till/klarbild
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat: compose + text-to-image generation modes (multi-image + free-text)

Browse Source 
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XNQ8ghPfzAfsyVYd6HgFb6
This commit is contained in:
Till Heidrich
2026-07-23 19:19:39 +00:00
parent 6d58c6c8fd
commit 044eb7f1ae
7 changed files with 172 additions and 35 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/lib/openrouter.ts
+4 -2
View File
@@ -30,7 +30,8 @@ function friendlyFor(status: number): string {
export interface GenerateOpts {
model: string;
prompt: string;
inputUrl?: string; // vorsignierte URL des Originals
inputUrl?: string; // eine Vorlage (Data-URL) — Kurzform
inputUrls?: string[]; // mehrere Vorlagen (compose) — bis zu 14/16 je Modell
aspectRatio?: string; // "16:9", "3:4" …
resolution?: string; // "2K" | "4K" — NICHT zusammen mit expliziten Pixeln
background?: 'transparent' | 'opaque';
@@ -56,7 +57,8 @@ export async function generateImage(opts: GenerateOpts): Promise<GenerateResult>
n: 1,
output_format: opts.outputFormat || 'png',
};
if (opts.inputUrl) body.input_references = [{ type: 'image_url', image_url: { url: opts.inputUrl } }];
const refs = opts.inputUrls?.length ? opts.inputUrls : (opts.inputUrl ? [opts.inputUrl] : []);
if (refs.length) body.input_references = refs.map((url) => ({ type: 'image_url', image_url: { url } }));
if (opts.aspectRatio) body.aspect_ratio = opts.aspectRatio;
if (opts.resolution && !opts.aspectRatio) body.resolution = opts.resolution; // nie beides
if (opts.background) body.background = opts.background;
src/lib/process.ts
+72 -21
View File
@@ -1,13 +1,15 @@
import { one, query } from './db';
import { getObject, putObject, resultKey } from './storage';
import { getObject, putObject, deleteObject, resultKey, thumbKey } from './storage';
import { generateImage } from './openrouter';
import { buildPrompt, type Task } from './prompts';
import { buildPrompt, buildComposePrompt, buildGeneratePrompt, type Task } from './prompts';
import { resolveDimensions, aspectRatio, type Orientation } from './format';
import { finalizeToFormat, stickerContour, buildResultFilename, formatToken, type CropMode } from './pipeline';
import sharp from 'sharp';
const DEFAULT_MODEL = 'google/gemini-3.1-flash-image';
type Mode = 'each' | 'compose' | 'generate';
interface RecipeSnapshot {
tasks: Task[];
output_format?: string;
@@ -17,6 +19,7 @@ interface RecipeSnapshot {
contour_mm?: number | null;
model_key?: string | null;
custom_instruction?: string | null;
prompt_text?: string | null; // Beschreibung für compose/generate
delivery?: 'library' | 'picdrop' | 'both';
picdrop_gallery?: string | null;
}
@@ -42,47 +45,70 @@ async function toDataUrl(buf: Buffer): Promise<string> {
return `data:image/jpeg;base64,${small.toString('base64')}`;
}
/** Kleines Vorschaubild (webp) für die Bibliothek — spart Bandbreite & Speicher. */
async function makeThumb(buf: Buffer): Promise<Buffer> {
return sharp(buf, { failOn: 'none' })
.resize(600, 600, { fit: 'inside', withoutEnlargement: true })
.webp({ quality: 72 })
.toBuffer();
}
export interface ProcessResult { ok: boolean; cost: number; error?: string; }
/** Verarbeitet eine Position vollständig und aktualisiert ihren Datensatz. */
export async function processItem(itemId: string): Promise<ProcessResult> {
const item = await one<{ id: string; job_id: string; source_path: string; filename: string | null }>(
'SELECT id, job_id, source_path, filename FROM items WHERE id=$1', [itemId]);
const item = await one<{ id: string; job_id: string; source_path: string | null; source_paths: string[] | null; filename: string | null }>(
'SELECT id, job_id, source_path, source_paths, filename FROM items WHERE id=$1', [itemId]);
if (!item) return { ok: false, cost: 0, error: 'Position nicht gefunden' };
const job = await one<{ recipe_snapshot: RecipeSnapshot }>(
'SELECT recipe_snapshot FROM jobs WHERE id=$1', [item.job_id]);
const job = await one<{ recipe_snapshot: RecipeSnapshot; mode: Mode }>(
'SELECT recipe_snapshot, mode FROM jobs WHERE id=$1', [item.job_id]);
const r = (job?.recipe_snapshot || {}) as RecipeSnapshot;
const mode: Mode = job?.mode || 'each';
const tasks = r.tasks || [];
