abb-opencode-local-rag 0.1.0

package/dist/parser/title-extractor.js

@@ -0,0 +1,145 @@
+// Title Extractor - Per-format document title extraction
+// Title is display-only metadata (NOT used for search scoring)
+// ============================================
+// Constants
+// ============================================
+/** Minimum font size (pt) for page-1 largest-font text to be treated as a title. */
+const TITLE_MIN_FONT_SIZE = 14;
+// ============================================
+// Shared Helper
+// ============================================
+/**
+ * Convert a file name to a human-readable title
+ * Strips the extension and replaces hyphens/underscores with spaces
+ *
+ * @param fileName - File name (e.g., "2024-annual-report.pdf")
+ * @returns Human-readable title (e.g., "2024 annual report")
+ */
+export function fileNameToTitle(fileName) {
+    // Strip extension (last dot and everything after)
+    const lastDotIndex = fileName.lastIndexOf('.');
+    const nameWithoutExt = lastDotIndex > 0 ? fileName.substring(0, lastDotIndex) : fileName;
+    // Replace hyphens and underscores with spaces
+    return nameWithoutExt.replace(/[-_]/g, ' ');
+}
+// ============================================
+// Per-Format Extractors
+// ============================================
+/**
+ * Extract title from Markdown content
+ * Priority: YAML frontmatter title -> first # H1 -> file name
+ *
+ * @param text - Markdown content
+ * @param fileName - File name for fallback
+ * @returns Title extraction result
+ */
+export function extractMarkdownTitle(text, fileName) {
+    // 1. Try YAML frontmatter
+    const frontmatterMatch = text.match(/^---\n[\s\S]*?title:\s*['"]?(.+?)['"]?\s*\n[\s\S]*?---/);
+    if (frontmatterMatch?.[1]) {
+        return { title: frontmatterMatch[1].trim(), source: 'metadata' };
+    }
+    // 2. Try first H1 heading
+    const h1Match = text.match(/^# (.+)$/m);
+    if (h1Match?.[1]) {
+        return { title: h1Match[1].trim(), source: 'content' };
+    }
+    // 3. Fall back to file name
+    return { title: fileNameToTitle(fileName), source: 'filename' };
+}
+/**
+ * Extract title from plain text content
+ * Priority: first line followed by empty line -> file name
+ *
+ * @param text - Plain text content
+ * @param fileName - File name for fallback
+ * @returns Title extraction result
+ */
+export function extractTxtTitle(text, fileName) {
+    // Try first line followed by empty line
+    if (text.length > 0) {
+        const lines = text.split('\n');
+        const firstLine = lines[0];
+        const secondLine = lines[1];
+        if (firstLine !== undefined &&
+            secondLine !== undefined &&
+            firstLine.trim().length > 0 &&
+            secondLine.trim().length === 0) {
+            return { title: firstLine.trim(), source: 'content' };
+        }
+    }
+    // Fall back to file name
+    return { title: fileNameToTitle(fileName), source: 'filename' };
+}
+/**
+ * Extract title from HTML content (using Readability title)
+ * Priority: readability title -> file name
+ *
+ * @param readabilityTitle - Title extracted by Readability
+ * @param fileName - File name for fallback
+ * @returns Title extraction result
+ */
+export function extractHtmlTitle(readabilityTitle, fileName) {
+    if (readabilityTitle && readabilityTitle.trim().length > 0) {
+        return { title: readabilityTitle.trim(), source: 'content' };
+    }
+    // Fall back to file name
+    return { title: fileNameToTitle(fileName), source: 'filename' };
+}
+/**
+ * Extract title from PDF metadata or first page chunk text
+ * Priority: PDF metadata /Title -> first page chunk 0 text -> file name
+ *
+ * Rejects metadata titles that look like file paths (contain / or \) or are empty/whitespace-only.
+ *
+ * @param metadataTitle - PDF metadata /Title value (may be undefined)
+ * @param firstPageChunkText - Text of chunk 0 from semantic chunking of page 1 (may be undefined)
+ * @param fileName - File name for fallback
+ * @param firstPageFontHint - Largest-font text item from page 1 (optional, used for title detection)
+ * @returns Title extraction result
+ */
+export function extractPdfTitle(metadataTitle, firstPageChunkText, fileName, firstPageFontHint) {
+    // 1. Try PDF metadata title (reject file paths and empty values)
+    if (metadataTitle && metadataTitle.trim().length > 0) {
+        const trimmed = metadataTitle.trim();
+        const looksLikeFilePath = trimmed.includes('/') || trimmed.includes('\\');
+        if (!looksLikeFilePath) {
+            return { title: trimmed, source: 'metadata' };
+        }
+    }
+    // 2. Try largest-font text from page 1 (font size > threshold indicates title)
+    if (firstPageFontHint &&
+        firstPageFontHint.fontSize > TITLE_MIN_FONT_SIZE &&
+        firstPageFontHint.text.trim().length > 0) {
+        return { title: firstPageFontHint.text.trim(), source: 'content' };
+    }
+    // 3. Try first chunk from page 1 semantic chunking
+    if (firstPageChunkText && firstPageChunkText.trim().length > 0) {
+        return { title: firstPageChunkText.trim(), source: 'content' };
+    }
+    // 4. Fall back to file name
+    return { title: fileNameToTitle(fileName), source: 'filename' };
+}
+/**
+ * Extract title from a parsed DOCX mammoth HTML document
+ * Priority: DOCX core title -> first non-empty <h1> from mammoth HTML -> file name
+ *
+ * @param document - Parsed HTML document generated from mammoth.convertToHtml()
+ * @param fileName - File name for fallback
+ * @param metadataTitle - Optional title extracted from docProps/core.xml
+ * @returns Title extraction result
+ */
+export function extractDocxTitle(document, fileName, metadataTitle) {
+    const normalizedMetadataTitle = metadataTitle?.replace(/\s+/g, ' ').trim();
+    if (normalizedMetadataTitle) {
+        return { title: normalizedMetadataTitle, source: 'metadata' };
+    }
+    for (const heading of document.querySelectorAll('h1')) {
+        const title = (heading.textContent ?? '').replace(/\s+/g, ' ').trim();
+        if (title)
+            return { title, source: 'content' };
+    }
+    // Fall back to file name
+    return { title: fileNameToTitle(fileName), source: 'filename' };
+}
+//# sourceMappingURL=title-extractor.js.map

