@dsh-libing/knowledge-base 0.1.0 → 0.1.4

package/lib/document/pdf-worker.js

@@ -14,11 +14,14 @@ const matrixSchema = z.tuple([z.number(), z.number(), z.number(), z.number(), z.
 const markTag = (value) => z.union([z.string(), z.object({ name: z.string() })]).transform(value => typeof value === 'string' ? value : value.name).parse(value);
 const numeric = (value) => z.array(z.number().finite()).parse(value instanceof Float32Array || value instanceof Float64Array ? Array.from(value) : value);
 // PDF.js uses napi-rs on Node. Correct only subpixel clip skew in this isolated worker's supported canvas prototype.
-const proto = Object.getPrototypeOf(createCanvas(1, 1).getContext('2d')), nativeClip = proto.clip;
+const proto = Object.getPrototypeOf(createCanvas(1, 1).getContext('2d'));
+const nativeClip = proto.clip;
 proto.clip = function (path, rule) {
-    const original = this.getTransform(), tiny = (value) => value !== 0 && Math.abs(value) < 1e-7;
+    const original = this.getTransform();
+    const tiny = (value) => value !== 0 && Math.abs(value) < 1e-7;
     if (path instanceof Path2D && (tiny(original.b) || tiny(original.c))) {
-        const matrix = this.getTransform(), transformed = new Path2D();
+        const matrix = this.getTransform();
+        const transformed = new Path2D();
         if (tiny(matrix.b))
             matrix.b = 0;
         if (tiny(matrix.c))
@@ -33,13 +36,25 @@ proto.clip = function (path, rule) {
         }
     }
     else
-        Reflect.apply(nativeClip, this, path === undefined ? [] : rule === undefined ? [path] : [path, rule]);
+        Reflect.apply(nativeClip, this, (() => {
+            if (path === undefined) {
+                return [];
+            }
+            if (rule === undefined) {
+                return [path];
+            }
+            return [path, rule];
+        })());
 };
-const require = createRequire(import.meta.url), base = dirname(require.resolve('pdfjs-dist/package.json')), config = input.config;
-let doc, activePage = 0;
+const require = createRequire(import.meta.url);
+const base = dirname(require.resolve('pdfjs-dist/package.json'));
+const config = input.config;
+let doc;
+let activePage = 0;
 const engineWarnings = new Map();
 console.warn = (...parts) => {
-    const value = parts.map(String).join(' ').replace(/^Warning:\s*/u, '').slice(0, 1000), list = engineWarnings.get(activePage) || [];
+    const value = parts.map(String).join(' ').replace(/^Warning:\s*/u, '').slice(0, 1000);
+    const list = engineWarnings.get(activePage) || [];
     if (list.length < 100 && !list.includes(value))
         list.push(value);
     engineWarnings.set(activePage, list);
@@ -70,7 +85,8 @@ async function info() {
         for (const entry of items || []) {
             if (outline.length >= 10000 || level > 32)
                 throw new Error('PDF 书签层级或数量超过上限');
-            let page = null, y = null;
+            let page = null;
+            let y = null;
             try {
                 const raw = typeof entry.dest === 'string' ? await doc.getDestination(entry.dest) : entry.dest;
                 if (Array.isArray(raw) && raw.length) {
@@ -93,8 +109,11 @@ async function nativePage(n) {
     activePage = n;
     const page = await doc.getPage(n);
     try {
-        const viewport = page.getViewport({ scale: 1 }), content = await page.getTextContent({ includeMarkedContent: true }), tree = prune(await page.getStructTree());
-        const spans = [], marks = [];
+        const viewport = page.getViewport({ scale: 1 });
+        const content = await page.getTextContent({ includeMarkedContent: true });
+        const tree = prune(await page.getStructTree());
+        const spans = [];
+        const marks = [];
         let seq = 0;
         for (const item of content.items) {
             if ('type' in item) {
@@ -106,17 +125,47 @@ async function nativePage(n) {
             }
             if (!item.str.trim())
                 continue;
-            const t = matrixSchema.parse(Util.transform(viewport.transform, item.transform)), h = Math.hypot(t[2], t[3]) || item.height || 10, style = content.styles[item.fontName], a = style?.ascent ?? .8;
-            const a0 = [t[4] + t[2] * a, t[5] + t[3] * a], b0 = [t[4] + t[2] * (a - 1), t[5] + t[3] * (a - 1)];
-            const angle = Math.atan2(t[1], t[0]), w = item.width || h / 2, dx = Math.cos(angle) * w, dy = Math.sin(angle) * w;
-            const points = [a0, b0, [a0[0] + dx, a0[1] + dy], [b0[0] + dx, b0[1] + dy]], bbox = [Math.min(...points.map(p => p[0])), Math.min(...points.map(p => p[1])), Math.max(...points.map(p => p[0])), Math.max(...points.map(p => p[1]))];
-            const rotation = ((page.rotate % 360) + 360) % 360, W = viewport.width, H = viewport.height;
-            const readingBox = rotation === 90 ? [bbox[1], W - bbox[2], bbox[3], W - bbox[0]] : rotation === 180 ? [W - bbox[2], H - bbox[3], W - bbox[0], H - bbox[1]] : rotation === 270 ? [H - bbox[3], bbox[0], H - bbox[1], bbox[2]] : bbox;
+            const t = matrixSchema.parse(Util.transform(viewport.transform, item.transform));
+            const h = Math.hypot(t[2], t[3]) || item.height || 10;
+            const style = content.styles[item.fontName];