const dpi = r.dpi ?? 300;
const cropMode: CropMode = r.crop_mode === 'extend' ? 'extend' : 'crop';
const wantCutout = tasks.includes('cutout');
const wantContour = tasks.includes('contour');
const description = (r.prompt_text || '').trim();
await query(`UPDATE items SET status='running', attempts=attempts+1 WHERE id=$1`, [itemId]);
try {
const source = await getObject(item.source_path);
const target = r.tasks.includes('format') && r.output_format
// Quellen je Modus einsammeln
const sourceKeys: string[] =
mode === 'compose' ? (item.source_paths || []).filter(Boolean)
: mode === 'generate' ? []
: (item.source_path ? [item.source_path] : []);
const sources = await Promise.all(sourceKeys.map((k) => getObject(k)));
const target = tasks.includes('format') && r.output_format
? resolveDimensions({ format: r.output_format, orientation: r.orientation, dpi })
: null;
const model = await pickModel(r.model_key);
const prompt = buildPrompt({ tasks, cropMode, customInstruction: r.custom_instruction });
// Nur Modell aufrufen, wenn eine generative Aufgabe dabei ist.
const needsModel = tasks.includes('clean') || wantCutout ||
(tasks.includes('format') && cropMode === 'extend') || !!r.custom_instruction;
// Prompt + ob das Modell überhaupt gebraucht wird, je Modus.
let prompt: string;
let needsModel: boolean;
if (mode === 'compose') { prompt = buildComposePrompt(description, cropMode === 'extend'); needsModel = true; }
else if (mode === 'generate') { prompt = buildGeneratePrompt(description); needsModel = true; }
else {
prompt = buildPrompt({ tasks, cropMode, customInstruction: r.custom_instruction });
needsModel = tasks.includes('clean') || wantCutout ||
(tasks.includes('format') && cropMode === 'extend') || !!r.custom_instruction;
}
let working = source;
let working = sources[0] || Buffer.alloc(0);
let modelUsed: string | null = null;
let cost = 0;
if (needsModel) {
const inputUrls = sources.length ? await Promise.all(sources.map(toDataUrl)) : undefined;
const gen = await generateImage({
model: model.id,
prompt,
inputUrl: await toDataUrl(source),
inputUrls,
aspectRatio: target ? aspectRatio(target.w, target.h) : undefined,
background: wantCutout ? 'transparent' : undefined,
outputFormat: 'png',
@@ -91,6 +117,7 @@ export async function processItem(itemId: string): Promise<ProcessResult> {
modelUsed = gen.model;
cost = gen.cost;
}
if (!working.length) throw new Error('Keine Bilddaten erzeugt');
// Lokale Nachbearbeitung: exakt aufs Zielmaß + dpi
const fin = await finalizeToFormat(working, target, cropMode, dpi);
@@ -105,27 +132,51 @@ export async function processItem(itemId: string): Promise<ProcessResult> {
const key = resultKey(item.id);
await putObject(key, outBuf, 'image/png');
// Vorschaubild (optional, laut Einstellung)
const cfg = await one<{ make_thumbnails: boolean; keep_sources: boolean; nas_enabled: boolean }>(
'SELECT make_thumbnails, keep_sources, nas_enabled FROM settings WHERE id=1');
let thumbPath: string | null = null;
if (cfg?.make_thumbnails !== false) {
try {
const tk = thumbKey(item.id);
await putObject(tk, await makeThumb(outBuf), 'image/webp');
thumbPath = tk;
} catch (e) { console.error('[process] Thumbnail fehlgeschlagen', e); }
}
const filename = buildResultFilename(item.filename, formatToken(r.output_format), 'png');
const outputPx = `${fin.width}x${fin.height}`;
const deliveryStatus = (r.delivery === 'picdrop' || r.delivery === 'both') ? 'pending' : 'none';
const nasStatus = cfg?.nas_enabled ? 'pending' : 'none';
await query(
`UPDATE items SET status='done', result_path=$2, filename=$3, output_px=$4, dpi=$5,
has_alpha=$6, model_used=$7, prompt_used=$8, cost=$9, error_message=NULL,
delivery_status=$10 WHERE id=$1`,
[itemId, key, filename, outputPx, dpi, hasAlpha, modelUsed, prompt.slice(0, 1000),
cost, deliveryStatus]);
`UPDATE items SET status='done', result_path=$2, thumb_path=$3, filename=$4, output_px=$5, dpi=$6,
has_alpha=$7, model_used=$8, prompt_used=$9, cost=$10, error_message=NULL,
delivery_status=$11, nas_status=$12 WHERE id=$1`,
[itemId, key, thumbPath, filename, outputPx, dpi, hasAlpha, modelUsed, prompt.slice(0, 1000),
cost, deliveryStatus, nasStatus]);
// Quellbilder löschen, wenn nicht behalten (Speicher sparen).
