Files
Genarrative/server-node/src/services/sceneImageService.ts
高物 09d4c0c31b
Some checks failed
CI / verify (push) Has been cancelled
11
2026-04-16 21:47:20 +08:00

434 lines
11 KiB
TypeScript

import fs from 'node:fs';
import { readFile } from 'node:fs/promises';
import path from 'node:path';
import { z } from 'zod';
import type { AppContext } from '../context.js';
import { badRequest } from '../errors.js';
import { extractApiErrorMessage } from '../http.js';
export const sceneImageSchema = z.object({
prompt: z.string().trim().min(1),
negativePrompt: z.string().trim().optional().default(''),
size: z.string().trim().optional().default('1280*720'),
model: z.string().trim().optional().default(''),
worldName: z.string().trim().optional().default(''),
profileId: z.string().trim().optional().default(''),
landmarkName: z.string().trim().optional().default(''),
landmarkId: z.string().trim().optional().default(''),
referenceImageSrc: z.string().trim().optional().default(''),
});
const TEXT_TO_IMAGE_SCENE_MODEL = 'wan2.2-t2i-flash';
const REFERENCE_IMAGE_SCENE_MODEL = 'qwen-image-2.0';
function parseImageDataUrl(source: string) {
const matched = /^data:(image\/[^;]+);base64,(.+)$/u.exec(source);
if (!matched) {
return null;
}
return {
buffer: Buffer.from(matched[2], 'base64'),
mimeType: matched[1],
};
}
async function resolveReferenceImageAsDataUrl(rootDir: string, source: string) {
const trimmedSource = source.trim();
if (!trimmedSource) {
return '';
}
const parsedDataUrl = parseImageDataUrl(trimmedSource);
if (parsedDataUrl) {
return trimmedSource;
}
if (!trimmedSource.startsWith('/')) {
throw badRequest('参考图必须是 Data URL 或 public 目录下的 URL。');
}
const normalizedSource = path.posix
.normalize(trimmedSource)
.replace(/^\/+/u, '');
const absolutePath = path.resolve(
rootDir,
'public',
...normalizedSource.split('/'),
);
const publicRoot = path.resolve(rootDir, 'public');
if (!absolutePath.startsWith(publicRoot)) {
throw badRequest('参考图路径越界。');
}
const buffer = await readFile(absolutePath);
const extension = path.extname(absolutePath).replace(/^\./u, '').toLowerCase();
const mimeType = (() => {
switch (extension) {
case 'jpg':
case 'jpeg':
return 'image/jpeg';
case 'webp':
return 'image/webp';
default:
return 'image/png';
}
})();
return `data:${mimeType};base64,${buffer.toString('base64')}`;
}
function collectStringsByKey(
value: unknown,
targetKey: string,
results: string[],
) {
if (typeof value === 'string') {
return;
}
if (Array.isArray(value)) {
value.forEach((entry) => collectStringsByKey(entry, targetKey, results));
return;
}
if (!value || typeof value !== 'object') {
return;
}
Object.entries(value).forEach(([key, nestedValue]) => {
if (key === targetKey && typeof nestedValue === 'string' && nestedValue.trim()) {
results.push(nestedValue.trim());
return;
}
collectStringsByKey(nestedValue, targetKey, results);
});
}
function findFirstStringByKey(value: unknown, targetKey: string) {
const results: string[] = [];
collectStringsByKey(value, targetKey, results);
return results[0] ?? '';
}
function extractTaskId(payload: Record<string, unknown>) {
return findFirstStringByKey(payload, 'task_id');
}
function extractImageUrls(payload: Record<string, unknown>) {
const urls: string[] = [];
collectStringsByKey(payload, 'image', urls);
collectStringsByKey(payload, 'url', urls);
return [...new Set(urls)];
}