package/dist/parser/title-extractor.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"title-extractor.js","sourceRoot":"","sources":["../../src/parser/title-extractor.ts"],"names":[],"mappings":"AAAA,yDAAyD;AACzD,+DAA+D;AAE/D,+CAA+C;AAC/C,YAAY;AACZ,+CAA+C;AAE/C,oFAAoF;AACpF,MAAM,mBAAmB,GAAG,EAAE,CAAA;AAc9B,+CAA+C;AAC/C,gBAAgB;AAChB,+CAA+C;AAE/C;;;;;;GAMG;AACH,MAAM,UAAU,eAAe,CAAC,QAAgB;IAC9C,kDAAkD;IAClD,MAAM,YAAY,GAAG,QAAQ,CAAC,WAAW,CAAC,GAAG,CAAC,CAAA;IAC9C,MAAM,cAAc,GAAG,YAAY,GAAG,CAAC,CAAC,CAAC,CAAC,QAAQ,CAAC,SAAS,CAAC,CAAC,EAAE,YAAY,CAAC,CAAC,CAAC,CAAC,QAAQ,CAAA;IACxF,8CAA8C;IAC9C,OAAO,cAAc,CAAC,OAAO,CAAC,OAAO,EAAE,GAAG,CAAC,CAAA;AAC7C,CAAC;AAED,+CAA+C;AAC/C,wBAAwB;AACxB,+CAA+C;AAE/C;;;;;;;GAOG;AACH,MAAM,UAAU,oBAAoB,CAAC,IAAY,EAAE,QAAgB;IACjE,0BAA0B;IAC1B,MAAM,gBAAgB,GAAG,IAAI,CAAC,KAAK,CAAC,wDAAwD,CAAC,CAAA;IAC7F,IAAI,gBAAgB,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;QAC1B,OAAO,EAAE,KAAK,EAAE,gBAAgB,CAAC,CAAC,CAAC,CAAC,IAAI,EAAE,EAAE,MAAM,EAAE,UAAU,EAAE,CAAA;IAClE,CAAC;IAED,0BAA0B;IAC1B,MAAM,OAAO,GAAG,IAAI,CAAC,KAAK,CAAC,WAAW,CAAC,CAAA;IACvC,IAAI,OAAO,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;QACjB,OAAO,EAAE,KAAK,EAAE,OAAO,CAAC,CAAC,CAAC,CAAC,IAAI,EAAE,EAAE,MAAM,EAAE,SAAS,EAAE,CAAA;IACxD,CAAC;IAED,4BAA4B;IAC5B,OAAO,EAAE,KAAK,EAAE,eAAe,CAAC,QAAQ,CAAC,EAAE,MAAM,EAAE,UAAU,EAAE,CAAA;AACjE,CAAC;AAED;;;;;;;GAOG;AACH,MAAM,UAAU,eAAe,CAAC,IAAY,EAAE,QAAgB;IAC5D,wCAAwC;IACxC,IAAI,IAAI,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;QACpB,MAAM,KAAK,GAAG,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,CAAA;QAC9B,MAAM,SAAS,GAAG,KAAK,CAAC,CAAC,CAAC,CAAA;QAC1B,MAAM,UAAU,GAAG,KAAK,CAAC,CAAC,CAAC,CAAA;QAC3B,IACE,SAAS,KAAK,SAAS;YACvB,UAAU,KAAK,SAAS;YACxB,SAAS,CAAC,IAAI,EAAE,CAAC,MAAM,GAAG,CAAC;YAC3B,UAAU,CAAC,IAAI,EAAE,CAAC,MAAM,KAAK,CAAC,EAC9B,CAAC;YACD,OAAO,EAAE,KAAK,EAAE,SAAS,CAAC,IAAI,EAAE,EAAE,MAAM,EAAE,SAAS,EAAE,CAAA;QACvD,CAAC;IACH,CAAC;IAED,yBAAyB;IACzB,OAAO,EAAE,KAAK,EAAE,eAAe,CAAC,QAAQ,CAAC,EAAE,MAAM,EAAE,UAAU,EAAE,CAAA;AACjE,CAAC;AAED;;;;;;;GAOG;AACH,MAAM,UAAU,gBAAgB,CAC9B,gBAAwB,EACxB,QAAgB;IAEhB,IAAI,gBAAgB,IAAI,gBAAgB,CAAC,IAAI,EAAE,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;QAC3D,OAAO,EAAE,KAAK,EAAE,gBAAgB,CAAC,IAAI,EAAE,EAAE,MAAM,EAAE,SAAS,EAAE,CAAA;IAC9D,CAAC;IAED,yBAAyB;IACzB,OAAO,EAAE,KAAK,EAAE,eAAe,CAAC,QAAQ,CAAC,EAAE,MAAM,EAAE,UAAU,EAAE,CAAA;AACjE,CAAC;AAED;;;;;;;;;;;GAWG;AACH,MAAM,UAAU,eAAe,CAC7B,aAAiC,EACjC,kBAAsC,EACtC,QAAgB,EAChB,iBAAsD;IAEtD,iEAAiE;IACjE,IAAI,aAAa,IAAI,aAAa,CAAC,IAAI,EAAE,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;QACrD,MAAM,OAAO,GAAG,aAAa,CAAC,IAAI,EAAE,CAAA;QACpC,MAAM,iBAAiB,GAAG,OAAO,CAAC,QAAQ,CAAC,GAAG,CAAC,IAAI,OAAO,CAAC,QAAQ,CAAC,IAAI,CAAC,CAAA;QACzE,IAAI,CAAC,iBAAiB,EAAE,CAAC;YACvB,OAAO,EAAE,KAAK,EAAE,OAAO,EAAE,MAAM,EAAE,UAAU,EAAE,CAAA;QAC/C,CAAC;IACH,CAAC;IAED,+EAA+E;IAC/E,IACE,iBAAiB;QACjB,iBAAiB,CAAC,QAAQ,GAAG,mBAAmB;QAChD,iBAAiB,CAAC,IAAI,CAAC,IAAI,EAAE,CAAC,MAAM,GAAG,CAAC,EACxC,CAAC;QACD,OAAO,EAAE,KAAK,EAAE,iBAAiB,CAAC,IAAI,CAAC,IAAI,EAAE,EAAE,MAAM,EAAE,SAAS,EAAE,CAAA;IACpE,CAAC;IAED,mDAAmD;IACnD,IAAI,kBAAkB,IAAI,kBAAkB,CAAC,IAAI,EAAE,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;QAC/D,OAAO,EAAE,KAAK,EAAE,kBAAkB,CAAC,IAAI,EAAE,EAAE,MAAM,EAAE,SAAS,EAAE,CAAA;IAChE,CAAC;IAED,4BAA4B;IAC5B,OAAO,EAAE,KAAK,EAAE,eAAe,CAAC,QAAQ,CAAC,EAAE,MAAM,EAAE,UAAU,EAAE,CAAA;AACjE,CAAC;AAED;;;;;;;;GAQG;AACH,MAAM,UAAU,gBAAgB,CAC9B,QAAkB,EAClB,QAAgB,EAChB,aAAsB;IAEtB,MAAM,uBAAuB,GAAG,aAAa,EAAE,OAAO,CAAC,MAAM,EAAE,GAAG,CAAC,CAAC,IAAI,EAAE,CAAA;IAC1E,IAAI,uBAAuB,EAAE,CAAC;QAC5B,OAAO,EAAE,KAAK,EAAE,uBAAuB,EAAE,MAAM,EAAE,UAAU,EAAE,CAAA;IAC/D,CAAC;IAED,KAAK,MAAM,OAAO,IAAI,QAAQ,CAAC,gBAAgB,CAAC,IAAI,CAAC,EAAE,CAAC;QACtD,MAAM,KAAK,GAAG,CAAC,OAAO,CAAC,WAAW,IAAI,EAAE,CAAC,CAAC,OAAO,CAAC,MAAM,EAAE,GAAG,CAAC,CAAC,IAAI,EAAE,CAAA;QACrE,IAAI,KAAK;YAAE,OAAO,EAAE,KAAK,EAAE,MAAM,EAAE,SAAS,EAAE,CAAA;IAChD,CAAC;IAED,yBAAyB;IACzB,OAAO,EAAE,KAAK,EAAE,eAAe,CAAC,QAAQ,CAAC,EAAE,MAAM,EAAE,UAAU,EAAE,CAAA;AACjE,CAAC"}