+            const a = style?.ascent ?? .8;
+            const a0 = [t[4] + t[2] * a, t[5] + t[3] * a];
+            const b0 = [t[4] + t[2] * (a - 1), t[5] + t[3] * (a - 1)];
+            const angle = Math.atan2(t[1], t[0]);
+            const w = item.width || h / 2;
+            const dx = Math.cos(angle) * w;
+            const dy = Math.sin(angle) * w;
+            const points = [a0, b0, [a0[0] + dx, a0[1] + dy], [b0[0] + dx, b0[1] + dy]];
+            const bbox = [Math.min(...points.map(p => p[0])), Math.min(...points.map(p => p[1])), Math.max(...points.map(p => p[0])), Math.max(...points.map(p => p[1]))];
+            const rotation = ((page.rotate % 360) + 360) % 360;
+            const W = viewport.width;
+            const H = viewport.height;
+            let readingBox;
+            if (rotation === 90) {
+                readingBox = [bbox[1], W - bbox[2], bbox[3], W - bbox[0]];
+            }
+            else {
+                if (rotation === 180) {
+                    readingBox = [W - bbox[2], H - bbox[3], W - bbox[0], H - bbox[1]];
+                }
+                else {
+                    if (rotation === 270) {
+                        readingBox = [H - bbox[3], bbox[0], H - bbox[1], bbox[2]];
+                    }
+                    else {
+                        readingBox = bbox;
+                    }
+                }
+            }
             spans.push({ id: `s${++seq}`, text: item.str, mcid: [...marks].reverse().find(m => m.id)?.id || null, artifact: marks.some(m => m.artifact), bbox, readingBox, fontSize: h, fontName: style?.fontFamily || item.fontName });
             if (spans.length > 100000)
                 throw new Error(`PDF 第 ${n} 页文字对象过多`);
         }
-        const operators = await page.getOperatorList(), drawings = [], stack = [], mc = [], prefix = `p${page.ref?.num || n}R_mc`;
+        const operators = await page.getOperatorList();
+        const drawings = [];
+        const stack = [];
+        const mc = [];
+        const prefix = `p${page.ref?.num || n}R_mc`;
         let matrix = [1, 0, 0, 1, 0, 0];
         const add = (b, kind, shape = null) => {
             const bounds = box(matrixSchema.parse(Util.transform(viewport.transform, matrix)), b);
@@ -126,7 +175,8 @@ async function nativePage(n) {
             drawings.push({ bbox: bounds, kind, shape, mcid: id === undefined ? null : prefix + id });
         };
         for (let i = 0; i < operators.fnArray.length; i++) {
-            const op = required(operators.fnArray[i]), args = operators.argsArray[i] || [];
+            const op = required(operators.fnArray[i]);
+            const args = operators.argsArray[i] || [];
             if (op === OPS.save)
                 stack.push([...matrix]);
             else if (op === OPS.restore)
@@ -147,7 +197,8 @@ async function nativePage(n) {
             else if (op === OPS.endMarkedContent)
                 mc.pop();
             else if (op === OPS.constructPath && [OPS.stroke, OPS.fill, OPS.eoFill, OPS.fillStroke, OPS.eoFillStroke].includes(z.number().parse(args[0])) && args[2]) {
-                const paths = Array.isArray(args[1]) ? args[1] : [], path = paths[0] === undefined ? [] : numeric(paths[0]);
+                const paths = Array.isArray(args[1]) ? args[1] : [];
+                const path = paths[0] === undefined ? [] : numeric(paths[0]);
                 add(bboxSchema.parse(numeric(args[2])), 'path', path.length === 13 && path[0] === 0 && path[3] === 1 && path[6] === 1 && path[9] === 1 && path[12] === 4 ? 'rect' : null);
             }
             else if ([OPS.paintImageXObject, OPS.paintInlineImageXObject, OPS.paintImageMaskXObject].includes(op))
@@ -163,13 +214,17 @@ async function nativePage(n) {
 }
 async function render(n, bbox) {
     activePage = n;
-    const page = await doc.getPage(n), v = page.getViewport({ scale: 1 }), b = bbox || [0, 0, v.width, v.height];
+    const page = await doc.getPage(n);
+    const v = page.getViewport({ scale: 1 });
+    const b = bbox || [0, 0, v.width, v.height];
     if (b[0] < 0 || b[1] < 0 || b[2] > v.width + .1 || b[3] > v.height + .1 || b[2] <= b[0] || b[3] <= b[1])
         throw new Error('PDF 裁切区域无效');
     const scale = Math.min(2, Math.sqrt(config.pdfMaxPagePixels / (v.width * v.height)));
     if (scale < .5)
         throw new Error(`PDF 第 ${n} 页页面尺寸过大`);
-    const viewport = page.getViewport({ scale }), canvas = createCanvas(Math.ceil(viewport.width), Math.ceil(viewport.height)), crop = createCanvas(Math.max(1, Math.ceil((b[2] - b[0]) * scale)), Math.max(1, Math.ceil((b[3] - b[1]) * scale)));
+    const viewport = page.getViewport({ scale });
+    const canvas = createCanvas(Math.ceil(viewport.width), Math.ceil(viewport.height));
+    const crop = createCanvas(Math.max(1, Math.ceil((b[2] - b[0]) * scale)), Math.max(1, Math.ceil((b[3] - b[1]) * scale)));
     try {
         // PDF.js publishes DOM-only render types for its supported Node napi-rs canvas. This bridge uses that same tested Canvas2D implementation.
         await page.render({ canvas: null, canvasContext: canvas.getContext('2d'), viewport, background: 'rgb(255,255,255)' }).promise;