async function createSceneImageTask(params: {
baseUrl: string;
apiKey: string;
payload: z.infer<typeof sceneImageSchema>;
}) {
const { baseUrl, apiKey, payload } = params;
const response = await fetch(`${baseUrl}/services/aigc/image-generation/generation`, {
method: 'POST',
headers: {
Authorization: `Bearer ${apiKey}`,
'Content-Type': 'application/json',
'X-DashScope-Async': 'enable',
},
body: JSON.stringify({
model: payload.model,
input: {
messages: [
{
role: 'user',
content: [{ text: payload.prompt }],
},
],
},
parameters: {
n: 1,
size: payload.size,
prompt_extend: true,
watermark: false,
...(payload.negativePrompt
? { negative_prompt: payload.negativePrompt }
: {}),
},
}),
});
const responseText = await response.text();
if (!response.ok) {
return {
ok: false as const,
errorMessage: extractApiErrorMessage(
responseText,
'创建场景图片生成任务失败',
),
};
}
return {
ok: true as const,
payload: JSON.parse(responseText) as Record<string, unknown>,
};
}
async function createSceneImageFromReference(params: {
baseUrl: string;
apiKey: string;
payload: z.infer<typeof sceneImageSchema>;
referenceImage: string;
}) {
const { baseUrl, apiKey, payload, referenceImage } = params;
const response = await fetch(
`${baseUrl}/services/aigc/multimodal-generation/generation`,
{
method: 'POST',
headers: {
Authorization: `Bearer ${apiKey}`,
'Content-Type': 'application/json',
},
body: JSON.stringify({
model: payload.model,
input: {
messages: [
{
role: 'user',
content: [{ image: referenceImage }, { text: payload.prompt }],
},
],
},
parameters: {
n: 1,
size: payload.size,
prompt_extend: true,
watermark: false,
...(payload.negativePrompt
? { negative_prompt: payload.negativePrompt }
: {}),
},
}),
},
);
const responseText = await response.text();
if (!response.ok) {
return {
ok: false as const,
errorMessage: extractApiErrorMessage(
responseText,
'创建参考图场景编辑任务失败',
),
};
}
const responsePayload = JSON.parse(responseText) as Record<string, unknown>;
const imageUrl = extractImageUrls(responsePayload)[0] ?? '';
if (!imageUrl) {
return {
ok: false as const,
errorMessage: '参考图场景编辑未返回图片地址',
};
}
return {
ok: true as const,
imageUrl,
actualPrompt: findFirstStringByKey(responsePayload, 'actual_prompt').trim(),
taskId: `scene-edit-${Date.now()}`,
};
}
function ensurePayload(
payload: z.infer<typeof sceneImageSchema>,
_defaultModel: string,
) {
if (!payload.landmarkName && !payload.landmarkId) {
throw badRequest('landmarkName 或 landmarkId 至少要提供一个');
}
const referenceImageSrc =
typeof payload.referenceImageSrc === 'string'
? payload.referenceImageSrc.trim()
: '';
return {
...payload,
referenceImageSrc,
model: referenceImageSrc
? REFERENCE_IMAGE_SCENE_MODEL
: TEXT_TO_IMAGE_SCENE_MODEL,
};
}
async function saveSceneImageAsset(params: {
context: AppContext;
payload: z.infer<typeof sceneImageSchema>;
imageUrl: string;
taskId: string;
actualPrompt: string;
}) {
const { context, payload, imageUrl, taskId, actualPrompt } = params;
const imageResponse = await fetch(imageUrl);
if (!imageResponse.ok) {
throw badRequest('下载生成图片失败');
}
const imageBuffer = Buffer.from(await imageResponse.arrayBuffer());
const contentType = imageResponse.headers.get('content-type') || '';
const extension = contentType.includes('png')
? 'png'
: contentType.includes('webp')
? 'webp'
: 'jpg';
const assetId = `custom-scene-${Date.now()}`;
const worldSegment = (payload.profileId || payload.worldName || 'world')