package/dist/pdf-visual/captioner.d.ts

@@ -0,0 +1,16 @@
+import type { Captioner, CaptionerConfig } from './types.js';
+/**
+ * Create a captioner for the requested visual-quality profile. Sets
+ * `env.cacheDir` immediately so the global is correct even if the captioner
+ * is constructed before any embedder initializes.
+ *
+ * Concurrency assumption: `env.cacheDir` is a process-global from
+ * `@huggingface/transformers`. Setting it here at construction time is safe
+ * for the current single-instance usage (one captioner per ingest run). If
+ * the codebase ever constructs multiple captioners with DIFFERENT `cacheDir`
+ * values in parallel, the last writer wins and the first captioner's
+ * `from_pretrained` may resolve against the wrong cache. Avoid concurrent
+ * construction with differing cacheDirs.
+ */
+export declare function createCaptioner(config: CaptionerConfig): Captioner;
+//# sourceMappingURL=captioner.d.ts.map

package/dist/pdf-visual/captioner.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"captioner.d.ts","sourceRoot":"","sources":["../../src/pdf-visual/captioner.ts"],"names":[],"mappings":"AAuBA,OAAO,KAAK,EAAE,SAAS,EAAE,eAAe,EAAE,MAAM,YAAY,CAAA;AAE5D;;;;;;;;;;;;GAYG;AACH,wBAAgB,eAAe,CAAC,MAAM,EAAE,eAAe,GAAG,SAAS,CA6BlE"}

package/dist/pdf-visual/captioner.js

@@ -0,0 +1,63 @@
+// Captioner dispatcher for the visual ingest path.
+//
+// `createCaptioner(config)` selects the underlying VLM family based on the
+// `QualityProfile` and returns a `Captioner`. Each profile is implemented as a
+// self-contained module under `./captioners/` so that prompt, chat template,
+// processor signature, generation options, and model class stay coherent per
+// profile.
+//
+// Profiles:
+//   - `fast`    → `captioners/fast.ts`    (SmolVLM-256M-Instruct, IDEFICS3).
+//                 Lightweight default; ~250 MB cache.
+//   - `quality` → `captioners/quality.ts` (Qwen2.5-VL-3B-Instruct-ONNX).
+//                 Higher fidelity on figures with in-image text; ~2.9 GB cache,
+//                 ~2× per-page inference relative to `fast`.
+//
+// `env.cacheDir` is set once here (not inside the per-profile modules) so the
+// shared global is configured before either profile's `from_pretrained` runs
+// and the per-profile modules stay free of the global side effect.
+import { env } from '@huggingface/transformers';
+import { createFastCaptioner } from './captioners/fast.js';
+import { createQualityCaptioner } from './captioners/quality.js';
+/**
+ * Create a captioner for the requested visual-quality profile. Sets
+ * `env.cacheDir` immediately so the global is correct even if the captioner
+ * is constructed before any embedder initializes.
+ *
+ * Concurrency assumption: `env.cacheDir` is a process-global from
+ * `@huggingface/transformers`. Setting it here at construction time is safe
+ * for the current single-instance usage (one captioner per ingest run). If
+ * the codebase ever constructs multiple captioners with DIFFERENT `cacheDir`
+ * values in parallel, the last writer wins and the first captioner's
+ * `from_pretrained` may resolve against the wrong cache. Avoid concurrent
+ * construction with differing cacheDirs.
+ */
+export function createCaptioner(config) {
+    // Defensive ordering: set the global cacheDir at construction so the very
+    // first `from_pretrained` call sees the right value. Setting the same
+    // global twice with the same value is idempotent (shared with the
+    // embedder).
+    env.cacheDir = config.cacheDir;
+    const resolvedDevice = config.device || 'cpu';
+    switch (config.profile) {
+        case 'fast':
+            return createFastCaptioner(resolvedDevice);
+        case 'quality':
+            // No silent fallback to `fast` on load failure. The heavier Qwen2.5-VL
+            // model surfaces its load error as a wrapped `VlmError` per page (see
+            // `processVisualRegions` in `./index.ts`); per FR-3 the file ingest
+            // as a whole still completes text-only, so a misconfigured install
+            // degrades each candidate page rather than masking the misconfig by
+            // switching to `fast`. Operators see one warn line per candidate page.
+            return createQualityCaptioner(resolvedDevice);
+        default: {
+            // Exhaustiveness guard. `QualityProfile` is statically narrow at the
+            // call sites (CLI + MCP both validate before reaching here) so this
+            // branch is unreachable today; the throw is defensive for future
+            // ProfileType additions that forget to extend this switch.
+            const _exhaustive = config.profile;
+            throw new Error(`Unknown QualityProfile: ${_exhaustive}`);
+        }
+    }
+}
+//# sourceMappingURL=captioner.js.map