package/lib/document/pdf.js

@@ -10,10 +10,14 @@ import { blockMarkdown } from "./convert.js";
 import { BudgetError } from "../models.js";
 export async function convertPdf(file, destination, job, models, signal) {
     let session = new PdfSession(file, job.config, signal);
-    const assets = new Map(), pages = [], blocks = [], warnings = [];
+    const assets = new Map();
+    const pages = [];
+    const blocks = [];
+    const warnings = [];
     job.parserVersion = PDF_PARSER_VERSION;
     const addAsset = async (data, origin, source) => {
-        const assetId = `a-${hash(data).slice(0, 20)}`, asset = { assetId, mime: 'image/png', path: `assets/${assetId}.png`, bytes: data.length, origin, source };
+        const assetId = `a-${hash(data).slice(0, 20)}`;
+        const asset = { assetId, mime: 'image/png', path: `assets/${assetId}.png`, bytes: data.length, origin, source };
         if (!assets.has(assetId)) {
             await atomicWrite(join(destination, asset.path), Buffer.from(data));
             assets.set(assetId, asset);
@@ -21,7 +25,8 @@ export async function convertPdf(file, destination, job, models, signal) {
         return asset;
     };
     try {
-        const info = await session.ready, range = pageRange(job.pageStart, job.pageEnd, info.pageCount);
+        const info = await session.ready;
+        const range = pageRange(job.pageStart, job.pageEnd, info.pageCount);
         warnings.push(...(info.warnings || []).map(message => ({ code: 'pdf-engine-warning', message: `PDF 打开警告：${message}`, blocking: true })));
         job.pdf = { pageCount: info.pageCount, selectedPages: range, outlineCount: info.outline.length };
         job.progress = { ...job.progress, totalPages: range.count, parsedPages: 0, ocrPages: 0, tables: 0 };
@@ -34,7 +39,9 @@ export async function convertPdf(file, destination, job, models, signal) {
             const receipt = join(destination, 'pdf-pages', `${n}-${key}.json`);
             let result = await readJson(receipt, pageReceiptSchema, null);
             if (!result) {
-                let page, layout, previewAsset;
+                let page;
+                let layout;
+                let previewAsset;
                 try {
                     if (session.closed) {
                         session = new PdfSession(file, job.config, signal);
@@ -94,7 +101,15 @@ export async function convertPdf(file, destination, job, models, signal) {
                         }
                         return { ...value, localIndex: i };
                     });
-                    result = { blocks: pageBlocks, assets: localAssets, warnings: layout.warnings, audit: { page: n, width: page.width, height: page.height, pageLabel: info.pageLabels?.[n - 1] || null, status: layout.warnings.some(w => w.blocking) ? 'needs-review' : layout.navigation ? 'navigation' : 'parsed', ocr: layout.ocr, spanCount: layout.spanCount, artifactCount: layout.artifactCount, excluded: layout.excluded, tables: pageBlocks.filter(b => b.type === 'table').length, figures: pageBlocks.filter(b => b.type === 'figure').length, headings: pageBlocks.filter(b => b.type === 'heading').map(b => ({ text: b.text, basis: b.headingBasis, level: b.level })), previewAssetId: previewAsset?.assetId || null } };
+                    result = { blocks: pageBlocks, assets: localAssets, warnings: layout.warnings, audit: { page: n, width: page.width, height: page.height, pageLabel: info.pageLabels?.[n - 1] || null, status: (() => {
+                                if (layout.warnings.some(w => w.blocking)) {
+                                    return 'needs-review';
+                                }
+                                if (layout.navigation) {
+                                    return 'navigation';
+                                }
+                                return 'parsed';
+                            })(), ocr: layout.ocr, spanCount: layout.spanCount, artifactCount: layout.artifactCount, excluded: layout.excluded, tables: pageBlocks.filter(b => b.type === 'table').length, figures: pageBlocks.filter(b => b.type === 'figure').length, headings: pageBlocks.filter(b => b.type === 'heading').map(b => ({ text: b.text, basis: b.headingBasis, level: b.level })), previewAssetId: previewAsset?.assetId || null } };
                 }
                 catch (error) {
                     if (error instanceof BudgetError || signal?.aborted)
@@ -148,10 +163,23 @@ export async function convertPdf(file, destination, job, models, signal) {
         if (pages.length >= 3) {
             const repeated = new Map();
             for (const b of blocks) {
-                const box = b.source?.bbox, p = pages.find(p => p.page === b.source?.page);
+                const box = b.source?.bbox;
+                const p = pages.find(p => p.page === b.source?.page);
                 if (b.type !== 'paragraph' || !box || b.text.length > 160)
                     continue;
-                const key = normalize(b.text).replace(/\d+/g, '#'), edge = box[1] < 40 ? 'top' : box[1] > (p?.height || 842) - 45 ? 'bottom' : null;
+                const key = normalize(b.text).replace(/\d+/g, '#');
+                let edge;
+                if (box[1] < 40) {
+                    edge = 'top';
+                }
+                else {
+                    if (box[1] > (p?.height || 842) - 45) {
+                        edge = 'bottom';
+                    }
+                    else {
+                        edge = null;
+                    }
+                }
                 if (edge) {
                     const k = edge + ':' + key;
                     repeated.set(k, [...(repeated.get(k) || []), b]);
@@ -179,11 +207,13 @@ export async function convertPdf(file, destination, job, models, signal) {
                     repeated.push(item);
                 previous--;
             }
-            const a = blocks[previous], last = a?.source?.locations?.at(-1) || a?.source;
+            const a = blocks[previous];
+            const last = a?.source?.locations?.at(-1) || a?.source;
             const tableId = (t) => /^(?:表|Table)\s*([A-Z\d][\w.\-–]*)/iu.exec(t.caption)?.[1];
             if (!a || a.type !== 'table' || !a.source || !last?.bbox || !a.rows[0] || !b.rows[0])
                 continue;
-            const height = pages.find(p => p.page === last.page)?.height || 842, nextHeight = pages.find(p => p.page === b.source?.page)?.height || 842;
+            const height = pages.find(p => p.page === last.page)?.height || 842;
+            const nextHeight = pages.find(p => p.page === b.source?.page)?.height || 842;
             if (b.source.page === last.page + 1 && tableId(a) && tableId(a) === tableId(b) && last.bbox[3] > height * .8 && b.source.bbox[1] < nextHeight * .25 && !a.needsReview && !b.needsReview && JSON.stringify(a.rows[0].map(c => normalize(c.text))) === JSON.stringify(b.rows[0].map(c => normalize(c.text))) && a.rows[0].length === b.rows[0].length && [...a.rows, ...b.rows].every(r => r.length === required(a.rows[0]).length && r.every(c => c.rowspan === 1 && c.colspan === 1))) {
                 a.rows.push(...b.rows.slice(1));
                 const { locations, ...first } = a.source;

