Files
pastilhasandClaude Opus 5 d7f3223b6a qwen-local: the first provider that runs on our own GPU
sd-server (stable-diffusion.cpp, Vulkan) serving Qwen-Image 2.1 on this
machine. It already speaks OpenAI's image API, which is the vocabulary
`ImageRequest` is written in, so there is no translation layer — only the parts
that are true of a local GPU and not of a hosted one.

── keyless is a category, not an exception ──

`credentials.ts` decided a provider was usable by asking whether a key existed,
so a service that takes none read as unconfigured forever. The fix is an
adapter declaring `keyless`, not a dummy secret: a dummy would record a
credential that does not exist and make the secret store lie about what is
configured, and that store is the one place meant to be the truth about them.

A keyless adapter never reaches `keyFor` at all. That function's own comment
says why it uses `getSecret` rather than `getKey` — so a missing FOREIGN
credential is never minted — and a provider with no credential by design has no
business in that path.

whisper, OCR and TTS already run on this machine, so this will not be the last.

── measured, not assumed ──

The integration doc said "tens of seconds, not two". The server's own log says
512x512 completes in ~185s and 1024x1024 samples at ~41.4s/step over 20 steps,
about 14 minutes. And it serialises strictly: a second caller waits for the
first to finish entirely, so a queued 512 can wait 17 minutes before a byte
comes back.

So: 512 is the default and the substitute for `auto`, 1024 is offered with the
cost in its label, and `timeout: false` is a requirement rather than caution —
every default anyone would reach for is far under 17 minutes. The caller's
`signal` still aborts, because a user cancelling is not a client giving up.

── the model id moved, because only it could ──

`GET /v1/models` returns `sd-cpp-local`, hardcoded in sd-server with no
override. The catalogue now agrees with the server and carries "Qwen-Image 2.1
(local)" in the label, so requests name a model the server knows while the UI
still says something meaningful.

── not shipped: edit() ──

Same GPU, same queue. It doubles the ways generation can be blocked while the
generate path is unproven in anger, and `ImageAdapter` makes absence the way to
say so — the route answers 400 instead of the queue finding out late.

Verified: /v1/models answers, provider statuses report qwen-local as
configured/keyless/usable, and the shipped entry survives the merge with the
stored catalogue.json. NOT verified end to end — a generation is occupying the
GPU and I left it alone.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Officer-Session: de3ed4e6-f88a-429c-94e6-a760336ba83b
2026-09-21 13:00:05 +01:00