if (cfg?.keep_sources === false) {
for (const k of sourceKeys) await deleteObject(k).catch(() => {});
}
// Zweite Sicherung auf NAS (best effort, blockiert den Erfolg nicht).
if (cfg?.nas_enabled) {
try {
const { mirrorItemToNas } = await import('./nas');
await mirrorItemToNas(itemId);
} catch (e) { console.error('[process] NAS-Spiegelung fehlgeschlagen', e); }
}
return { ok: true, cost };
} catch (e: any) {
const real = e?.message || String(e);
const status = e?.status ? ` [status ${e.status}]` : '';
console.error(`[process] Item ${itemId} fehlgeschlagen${status}: ${real}`, e?.stack || '');
// Nutzeranzeige: freundlich, aber mit kompaktem Grund fürs Debugging in dieser Phase
const msg = (e?.friendly ? `${e.friendly}` : real) + (e?.status ? ` (${e.status})` : '');
await query(`UPDATE items SET status='failed', error_message=$2 WHERE id=$1`,
[itemId, String(msg).slice(0, 500)]);
// 402/Guthaben: nach oben durchreichen, damit der Worker den Auftrag anhält
if (e?.status === 402) throw e;
return { ok: false, cost: 0, error: msg };
}
src/lib/prompts.ts
+22
View File
@@ -43,3 +43,25 @@ export function buildPrompt({ tasks, cropMode, customInstruction }: PromptOpts):
parts.push('Output only the resulting image.');
return parts.join(' ');
}
/** Kombinieren: mehrere Vorlagen + Beschreibung → EIN neues Bild.
* Die erste Vorlage ist das Leitbild; weitere liefern Elemente/Personen/Motive. */
export function buildComposePrompt(description: string, extend = false): string {
const parts = [
'You are given several reference images. Combine them into ONE new, coherent, ' +
'high-resolution image that follows the instruction below. Treat the first image as ' +
'the main scene/style reference and use the other images as elements to integrate ' +
'(people, pets, objects) — match their identity, colors and lighting faithfully so ' +
'they look naturally part of the same photo.',
];
if (description.trim()) parts.push(`Instruction: ${description.trim()}`);
if (extend) parts.push('Extend the scene naturally to fill the requested aspect ratio (outpainting).');
parts.push('Output only the resulting image.');
return parts.join(' ');
}
/** Freitext: gar keine Vorlage → ein komplett neues Bild aus der Beschreibung. */
export function buildGeneratePrompt(description: string): string {
const d = description.trim() || 'A clean, high-quality image.';
return `Create a new, high-resolution image. ${d} Output only the resulting image.`;
}
src/lib/storage.ts
+3
View File
@@ -90,3 +90,6 @@ export function sourceKey(uuid: string, ext: string): string {
export function resultKey(uuid: string): string {
return `results/${new Date().getFullYear()}/${uuid}.png`;
}
export function thumbKey(uuid: string): string {
return `thumbs/${new Date().getFullYear()}/${uuid}.webp`;
}