package/lib/document/zip.js

@@ -12,17 +12,24 @@ export function unpackDocx(buffer, config) {
     }
     if (end < 0)
         throw new Error('无效 DOCX ZIP 文件');
-    const count = buffer.readUInt16LE(end + 10), size = buffer.readUInt32LE(end + 12), start = buffer.readUInt32LE(end + 16);
+    const count = buffer.readUInt16LE(end + 10);
+    const size = buffer.readUInt32LE(end + 12);
+    const start = buffer.readUInt32LE(end + 16);
     if (count === 65535 || count > config.maxArchiveEntries || start + size > end || buffer.readUInt16LE(end + 4) !== 0)
         throw new Error('不支持的分卷/ZIP64 或过多 ZIP 条目');
-    let pos = start, total = 0;
+    let pos = start;
+    let total = 0;
     const entries = new Map();
     for (let n = 0; n < count; n++) {
         if (pos + 46 > end || buffer.readUInt32LE(pos) !== 0x02014b50)
             throw new Error('损坏的 ZIP 目录');
-        const flags = buffer.readUInt16LE(pos + 8), method = buffer.readUInt16LE(pos + 10);
-        const compressed = buffer.readUInt32LE(pos + 20), uncompressed = buffer.readUInt32LE(pos + 24);
-        const nameLength = buffer.readUInt16LE(pos + 28), extraLength = buffer.readUInt16LE(pos + 30), commentLength = buffer.readUInt16LE(pos + 32);
+        const flags = buffer.readUInt16LE(pos + 8);
+        const method = buffer.readUInt16LE(pos + 10);
+        const compressed = buffer.readUInt32LE(pos + 20);
+        const uncompressed = buffer.readUInt32LE(pos + 24);
+        const nameLength = buffer.readUInt16LE(pos + 28);
+        const extraLength = buffer.readUInt16LE(pos + 30);
+        const commentLength = buffer.readUInt16LE(pos + 32);
         const next = pos + 46 + nameLength + extraLength + commentLength;
         if (next > end || uncompressed === 0xffffffff || compressed === 0xffffffff)
             throw new Error('损坏或不支持的 ZIP 条目');
@@ -34,7 +41,8 @@ export function unpackDocx(buffer, config) {
         const local = buffer.readUInt32LE(pos + 42);
         if (local + 30 > start || buffer.readUInt32LE(local) !== 0x04034b50)
             throw new Error('损坏的 ZIP 本地条目');
-        const localNameLength = buffer.readUInt16LE(local + 26), localExtra = buffer.readUInt16LE(local + 28);
+        const localNameLength = buffer.readUInt16LE(local + 26);
+        const localExtra = buffer.readUInt16LE(local + 28);
         if (buffer.subarray(local + 30, local + 30 + localNameLength).toString('utf8') !== name || local + 30 + localNameLength + localExtra + compressed > start)
             throw new Error('ZIP 条目不一致');
         total += uncompressed;