package/dist/pdf-visual/captioner.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"captioner.js","sourceRoot":"","sources":["../../src/pdf-visual/captioner.ts"],"names":[],"mappings":"AAAA,mDAAmD;AACnD,EAAE;AACF,2EAA2E;AAC3E,+EAA+E;AAC/E,6EAA6E;AAC7E,6EAA6E;AAC7E,WAAW;AACX,EAAE;AACF,YAAY;AACZ,6EAA6E;AAC7E,sDAAsD;AACtD,yEAAyE;AACzE,gFAAgF;AAChF,6DAA6D;AAC7D,EAAE;AACF,8EAA8E;AAC9E,6EAA6E;AAC7E,mEAAmE;AAEnE,OAAO,EAAE,GAAG,EAAE,MAAM,2BAA2B,CAAA;AAE/C,OAAO,EAAE,mBAAmB,EAAE,MAAM,sBAAsB,CAAA;AAC1D,OAAO,EAAE,sBAAsB,EAAE,MAAM,yBAAyB,CAAA;AAGhE;;;;;;;;;;;;GAYG;AACH,MAAM,UAAU,eAAe,CAAC,MAAuB;IACrD,0EAA0E;IAC1E,sEAAsE;IACtE,kEAAkE;IAClE,aAAa;IACb,GAAG,CAAC,QAAQ,GAAG,MAAM,CAAC,QAAQ,CAAA;IAE9B,MAAM,cAAc,GAAG,MAAM,CAAC,MAAM,IAAI,KAAK,CAAA;IAE7C,QAAQ,MAAM,CAAC,OAAO,EAAE,CAAC;QACvB,KAAK,MAAM;YACT,OAAO,mBAAmB,CAAC,cAAc,CAAC,CAAA;QAC5C,KAAK,SAAS;YACZ,uEAAuE;YACvE,sEAAsE;YACtE,oEAAoE;YACpE,mEAAmE;YACnE,oEAAoE;YACpE,uEAAuE;YACvE,OAAO,sBAAsB,CAAC,cAAc,CAAC,CAAA;QAC/C,SAAS,CAAC;YACR,qEAAqE;YACrE,oEAAoE;YACpE,iEAAiE;YACjE,2DAA2D;YAC3D,MAAM,WAAW,GAAU,MAAM,CAAC,OAAO,CAAA;YACzC,MAAM,IAAI,KAAK,CAAC,2BAA2B,WAAqB,EAAE,CAAC,CAAA;QACrE,CAAC;IACH,CAAC;AACH,CAAC"}

package/dist/pdf-visual/captioners/fast.d.ts

@@ -0,0 +1,7 @@
+import type { Captioner } from '../types.js';
+/**
+ * Create a `fast` profile captioner. The dispatcher has already configured
+ * `env.cacheDir`; this profile only owns lazy model loading and inference.
+ */
+export declare function createFastCaptioner(resolvedDevice: string): Captioner;
+//# sourceMappingURL=fast.d.ts.map

package/dist/pdf-visual/captioners/fast.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"fast.d.ts","sourceRoot":"","sources":["../../../src/pdf-visual/captioners/fast.ts"],"names":[],"mappings":"AAiCA,OAAO,KAAK,EAAE,SAAS,EAAE,MAAM,aAAa,CAAA;AAkB5C;;;GAGG;AACH,wBAAgB,mBAAmB,CAAC,cAAc,EAAE,MAAM,GAAG,SAAS,CAkErE"}

