/image.json?prompt=…                        {cacheStatus, layer, imageId, promptKey, url, prompt, latencyMs, costUsd}
     both also take detail= (generation only) and exclude=<id,id> (never serve these)
GET  /v1/image-set.json?prompt=&count=N             N distinct pictures of one subject {images, missing, costUsd}
POST /v1/resolve {slots:[{id,prompt}], detail, exclude, generate, maxGenerate}   many slots, all distinct
POST /v1/photos {runId, slots:[{id,caption}], generationCap, style, exclude}   202 at once: one link per slot
GET  /v1/photos/<job>/<slot>.jpg                    302 to the picture | 202 pending | 404 failed (public)
GET  /v1/photos/<job>/<slot>.json?wait=             {id, state, layer, imageId, url} (public)
GET  /v1/photos/<job>/status                        {complete, createdAt, slots} (public)
GET  /v1/images/<id>.jpg                            stored image (immutable, JPEG)
POST /v1/illustrations {runId, style, slots:[{id,subject,mood}], generationCap}   202 at once: one link per slot
GET  /v1/illustrations/<job>/<slot>.png             302 to the transparent PNG | 202 pending | 404 failed (public)
GET  /v1/illustrations/<job>/<slot>.json?wait=      {id, state, layer, imageId, url, mood} (public)
GET  /v1/illustrations/<job>/status                 {complete, createdAt, slots} (public)
GET  /v1/illustration-images/<style>/<id>.png       stored illustration (immutable, PNG with alpha)
POST /v1/icon {name, pitch, audience, describe, accent, style, darkStyle, count, judge, letter, motif}
                                                    app icon candidates (never cached); describe: the maker's words
POST /v1/icon/restyle {symbolId | letter, name, text, font, style, accent, darkStyle, motif}   a symbol in another style
                                                    or color, free, content-hashed
POST /v1/icon/appearances {symbolId | letter: {name, text, font, motif}, style, accent, darkStyle, tint} | {icon, tint}
                                                    the icon as phones show it (256 px, masked), free, cached
GET  /v1/icon/letter?name=&accent=&style=&motif=&mirrored=   a Letter icon {iconId, symbolId, urls}, free,
                                                    deterministic
GET  /v1/icons/<id>.png                             one icon output (immutable, PNG, public)
POST /api/query {text, auto_judge}                  candidates + decision (demo/introspection)
POST /api/generate {text, model}                    generate + cache
POST /api/judge {text, id}                          one judge verdict
GET  /api/stats · GET /api/examples