.replace(/[^\w\u4e00-\u9fa5-]+/gu, '-')
.slice(0, 48);
const landmarkSegment = (payload.landmarkId || payload.landmarkName || 'landmark')
.replace(/[^\w\u4e00-\u9fa5-]+/gu, '-')
.slice(0, 48);
const relativeDir = path.join(
'generated-custom-world-scenes',
worldSegment || 'world',
landmarkSegment || 'landmark',
assetId,
);
const outputDir = path.join(context.config.publicDir, relativeDir);
fs.mkdirSync(outputDir, { recursive: true });
const fileName = `scene.${extension}`;
fs.writeFileSync(path.join(outputDir, fileName), imageBuffer);
const imageSrc = `/${relativeDir.replace(/\\/gu, '/')}/${fileName}`;
fs.writeFileSync(
path.join(outputDir, 'manifest.json'),
`${JSON.stringify(
{
assetId,
taskId,
model: payload.model,
size: payload.size,
prompt: payload.prompt,
negativePrompt: payload.negativePrompt,
actualPrompt,
imageSrc,
worldName: payload.worldName,
landmarkName: payload.landmarkName,
createdAt: new Date().toISOString(),
},
null,
2,
)}\n`,
);
return {
ok: true,
imageSrc,
assetId,
taskId,
model: payload.model,
size: payload.size,
prompt: payload.prompt,
actualPrompt,
};
}
export async function generateSceneImage(
context: AppContext,
input: z.infer<typeof sceneImageSchema>,
) {
const payload = ensurePayload(input, context.config.dashScope.imageModel);
const baseUrl = context.config.dashScope.baseUrl.replace(/\/+$/u, '');
const referenceImage = payload.referenceImageSrc.trim()
? await resolveReferenceImageAsDataUrl(
context.config.projectRoot,
payload.referenceImageSrc,
)
: '';
if (referenceImage) {
const referenceResult = await createSceneImageFromReference({
baseUrl,
apiKey: context.config.dashScope.apiKey,
payload,
referenceImage,
});
if (!referenceResult.ok) {
throw badRequest(referenceResult.errorMessage);
}
return saveSceneImageAsset({
context,
payload,
imageUrl: referenceResult.imageUrl,
taskId: referenceResult.taskId,
actualPrompt: referenceResult.actualPrompt,
});
}
const createTaskResult = await createSceneImageTask({
baseUrl,
apiKey: context.config.dashScope.apiKey,
payload,
});
if (!createTaskResult.ok) {
throw badRequest(createTaskResult.errorMessage);
}
const createPayload = createTaskResult.payload;
const taskId = extractTaskId(createPayload);
if (!taskId) {
throw badRequest('场景图片生成任务未返回 task_id');
}
const deadline = Date.now() + context.config.dashScope.requestTimeoutMs;
let imageUrl = '';
let actualPrompt = '';
while (Date.now() < deadline) {
const pollResponse = await fetch(`${baseUrl}/tasks/${taskId}`, {
headers: {
Authorization: `Bearer ${context.config.dashScope.apiKey}`,
},
});
const pollText = await pollResponse.text();
if (!pollResponse.ok) {
throw badRequest(
extractApiErrorMessage(pollText, '查询场景图片任务失败'),
);
}
const pollPayload = JSON.parse(pollText) as Record<string, unknown>;
const status = findFirstStringByKey(pollPayload, 'task_status').trim();
if (status === 'SUCCEEDED') {
imageUrl = extractImageUrls(pollPayload)[0] ?? '';
actualPrompt = findFirstStringByKey(pollPayload, 'actual_prompt').trim();
break;
}
if (status === 'FAILED' || status === 'UNKNOWN') {
throw badRequest(
extractApiErrorMessage(pollText, '场景图片生成任务失败'),
);
}
await new Promise((resolve) => setTimeout(resolve, 2000));
}
if (!imageUrl) {
throw badRequest('场景图片生成超时或未返回图片地址');
}
return saveSceneImageAsset({
context,
payload,
imageUrl,
taskId,
actualPrompt,
});
}