package/dist/pdf-visual/captioners/fast.js

@@ -0,0 +1,103 @@
+// `fast` visual-quality profile — SmolVLM-256M-Instruct / IDEFICS3.
+//
+// Shares the profile-agnostic load/decode mechanics with `quality` via
+// `shared.ts` (`createModelLoader`, `decodePngToRawImage`, `VLM_DTYPE`,
+// post-processing). The profile-specific parts — model class, prompt,
+// processor call shape, and generation options — stay here so `fast` and
+// `quality` can diverge on exactly the parts that genuinely differ.
+//
+// Implementation contract (matches HEAD captioner contract steps 2–8):
+//   1. Lazy-load processor + model on first `caption()` call with the pinned
+//      `VLM_DTYPE` and the resolved device. `env.cacheDir` is set by the
+//      dispatcher (`captioner.ts`) before this profile is constructed.
+//   2. Decode PNG bytes via `RawImage.fromBlob(new Blob([pngBytes], { type:
+//      'image/png' }))`. SmolVLM-256M is NOT resized client-side — IDEFICS3's
+//      processor handles its own preprocessing.
+//   3. Build chat-style input via `processor.apply_chat_template(messages,
+//      { add_generation_prompt: true })` with the IDEFICS3 conversation shape
+//      `[{role:'user', content:[{type:'image'},{type:'text',text:...}]}]`.
+//      Probe-verified.
+//   4. Call `model.generate({ ...inputs, max_new_tokens: 128,
+//      repetition_penalty: 1.15, no_repeat_ngram_size: 3 })` — these
+//      generation options are tuned for SmolVLM-256M and MUST stay on the
+//      `fast` profile (they cause forced variant generation on Qwen2.5-VL,
+//      which is why the `quality` profile drops them).
+//   5. Decode via `processor.batch_decode(newTokens, { skip_special_tokens: true })`
+//      where `newTokens = outputs.slice(null, [inputs.input_ids.dims[1], null])`.
+//   6. Post-processing via `shared.postProcess` (control-char strip, trim,
+//      empty → `null`, length > 1000 → truncate + `…`).
+//   7. On model load / image decode / generation failure throw `VlmError`
+//      with `pageNum` + `cause`.
+import { AutoModelForImageTextToText, AutoProcessor } from '@huggingface/transformers';
+import { VlmError } from '../types.js';
+import { createModelLoader, decodePngToRawImage, postProcess } from './shared.js';
+const MODEL_NAME = 'HuggingFaceTB/SmolVLM-256M-Instruct';
+/**
+ * Static prompt — tuned for "describe for search retrieval" not "describe for
+ * a blind reader". It asks for retrieval-relevant detail rather than a short
+ * summary, while keeping claims grounded in visible evidence.
+ */
+const PROMPT = 'Write search text for this PDF page image. Include visible section names, visual titles, ' +
+    'headings, labels, legends, axes, row or column names, UI text, metric names, identifiers, ' +
+    'proper nouns, and flow or diagram step names. Prefer exact readable words from the image. ' +
+    'Cover the main visual regions across the page. Use short searchable phrases separated by ' +
+    'commas or semicolons. Use only readable or visually evident details. Use each phrase once.';
+/**
+ * Create a `fast` profile captioner. The dispatcher has already configured
+ * `env.cacheDir`; this profile only owns lazy model loading and inference.
+ */
+export function createFastCaptioner(resolvedDevice) {
+    // Both classes accept `{ dtype }` (probe-verified). They load in sequence
+    // because the second resolves the runtime class
+    // (`Idefics3ForConditionalGeneration` for the default model) via the
+    // architecture-agnostic `AutoModelForImageTextToText` entry point.
+    const loader = createModelLoader(MODEL_NAME, resolvedDevice, async ({ dtypeOpt, modelOpt }) => {
+        const processor = await AutoProcessor.from_pretrained(MODEL_NAME, dtypeOpt);
+        const model = await AutoModelForImageTextToText.from_pretrained(MODEL_NAME, modelOpt);
+        return { processor, model };
+    });
+    return {
+        async caption(pngBytes, pageNum) {
+            try {
+                const { processor, model } = await loader.ensureLoaded();
+                // Decode PNG → RawImage. SmolVLM-256M is NOT resized client-side —
+                // IDEFICS3's processor handles its own preprocessing.
+                const rawImage = await decodePngToRawImage(pngBytes);
+                // Build chat-style input. The IDEFICS3 conversation shape is
+                // probe-verified for `Idefics3Processor.apply_chat_template`.
+                const messages = [
+                    {
+                        role: 'user',
+                        content: [{ type: 'image' }, { type: 'text', text: PROMPT }],
+                    },
+                ];
+                // The processor and model are dynamic in type at the boundary;
+                // narrow to a minimal callable / generate-able shape here. IDEFICS3
+                // processor takes an array of images.
+                const proc = processor;
+                const mdl = model;
+                const chatPrompt = proc.apply_chat_template(messages, { add_generation_prompt: true });
+                const inputs = await proc(chatPrompt, [rawImage]);
+                const outputs = await mdl.generate({
+                    ...inputs,
+                    max_new_tokens: 128,
+                    repetition_penalty: 1.15,
+                    no_repeat_ngram_size: 3,
+                });
+                // `outputs.slice(null, [inputLen, null])` strips the prompt tokens.
+                const inputLen = inputs.input_ids.dims[1];
+                const newTokens = outputs.slice(null, [inputLen, null]);
+                const decoded = proc.batch_decode(newTokens, { skip_special_tokens: true });
+                const text = decoded[0] ?? '';
+                return postProcess(text);
+            }
+            catch (err) {
+                if (err instanceof VlmError)
+                    throw err;
+                const cause = err instanceof Error ? err : new Error(String(err));
+                throw new VlmError(`Captioning failed for page ${pageNum}`, { cause, pageNum });
+            }
+        },
+    };
+}
+//# sourceMappingURL=fast.js.map