package/lib/figures/image-client.d.ts

@@ -0,0 +1,5 @@
+import type { Settings } from '../config.ts';
+export declare function runImageWorker(file: string, config: Settings, save: (data: Uint8Array, coordinates: Record<string, string | number>) => Promise<void>, signal?: AbortSignal): Promise<{
+    width: number;
+    height: number;
+}>;

package/lib/figures/image-client.js

@@ -0,0 +1,63 @@
+import { Worker } from 'node:worker_threads';
+import { existsSync } from 'node:fs';
+import { fileURLToPath } from 'node:url';
+import { checkAbort, deferred } from "../util.js";
+import { imageGate } from "./image-gate.js";
+import { imageMessageSchema, imageOptionsSchema } from "./image-protocol.js";
+export async function runImageWorker(file, config, save, signal) {
+    const options = imageOptionsSchema.parse({ visionMaxEdge: config.visionMaxEdge, visionTileSize: config.visionTileSize, visionMaxTiles: config.visionMaxTiles });
+    const release = await imageGate.acquire(signal);
+    let worker;
+    let saving;
+    const result = deferred();
+    const abort = () => result.reject(signal?.reason ?? new Error('图片预处理已取消'));
+    let timer;
+    let complete = false;
+    let next = 1;
+    try {
+        checkAbort(signal);
+        const js = new URL('./image-worker.js', import.meta.url);
+        const source = !existsSync(fileURLToPath(js));
+        worker = new Worker(source ? new URL('./image-worker.ts', import.meta.url) : js, { workerData: { file, options }, execArgv: source ? ['--import', 'tsx/esm'] : [], resourceLimits: { maxOldGenerationSizeMb: 512 } });
+        const current = worker;
+        timer = setTimeout(() => result.reject(new Error('图片预处理超时')), Math.min(config.imageTimeoutMs, 120000));
+        signal?.addEventListener('abort', abort, { once: true });
+        if (signal?.aborted)
+            abort();
+        current.on('error', error => result.reject(error));
+        current.on('exit', code => { if (!complete)
+            result.reject(new Error(`图片预处理进程退出（${code}）`)); });
+        current.on('message', (raw) => {
+            try {
+                const value = imageMessageSchema.parse(raw);
+                if (value.kind === 'error')
+                    result.reject(new Error(value.message));
+                else if (value.kind === 'done') {
+                    complete = true;
+                    result.resolve({ width: value.width, height: value.height });
+                }
+                else {
+                    if (saving || value.id !== next++ || value.data.length > 8 * 1024 * 1024)
+                        throw new Error('图片预处理输出无效');
+                    saving = Promise.resolve().then(() => { checkAbort(signal); return save(value.data, value.coordinates); }).then(() => { checkAbort(signal); saving = undefined; current.postMessage({ ack: value.id }); });
+                    void saving.catch(error => result.reject(error));
+                }
+            }
+            catch (error) {
+                result.reject(error);
+            }
+        });
+        return await result.promise;
+    }
+    finally {
+        clearTimeout(timer);
+        signal?.removeEventListener('abort', abort);
+        try {
+            await worker?.terminate();
+            await saving?.catch(() => { });
+        }
+        finally {
+            release();
+        }
+    }
+}

package/lib/figures/image-gate.d.ts

@@ -0,0 +1,12 @@
+/** One process-wide bitmap worker, including all knowledge bases and runtime instances. */
+export declare class ImageGate {
+    active: boolean;
+    private queue;
+    get waiting(): number;
+    acquire(signal?: AbortSignal): Promise<() => void>;
+    private release;
+}
+declare global {
+    var __dshLibingKnowledgeImageGateV1: ImageGate | undefined;
+}
+export declare const imageGate: ImageGate;