187 lines
7.6 KiB
TypeScript

import { getHeaders, useClient } from 'hooks/useClient';
// Every call the panels make to this plugin's own sidecar, through `/api/atelier`.
//
// `useClient('/api/atelier')` gives typed verbs for JSON; `fileUrl` below is the one thing a client
// call cannot do — put a bearer token on an `<img src>`.
const ATELIER_API = '/api/atelier';
/**
* A provider: a vendor Officer curates, and whether we hold a key for it.
*
* Every provider in the catalogue is listed, set up or not — a vendor you have no key for is exactly
* what you need to see in order to add one. There is no "configured provider" as a separate thing;
* `configured` is a fact about the key, and the model is chosen per generation.
*/
export type ProviderStatus = {
name: string;
label: string;
configured: boolean;
/**
* This provider authenticates with nothing — a service of ours rather than a vendor's. It is
* `configured` by existing, so anywhere that offers to add or change a key must skip it.
*/
keyless: boolean;
key_source: 'store' | 'environment' | 'keyless' | null;
/** Null for a keyless provider: naming a variable nobody can usefully set would be a wrong hint. */
env_var: string | null;
/** False when the catalogue lists a vendor this build has no adapter for. */
usable: boolean;
supports_edit: boolean;
refreshable: boolean;
default_model: string | null;
models: string[];
model_details: CatalogueModel[];
updated_at: string | null;
};
/** One model in the catalogue — what it makes, what it accepts, what it costs. */
export type CatalogueModel = {
id: string;
label?: string;
kind: 'image' | 'video';
sizes?: string[];
qualities?: string[];
transparency?: boolean;
edit?: boolean;
pricing?: Record<string, unknown>;
status?: 'current' | 'deprecated';
source?: 'shipped' | 'api' | 'scraped';
};
/** Snake case, matching the server — one spelling per field across every endpoint. */
export type GenerationRow = {
id: string;
request_id: string;
index: number;
operation: 'generate' | 'edit';
status: 'queued' | 'running' | 'succeeded' | 'failed';
provider: string;
model: string;
prompt: string;
options: Record<string, unknown> | null;
input_images: string[] | null;
mask_image: string | null;
image_path: string | null;
url: string | null;
width: number | null;
height: number | null;
has_alpha: boolean | null;
opaque: boolean | null;
cost_usd: string | null;
duration_ms: number | null;
error: string | null;
created_at: string;
};
export type JobResponse = {
id?: string;
status?: string;
created?: number;
data?: { url: string; revised_prompt?: string; b64_json?: string }[];
atelier?: {
id: string;
status: 'queued' | 'running' | 'succeeded' | 'failed' | 'partial';
cost_usd: string | null;
duration_ms: number;
images: { id: string; status: string; url: string | null; error: string | null }[];
};
error?: { message: string; param?: string | null };
};
export type Collection = {
id: string;
name: string;
description: string | null;
created_at: string;
updated_at: string;
};
export type CollectionSummary = Collection & { item_count: number; cover_image_path: string | null; cover_url: string | null };
export type CollectionDetail = Collection & { images: GenerationRow[] };
/**
* `row.url` (e.g. `/files/<name>`) with this session's bearer token appended, for an `<img src>`.
*
* Browsers cannot set an `Authorization` header on an `<img>` request, so the token travels in the
* query string instead — the same escape hatch `FileViewer/file-types.ts` and the music plugin's
* `MusicPlayerHost` already use for exactly this reason (`permission-gate.ts` and `user-middleware.ts`
* both read `?token=` as well as the header).
*/
export function fileUrl(path: string | null): string | null {
if (!path) return null;
const token = getHeaders()['Authorization']?.replace('Bearer ', '') ?? '';
return `${ATELIER_API}${path}?token=${encodeURIComponent(token)}`;
}
export function useAtelierClient() {
const client = useClient(ATELIER_API);
return {
...client,
listProviders: () => client.get<{ data: ProviderStatus[] }>('providers').then((r) => r.data),
/** Save a provider's key. Goes to the platform's secret store; nothing ever reads it back. */
setProviderKey: (name: string, apiKey: string) =>
client.put<{ data: ProviderStatus[] }>(`providers/${name}/key`, { apiKey }).then((r) => r.data),
/** Forget a stored key. Refused for a key that came from the environment — that one is the install's. */
clearProviderKey: (name: string) =>
client.delete<{ data: ProviderStatus[] }>(`providers/${name}/key`).then((r) => r.data),
/**
* Re-read one provider's models and prices from the vendor, or every provider when omitted.
*
* The only thing that updates the catalogue — there is no schedule. It runs a live fetch, so it is
* as slow as the vendor is and the caller should show that it is working.
*/
refreshCatalogue: (provider?: string) =>
client
.post<{ data: { updated_at: string | null; providers: ProviderStatus[]; changed: boolean } }>(
`providers/catalogue/refresh${provider ? `?provider=${encodeURIComponent(provider)}` : ''}`,
{},
)
.then((r) => r.data),
listGenerations: (limit = 60) => client.get<{ data: GenerationRow[] }>(`images?limit=${limit}`).then((r) => r.data),
startGeneration: (body: Record<string, unknown>) => client.post<{ id: string }>('images/generations', body),
/** Long-polls up to `waitSeconds` server-side, so the panel does not run its own retry loop. */
awaitGeneration: (id: string, waitSeconds = 30) => client.get<JobResponse>(`images/generations/${id}?wait=${waitSeconds}`),
/** A provider row is adapter + model; the key is set separately, and never comes back. */
createProvider: (adapter: string, model: string) =>
client.post<{ data: ProviderStatus[] }>('providers', { adapter, model }).then((r) => r.data),
updateProvider: (id: string, values: Record<string, unknown>) =>
client.patch<{ data: ProviderStatus[] }>(`providers/${id}`, values).then((r) => r.data),
deleteProvider: (id: string) => client.delete<{ data: ProviderStatus[] }>(`providers/${id}`).then((r) => r.data),
listCollections: () => client.get<{ data: CollectionSummary[] }>('collections').then((r) => r.data),
getCollection: (id: string) => client.get<{ data: CollectionDetail }>(`collections/${id}`).then((r) => r.data),
createCollection: (name: string) =>
client.post<{ data: Collection }>('collections', { name }).then((r) => r.data),
renameCollection: (id: string, name: string) =>
client.patch<{ data: Collection }>(`collections/${id}`, { name }).then((r) => r.data),
/**
* `deleteImages` is the difference between filing something wrongly and regretting the whole
* afternoon — the server keeps the images unless it is asked, by name, not to.
*/
deleteCollection: (id: string, deleteImages = false) =>
client
.delete<{ data: { id: string; deleted: boolean; deletedImages: number } }>(
`collections/${id}${deleteImages ? '?images=delete' : ''}`,
)
.then((r) => r.data),
addToCollection: (id: string, generationIds: string[]) =>
client
.post<{ data: { added: number; images: GenerationRow[] } }>(`collections/${id}/items`, { generationIds })
.then((r) => r.data),
removeFromCollection: (id: string, generationId: string) =>
client.delete<{ data: { images: GenerationRow[] } }>(`collections/${id}/items/${generationId}`).then((r) => r.data),
};
}
export type UseAtelierClientType = ReturnType<typeof useAtelierClient>;