package/dist/pdf-visual/captioners/fast.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"fast.js","sourceRoot":"","sources":["../../../src/pdf-visual/captioners/fast.ts"],"names":[],"mappings":"AAAA,oEAAoE;AACpE,EAAE;AACF,uEAAuE;AACvE,wEAAwE;AACxE,sEAAsE;AACtE,yEAAyE;AACzE,oEAAoE;AACpE,EAAE;AACF,uEAAuE;AACvE,6EAA6E;AAC7E,yEAAyE;AACzE,uEAAuE;AACvE,4EAA4E;AAC5E,8EAA8E;AAC9E,gDAAgD;AAChD,2EAA2E;AAC3E,8EAA8E;AAC9E,2EAA2E;AAC3E,uBAAuB;AACvB,8DAA8D;AAC9D,qEAAqE;AACrE,0EAA0E;AAC1E,2EAA2E;AAC3E,uDAAuD;AACvD,qFAAqF;AACrF,kFAAkF;AAClF,2EAA2E;AAC3E,wDAAwD;AACxD,0EAA0E;AAC1E,iCAAiC;AAEjC,OAAO,EAAE,2BAA2B,EAAE,aAAa,EAAE,MAAM,2BAA2B,CAAA;AAGtF,OAAO,EAAE,QAAQ,EAAE,MAAM,aAAa,CAAA;AACtC,OAAO,EAAE,iBAAiB,EAAE,mBAAmB,EAAE,WAAW,EAAE,MAAM,aAAa,CAAA;AAEjF,MAAM,UAAU,GAAG,qCAAqC,CAAA;AAExD;;;;GAIG;AACH,MAAM,MAAM,GACV,2FAA2F;IAC3F,4FAA4F;IAC5F,4FAA4F;IAC5F,2FAA2F;IAC3F,4FAA4F,CAAA;AAE9F;;;GAGG;AACH,MAAM,UAAU,mBAAmB,CAAC,cAAsB;IACxD,0EAA0E;IAC1E,gDAAgD;IAChD,qEAAqE;IACrE,mEAAmE;IACnE,MAAM,MAAM,GAAG,iBAAiB,CAAC,UAAU,EAAE,cAAc,EAAE,KAAK,EAAE,EAAE,QAAQ,EAAE,QAAQ,EAAE,EAAE,EAAE;QAC5F,MAAM,SAAS,GAAG,MAAM,aAAa,CAAC,eAAe,CAAC,UAAU,EAAE,QAAQ,CAAC,CAAA;QAC3E,MAAM,KAAK,GAAG,MAAM,2BAA2B,CAAC,eAAe,CAAC,UAAU,EAAE,QAAQ,CAAC,CAAA;QACrF,OAAO,EAAE,SAAS,EAAE,KAAK,EAAE,CAAA;IAC7B,CAAC,CAAC,CAAA;IAEF,OAAO;QACL,KAAK,CAAC,OAAO,CAAC,QAAoB,EAAE,OAAe;YACjD,IAAI,CAAC;gBACH,MAAM,EAAE,SAAS,EAAE,KAAK,EAAE,GAAG,MAAM,MAAM,CAAC,YAAY,EAAE,CAAA;gBAExD,mEAAmE;gBACnE,sDAAsD;gBACtD,MAAM,QAAQ,GAAG,MAAM,mBAAmB,CAAC,QAAQ,CAAC,CAAA;gBAEpD,6DAA6D;gBAC7D,8DAA8D;gBAC9D,MAAM,QAAQ,GAAG;oBACf;wBACE,IAAI,EAAE,MAAM;wBACZ,OAAO,EAAE,CAAC,EAAE,IAAI,EAAE,OAAO,EAAE,EAAE,EAAE,IAAI,EAAE,MAAM,EAAE,IAAI,EAAE,MAAM,EAAE,CAAC;qBAC7D;iBACF,CAAA;gBACD,+DAA+D;gBAC/D,oEAAoE;gBACpE,sCAAsC;gBACtC,MAAM,IAAI,GAAG,SAG0E,CAAA;gBACvF,MAAM,GAAG,GAAG,KAIX,CAAA;gBAED,MAAM,UAAU,GAAG,IAAI,CAAC,mBAAmB,CAAC,QAAQ,EAAE,EAAE,qBAAqB,EAAE,IAAI,EAAE,CAAC,CAAA;gBACtF,MAAM,MAAM,GAAG,MAAM,IAAI,CAAC,UAAU,EAAE,CAAC,QAAQ,CAAC,CAAC,CAAA;gBAEjD,MAAM,OAAO,GAAG,MAAM,GAAG,CAAC,QAAQ,CAAC;oBACjC,GAAG,MAAM;oBACT,cAAc,EAAE,GAAG;oBACnB,kBAAkB,EAAE,IAAI;oBACxB,oBAAoB,EAAE,CAAC;iBACxB,CAAC,CAAA;gBAEF,oEAAoE;gBACpE,MAAM,QAAQ,GAAG,MAAM,CAAC,SAAS,CAAC,IAAI,CAAC,CAAC,CAAW,CAAA;gBACnD,MAAM,SAAS,GAAG,OAAO,CAAC,KAAK,CAAC,IAAI,EAAE,CAAC,QAAQ,EAAE,IAAI,CAAC,CAAC,CAAA;gBAEvD,MAAM,OAAO,GAAG,IAAI,CAAC,YAAY,CAAC,SAAS,EAAE,EAAE,mBAAmB,EAAE,IAAI,EAAE,CAAC,CAAA;gBAC3E,MAAM,IAAI,GAAG,OAAO,CAAC,CAAC,CAAC,IAAI,EAAE,CAAA;gBAE7B,OAAO,WAAW,CAAC,IAAI,CAAC,CAAA;YAC1B,CAAC;YAAC,OAAO,GAAG,EAAE,CAAC;gBACb,IAAI,GAAG,YAAY,QAAQ;oBAAE,MAAM,GAAG,CAAA;gBACtC,MAAM,KAAK,GAAG,GAAG,YAAY,KAAK,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,IAAI,KAAK,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAA;gBACjE,MAAM,IAAI,QAAQ,CAAC,8BAA8B,OAAO,EAAE,EAAE,EAAE,KAAK,EAAE,OAAO,EAAE,CAAC,CAAA;YACjF,CAAC;QACH,CAAC;KACF,CAAA;AACH,CAAC"}

package/dist/pdf-visual/captioners/quality.d.ts

@@ -0,0 +1,7 @@
+import type { Captioner } from '../types.js';
+/**
+ * Create a `quality` profile captioner. The dispatcher has already configured
+ * `env.cacheDir`; this profile only owns lazy model loading and inference.
+ */
+export declare function createQualityCaptioner(resolvedDevice: string): Captioner;
+//# sourceMappingURL=quality.d.ts.map

package/dist/pdf-visual/captioners/quality.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"quality.d.ts","sourceRoot":"","sources":["../../../src/pdf-visual/captioners/quality.ts"],"names":[],"mappings":"AAsCA,OAAO,KAAK,EAAE,SAAS,EAAE,MAAM,aAAa,CAAA;AAkC5C;;;GAGG;AACH,wBAAgB,sBAAsB,CAAC,cAAc,EAAE,MAAM,GAAG,SAAS,CAwExE"}