package/lib/figures/image-gate.js

@@ -0,0 +1,39 @@
+import { checkAbort } from "../util.js";
+/** One process-wide bitmap worker, including all knowledge bases and runtime instances. */
+export class ImageGate {
+    active = false;
+    queue = [];
+    get waiting() { return this.queue.length; }
+    async acquire(signal) {
+        checkAbort(signal);
+        if (!this.active) {
+            this.active = true;
+            return this.release();
+        }
+        if (this.queue.length >= 32)
+            throw new Error('图片预处理等待队列已满，请稍后重试');
+        return new Promise((resolve, reject) => {
+            const entry = { ...signal ? { signal } : {}, resolve, reject, abort: () => { signal?.removeEventListener('abort', entry.abort); this.queue = this.queue.filter(e => e !== entry); reject(signal?.reason ?? new Error('图片预处理已取消')); } };
+            this.queue.push(entry);
+            signal?.addEventListener('abort', entry.abort, { once: true });
+            if (signal?.aborted)
+                entry.abort();
+        });
+    }
+    release() {
+        let released = false;
+        return () => {
+            if (released)
+                return;
+            released = true;
+            const next = this.queue.shift();
+            if (next) {
+                next.signal?.removeEventListener('abort', next.abort);
+                next.resolve(this.release());
+            }
+            else
+                this.active = false;
+        };
+    }
+}
+export const imageGate = globalThis.__dshLibingKnowledgeImageGateV1 ??= new ImageGate();

package/lib/figures/image-protocol.d.ts

@@ -0,0 +1,19 @@
+import { z } from 'zod';
+export declare const imageOptionsSchema: z.ZodObject<{
+    visionMaxEdge: z.ZodNumber;
+    visionTileSize: z.ZodNumber;
+    visionMaxTiles: z.ZodNumber;
+}, z.core.$strict>;
+export declare const imageMessageSchema: z.ZodDiscriminatedUnion<[z.ZodObject<{
+    kind: z.ZodLiteral<"image">;
+    id: z.ZodNumber;
+    data: z.ZodCustom<Uint8Array<ArrayBuffer>, Uint8Array<ArrayBuffer>>;
+    coordinates: z.ZodRecord<z.ZodString, z.ZodUnion<readonly [z.ZodString, z.ZodNumber]>>;
+}, z.core.$strip>, z.ZodObject<{
+    kind: z.ZodLiteral<"done">;
+    width: z.ZodNumber;
+    height: z.ZodNumber;
+}, z.core.$strip>, z.ZodObject<{
+    kind: z.ZodLiteral<"error">;
+    message: z.ZodString;
+}, z.core.$strip>], "kind">;

package/lib/figures/image-protocol.js

@@ -0,0 +1,9 @@
+import { z } from 'zod';
+export const imageOptionsSchema = z.strictObject({
+    visionMaxEdge: z.number().int().min(1).max(4096), visionTileSize: z.number().int().min(1).max(4096), visionMaxTiles: z.number().int().min(1).max(64),
+});
+export const imageMessageSchema = z.discriminatedUnion('kind', [
+    z.object({ kind: z.literal('image'), id: z.number().int().positive(), data: z.instanceof(Uint8Array), coordinates: z.record(z.string(), z.union([z.string(), z.number()])) }),
+    z.object({ kind: z.literal('done'), width: z.number().int().positive(), height: z.number().int().positive() }),
+    z.object({ kind: z.literal('error'), message: z.string() }),
+]);

package/lib/figures/image-worker.d.ts

@@ -0,0 +1 @@
+export {};