package/dist/pdf-visual/captioners/quality.js

@@ -0,0 +1,127 @@
+// `quality` visual-quality profile — Qwen2.5-VL-3B-Instruct-ONNX.
+//
+// Verbatim port of the working-tree captioner validated during the visual-
+// quality design discussion. Higher fidelity than `fast` on figures with
+// in-image text (axis labels, panel sub-labels, annotations), at the cost of
+// a materially larger model cache (~10× `fast`) and ~2× per-page inference
+// time on CPU. Shares the profile-agnostic load/decode mechanics with `fast`
+// via `shared.ts`; keeps its own model class, prompt, resize, processor call
+// shape, and generation options.
+//
+// Implementation contract:
+//   1. Lazy-load processor + model on first `caption()` call with the pinned
+//      DTYPE and the resolved device. `env.cacheDir` is set by the dispatcher
+//      (`captioner.ts`) before this profile is constructed.
+//   2. Decode PNG bytes via `RawImage.fromBlob(...)` and resize to 448x448 —
+//      matches the onnx-community Qwen2-VL reference example. Qwen2.5-VL
+//      supports dynamic resolution natively, but the reference example uses
+//      a fixed resize for stable behavior.
+//   3. Build chat-style input via `processor.apply_chat_template(messages,
+//      { add_generation_prompt: true })` with the Qwen2.5-VL conversation
+//      shape `[{role:'user', content:[{type:'image'},{type:'text',text:...}]}]`.
+//      The shape is hard-coded against the Qwen2.5-VL family; swapping in a
+//      non-Qwen-VL model requires a new profile.
+//   4. Call `model.generate({ ...inputs, max_new_tokens: 128 })`. The
+//      `repetition_penalty` / `no_repeat_ngram_size` options used by `fast`
+//      are intentionally absent — on Qwen2.5-VL they cause forced variant
+//      generation (e.g. "Cycles per cycle" → "Cpu cycles/cycle" → "Cnt
+//      cyles/clk") whenever a figure naturally repeats a phrase.
+//   5. Decode via `processor.batch_decode(newTokens, { skip_special_tokens: true })`
+//      where `newTokens = outputs.slice(null, [inputs.input_ids.dims.at(-1), null])`.
+//      `dims.at(-1)` reads the last dimension defensively — matches the
+//      onnx-community reference example.
+//   6. Post-processing via `shared.postProcess` (same pipeline as `fast`).
+//   7. On model load / image decode / generation failure throw `VlmError`
+//      with `pageNum` + `cause`. No silent fallback to `fast`.
+import { AutoProcessor, Qwen2_5_VLForConditionalGeneration } from '@huggingface/transformers';
+import { VlmError } from '../types.js';
+import { createModelLoader, decodePngToRawImage, postProcess } from './shared.js';
+const MODEL_NAME = 'onnx-community/Qwen2.5-VL-3B-Instruct-ONNX';
+/**
+ * Fixed input resolution (px) for the Qwen2.5-VL reference resize. Matches the
+ * onnx-community Qwen2-VL example's stable-behavior fixed resize.
+ */
+const QWEN_INPUT_SIZE = 448;
+/**
+ * Static prompt — tuned for retrieval search indexing. Asks the VLM to scan
+ * the whole image before composing output so coverage spans every region,
+ * then produces a two-part response (Summary + Keywords) suitable for
+ * embedding + downstream semantic / lexical search. No length specifiers —
+ * output length is controlled by `max_new_tokens` because length specs
+ * narrow coverage.
+ */
+const PROMPT = `Describe this PDF page image for retrieval search indexing.
+
+Procedure:
+1. Scan the whole image and identify every distinct region.
+2. Compose the output from across all regions identified.
+
+Output exactly two parts:
+
+Summary: Describe the page's content, including its type and subject when identifiable.
+
+Keywords: Phrases separated by semicolons. Capture readable text and visible labels from across the page — including section titles, sub-labels inside figures, tables, panels, or annotations. Use exact wording from the image when readable. Cover the visible regions of the page. List each phrase once.
+
+Use only details visible in the image. If a region is unreadable, skip it.`;
+/**
+ * Create a `quality` profile captioner. The dispatcher has already configured
+ * `env.cacheDir`; this profile only owns lazy model loading and inference.
+ */
+export function createQualityCaptioner(resolvedDevice) {
+    // The explicit `Qwen2_5_VLForConditionalGeneration` class matches the
+    // onnx-community reference example (rather than the architecture-agnostic
+    // AutoModelForImageTextToText entry point used by `fast`).
+    const loader = createModelLoader(MODEL_NAME, resolvedDevice, async ({ dtypeOpt, modelOpt }) => {
+        const processor = await AutoProcessor.from_pretrained(MODEL_NAME, dtypeOpt);
+        const model = await Qwen2_5_VLForConditionalGeneration.from_pretrained(MODEL_NAME, modelOpt);
+        return { processor, model };
+    });
+    return {
+        async caption(pngBytes, pageNum) {
+            try {
+                const { processor, model } = await loader.ensureLoaded();
+                // Decode PNG → RawImage, then resize to 448x448 to match the
+                // onnx-community Qwen2-VL reference example. Qwen2.5-VL supports dynamic
+                // resolution natively, but the reference example uses a fixed resize for
+                // stable behavior; revisit if small in-figure text is lost.
+                const rawImage = await (await decodePngToRawImage(pngBytes)).resize(QWEN_INPUT_SIZE, QWEN_INPUT_SIZE);
+                // Build chat-style input. The Qwen2.5-VL conversation shape mirrors
+                // the onnx-community Qwen2-VL reference: a single user turn with an
+                // image placeholder followed by the text prompt.
+                const messages = [
+                    {
+                        role: 'user',
+                        content: [{ type: 'image' }, { type: 'text', text: PROMPT }],
+                    },
+                ];
+                // The processor and model are dynamic in type at the boundary;
+                // narrow to a minimal callable / generate-able shape here. Qwen2.5-VL
+                // processor takes a single image (not an array), per the
+                // onnx-community reference example signature `processor(text, image)`.
+                const proc = processor;
+                const mdl = model;
+                const chatPrompt = proc.apply_chat_template(messages, { add_generation_prompt: true });
+                const inputs = await proc(chatPrompt, rawImage);
+                const outputs = await mdl.generate({
+                    ...inputs,
+                    max_new_tokens: 128,
+                });
+                // `outputs.slice(null, [inputLen, null])` strips the prompt tokens.
+                // `dims.at(-1)` reads the last dimension defensively — matches the
+                // onnx-community reference example.
+                const inputLen = inputs.input_ids.dims.at(-1);
+                const newTokens = outputs.slice(null, [inputLen, null]);
+                const decoded = proc.batch_decode(newTokens, { skip_special_tokens: true });
+                const text = decoded[0] ?? '';
+                return postProcess(text);
+            }
+            catch (err) {
+                if (err instanceof VlmError)
+                    throw err;
+                const cause = err instanceof Error ? err : new Error(String(err));
+                throw new VlmError(`Captioning failed for page ${pageNum}`, { cause, pageNum });
+            }
+        },
+    };
+}
+//# sourceMappingURL=quality.js.map