package/lib/figures/image-worker.js

@@ -0,0 +1,69 @@
+import { parentPort, workerData } from 'node:worker_threads';
+import { readFile, stat } from 'node:fs/promises';
+import { Jimp, ResizeStrategy } from 'jimp';
+import { imageSize } from 'image-size';
+import { z } from 'zod';
+import { imageOptionsSchema } from "./image-protocol.js";
+import { message, required } from "../util.js";
+const port = required(parentPort);
+async function run() {
+    const { file, options } = z.object({ file: z.string(), options: imageOptionsSchema }).parse(workerData);
+    if ((await stat(file)).size > 32 * 1024 * 1024)
+        throw new Error('单张图片超过 32 MiB，请提供较小的位图');
+    const data = await readFile(file);
+    const dimensions = imageSize(data);
+    if (data.length > 32 * 1024 * 1024)
+        throw new Error('单张图片超过 32 MiB，请提供较小的位图');
+    if (!dimensions.width || !dimensions.height || dimensions.width * dimensions.height > 50000000)
+        throw new Error('图片像素过大，请提供分辨率较低的版本');
+    const image = await Jimp.read(data);
+    const { width, height } = image.bitmap;
+    if (width * height > 50000000)
+        throw new Error('图片像素过大，请提供分辨率较低的版本');
+    const tile = options.visionTileSize;
+    const step = Math.max(1, Math.floor(tile * .85));
+    const tiled = Math.max(width, height) > options.visionMaxEdge;
+    const columns = Math.max(1, Math.ceil((width - tile) / step) + 1);
+    const rows = Math.max(1, Math.ceil((height - tile) / step) + 1);
+    if (tiled && columns * rows > options.visionMaxTiles)
+        throw new Error(`大图需要 ${columns * rows} 个裁片，超过 visionMaxTiles；请拆分图片或提高上限`);
+    let id = 0;
+    const send = async (bitmap, coordinates) => {
+        const bytes = await bitmap.getBuffer('image/png');
+        if (bytes.length > 8 * 1024 * 1024)
+            throw new Error('图像转换后超过单张 8 MiB，请降低 visionTileSize');
+        const packet = new Uint8Array(bytes);
+        const current = ++id;
+        const acknowledged = new Promise((resolve, reject) => port.once('message', raw => {
+            const ack = z.object({ ack: z.literal(current) }).safeParse(raw);
+            if (ack.success)
+                resolve();
+            else
+                reject(new Error('图片保存确认无效'));
+        }));
+        port.postMessage({ kind: 'image', id: current, data: packet, coordinates }, [packet.buffer]);
+        await acknowledged;
+    };
+    // A separate bitmap descriptor shares the read-only source buffer; resize allocates only the overview.
+    const overview = new Jimp({ ...image.bitmap });
+    if (tiled) {
+        const ratio = options.visionMaxEdge / Math.max(width, height);
+        overview.resize({ w: Math.max(1, Math.round(width * ratio)), h: Math.max(1, Math.round(height * ratio)), mode: ResizeStrategy.BILINEAR });
+    }
+    await send(overview, { kind: 'overview', originalWidth: width, originalHeight: height });
+    if (tiled)
+        for (let row = 0; row < rows; row++)
+            for (let col = 0; col < columns; col++) {
+                const x = Math.min(col * step, Math.max(0, width - tile));
+                const y = Math.min(row * step, Math.max(0, height - tile));
+                const w = Math.min(tile, width);
+                const h = Math.min(tile, height);
+                // Copy only the requested region; never clone the full source for a tile.
+                const bitmap = new Jimp({ width: w, height: h });
+                for (let line = 0; line < h; line++)
+                    image.bitmap.data.copy(bitmap.bitmap.data, line * w * 4, ((y + line) * width + x) * 4, ((y + line) * width + x + w) * 4);
+                await send(bitmap, { kind: 'tile', x, y, width: w, height: h, originalWidth: width, originalHeight: height });
+            }
+    port.postMessage({ kind: 'done', width, height });
+}
+void run().catch(error => port.postMessage({ kind: 'error', message: message(error) })).finally(() => port.close());

package/lib/figures/process.d.ts

@@ -5,9 +5,9 @@ import type { Block, Document, Job, Overlay } from '../data.ts';
 import type { Models, PreparedImage } from '../models.ts';
 export declare function figureContext(blocks: Block[], index: number, headingPath: string[]): string;
 export declare function prepareImage(ctx: Context, file: string, config: Settings, signal?: AbortSignal): Promise<{
-    images: PreparedImage[];
     width: number;
     height: number;
+    images: PreparedImage[];
 }>;
 export declare function figureEntries(document: Document): {
     block: Block;

package/lib/figures/process.js

@@ -2,9 +2,7 @@ import { z } from 'zod';
 import { figureSchema, figureDataSchema, FIGURE_TYPES } from "../data.js";
 import { required, message } from "../util.js";
 const classificationSchema = z.object({ types: z.array(z.enum(FIGURE_TYPES)).min(1), reason: z.string() });
-import { Jimp } from 'jimp';
-import { imageSize } from 'image-size';
-import { readFile } from 'node:fs/promises';
+import { runImageWorker } from "./image-client.js";
 import { join } from 'node:path';
 import { atomicWrite, checkAbort, clip, hash, readJson } from "../util.js";
 import { BudgetError } from "../models.js";
@@ -13,41 +11,14 @@ export function figureContext(blocks, index, headingPath) {
     return JSON.stringify({ headingPath, caption: required(blocks[index]).caption, nearby: blocks.slice(Math.max(0, index - 2), index + 3).filter(b => b.type !== 'figure').map(b => clip(b.text || '', 1000)) });
 }
 export async function prepareImage(ctx, file, config, signal) {
-    checkAbort(signal);
-    const data = await readFile(file);
-    if (data.length > 32 * 1024 * 1024)
-        throw new Error('单张图片超过 32 MiB，请提供较小的位图');
-    const dimensions = imageSize(data);
-    if (!dimensions.width || !dimensions.height || dimensions.width * dimensions.height > 50000000)
-        throw new Error('图片像素过大，请提供分辨率较低的版本');
-    const image = await Jimp.read(data);
-    const { width, height } = image.bitmap;
-    if (width * height > 50000000)
-        throw new Error('图片像素过大，请提供分辨率较低的版本');
     const images = [];
-    const add = async (bitmap, coordinates) => {
+    const dimensions = await runImageWorker(file, config, async (data, coordinates) => {
+        checkAbort(signal);
+        const attachment = await ctx.attachments.saveImage({ data: Buffer.from(data.buffer, data.byteOffset, data.byteLength), mediaType: 'image/png', name: 'kb-figure.png' });
         checkAbort(signal);
-        const bytes = await bitmap.getBuffer('image/png');
-        if (bytes.length > 8 * 1024 * 1024)
-            throw new Error('图像转换后超过单张 8 MiB，请降低 visionTileSize');
-        const attachment = await ctx.attachments.saveImage({ data: bytes, mediaType: 'image/png', name: 'kb-figure.png' });
         images.push({ attachment, coordinates });
-    };
-    const overview = image.clone();
-    if (Math.max(width, height) > config.visionMaxEdge)
-        overview.scale(config.visionMaxEdge / Math.max(width, height));
-    await add(overview, { kind: 'overview', originalWidth: width, originalHeight: height });
-    if (Math.max(width, height) > config.visionMaxEdge) {
-        const tile = config.visionTileSize, step = Math.max(1, Math.floor(tile * .85)), columns = Math.max(1, Math.ceil((width - tile) / step) + 1), rows = Math.max(1, Math.ceil((height - tile) / step) + 1);
-        if (columns * rows > config.visionMaxTiles)
-            throw new Error(`大图需要 ${columns * rows} 个裁片，超过 visionMaxTiles；请拆分图片或提高上限`);
-        for (let row = 0; row < rows; row++)
-            for (let col = 0; col < columns; col++) {
-                const x = Math.min(col * step, Math.max(0, width - tile)), y = Math.min(row * step, Math.max(0, height - tile)), w = Math.min(tile, width), h = Math.min(tile, height);
-                await add(image.clone().crop({ x, y, w, h }), { kind: 'tile', x, y, width: w, height: h, originalWidth: width, originalHeight: height });
-            }
-    }
-    return { images, width, height };
+    }, signal);
+    return { images, ...dimensions };
 }
 export function figureEntries(document) {
     let heading = [];
@@ -74,8 +45,10 @@ export async function processFigures(document, sourceDir, job, models, signal, o
             checkAbort(signal);
             job.progress.inflightFigures = [...(job.progress.inflightFigures || []), { occurrenceId: block.occurrenceId, caption: block.caption || block.occurrenceId }];
             await job.checkpoint();
-            const asset = document.assets.find(x => x.assetId === block.assetId), resultPath = join(sourceDir, 'figures', block.occurrenceId + '.json');
-            const overlayKey = hash({ assetId: block.assetId, context }), overlay = overlays[overlayKey];
+            const asset = document.assets.find(x => x.assetId === block.assetId);
+            const resultPath = join(sourceDir, 'figures', block.occurrenceId + '.json');
+            const overlayKey = hash({ assetId: block.assetId, context });
+            const overlay = overlays[overlayKey];
             const cacheKey = hash({ assetId: block.assetId, context, route: job.routes.vision, prompt: PROMPT_VERSION, params: { maxEdge: job.config.visionMaxEdge, tile: job.config.visionTileSize, maxChars: job.config.figureMaxChars } });
             let result = await readJson(resultPath, figureSchema, null);
             const overlayHash = hash(overlay || null);
@@ -89,7 +62,15 @@ export async function processFigures(document, sourceDir, job, models, signal, o
                     if (!asset)
                         throw new Error('未提取原图');
                     if (overlay)
-                        result = figureSchema.parse({ ...overlay, cacheKey, overlayKey, overlayHash, assetId: block.assetId, occurrenceId: block.occurrenceId, caption: block.caption, status: overlay.ignore ? 'ignored' : overlay.skip ? 'decoration' : 'reviewed', humanReviewed: true });
+                        result = figureSchema.parse({ ...overlay, cacheKey, overlayKey, overlayHash, assetId: block.assetId, occurrenceId: block.occurrenceId, caption: block.caption, status: (() => {
+                                if (overlay.ignore) {
+                                    return 'ignored';
+                                }
+                                if (overlay.skip) {
+                                    return 'decoration';
+                                }
+                                return 'reviewed';
+                            })(), humanReviewed: true });
                     else if (job.config.skipImages)
                         result = figureSchema.parse({ status: 'skipped', cacheKey, overlayKey, overlayHash, assetId: block.assetId, occurrenceId: block.occurrenceId, caption: block.caption, description: '本次明确跳过图片理解', needsReview: true, uncertainties: ['未分析图片'] });
                     else {