package/dist/pdf-visual/captioners/quality.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"quality.js","sourceRoot":"","sources":["../../../src/pdf-visual/captioners/quality.ts"],"names":[],"mappings":"AAAA,kEAAkE;AAClE,EAAE;AACF,2EAA2E;AAC3E,yEAAyE;AACzE,6EAA6E;AAC7E,2EAA2E;AAC3E,6EAA6E;AAC7E,6EAA6E;AAC7E,iCAAiC;AACjC,EAAE;AACF,2BAA2B;AAC3B,6EAA6E;AAC7E,8EAA8E;AAC9E,4DAA4D;AAC5D,6EAA6E;AAC7E,yEAAyE;AACzE,4EAA4E;AAC5E,2CAA2C;AAC3C,2EAA2E;AAC3E,0EAA0E;AAC1E,iFAAiF;AACjF,4EAA4E;AAC5E,iDAAiD;AACjD,sEAAsE;AACtE,4EAA4E;AAC5E,0EAA0E;AAC1E,uEAAuE;AACvE,iEAAiE;AACjE,qFAAqF;AACrF,sFAAsF;AACtF,wEAAwE;AACxE,yCAAyC;AACzC,2EAA2E;AAC3E,0EAA0E;AAC1E,+DAA+D;AAE/D,OAAO,EAAE,aAAa,EAAE,kCAAkC,EAAE,MAAM,2BAA2B,CAAA;AAG7F,OAAO,EAAE,QAAQ,EAAE,MAAM,aAAa,CAAA;AACtC,OAAO,EAAE,iBAAiB,EAAE,mBAAmB,EAAE,WAAW,EAAE,MAAM,aAAa,CAAA;AAEjF,MAAM,UAAU,GAAG,4CAA4C,CAAA;AAE/D;;;GAGG;AACH,MAAM,eAAe,GAAG,GAAG,CAAA;AAE3B;;;;;;;GAOG;AACH,MAAM,MAAM,GAAG;;;;;;;;;;;;2EAY4D,CAAA;AAE3E;;;GAGG;AACH,MAAM,UAAU,sBAAsB,CAAC,cAAsB;IAC3D,sEAAsE;IACtE,0EAA0E;IAC1E,2DAA2D;IAC3D,MAAM,MAAM,GAAG,iBAAiB,CAAC,UAAU,EAAE,cAAc,EAAE,KAAK,EAAE,EAAE,QAAQ,EAAE,QAAQ,EAAE,EAAE,EAAE;QAC5F,MAAM,SAAS,GAAG,MAAM,aAAa,CAAC,eAAe,CAAC,UAAU,EAAE,QAAQ,CAAC,CAAA;QAC3E,MAAM,KAAK,GAAG,MAAM,kCAAkC,CAAC,eAAe,CAAC,UAAU,EAAE,QAAQ,CAAC,CAAA;QAC5F,OAAO,EAAE,SAAS,EAAE,KAAK,EAAE,CAAA;IAC7B,CAAC,CAAC,CAAA;IAEF,OAAO;QACL,KAAK,CAAC,OAAO,CAAC,QAAoB,EAAE,OAAe;YACjD,IAAI,CAAC;gBACH,MAAM,EAAE,SAAS,EAAE,KAAK,EAAE,GAAG,MAAM,MAAM,CAAC,YAAY,EAAE,CAAA;gBAExD,6DAA6D;gBAC7D,yEAAyE;gBACzE,yEAAyE;gBACzE,4DAA4D;gBAC5D,MAAM,QAAQ,GAAG,MAAM,CAAC,MAAM,mBAAmB,CAAC,QAAQ,CAAC,CAAC,CAAC,MAAM,CACjE,eAAe,EACf,eAAe,CAChB,CAAA;gBAED,oEAAoE;gBACpE,oEAAoE;gBACpE,iDAAiD;gBACjD,MAAM,QAAQ,GAAG;oBACf;wBACE,IAAI,EAAE,MAAM;wBACZ,OAAO,EAAE,CAAC,EAAE,IAAI,EAAE,OAAO,EAAE,EAAE,EAAE,IAAI,EAAE,MAAM,EAAE,IAAI,EAAE,MAAM,EAAE,CAAC;qBAC7D;iBACF,CAAA;gBACD,+DAA+D;gBAC/D,sEAAsE;gBACtE,yDAAyD;gBACzD,uEAAuE;gBACvE,MAAM,IAAI,GAAG,SAGuE,CAAA;gBACpF,MAAM,GAAG,GAAG,KAIX,CAAA;gBAED,MAAM,UAAU,GAAG,IAAI,CAAC,mBAAmB,CAAC,QAAQ,EAAE,EAAE,qBAAqB,EAAE,IAAI,EAAE,CAAC,CAAA;gBACtF,MAAM,MAAM,GAAG,MAAM,IAAI,CAAC,UAAU,EAAE,QAAQ,CAAC,CAAA;gBAE/C,MAAM,OAAO,GAAG,MAAM,GAAG,CAAC,QAAQ,CAAC;oBACjC,GAAG,MAAM;oBACT,cAAc,EAAE,GAAG;iBACpB,CAAC,CAAA;gBAEF,oEAAoE;gBACpE,mEAAmE;gBACnE,oCAAoC;gBACpC,MAAM,QAAQ,GAAG,MAAM,CAAC,SAAS,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC,CAAC,CAAW,CAAA;gBACvD,MAAM,SAAS,GAAG,OAAO,CAAC,KAAK,CAAC,IAAI,EAAE,CAAC,QAAQ,EAAE,IAAI,CAAC,CAAC,CAAA;gBAEvD,MAAM,OAAO,GAAG,IAAI,CAAC,YAAY,CAAC,SAAS,EAAE,EAAE,mBAAmB,EAAE,IAAI,EAAE,CAAC,CAAA;gBAC3E,MAAM,IAAI,GAAG,OAAO,CAAC,CAAC,CAAC,IAAI,EAAE,CAAA;gBAE7B,OAAO,WAAW,CAAC,IAAI,CAAC,CAAA;YAC1B,CAAC;YAAC,OAAO,GAAG,EAAE,CAAC;gBACb,IAAI,GAAG,YAAY,QAAQ;oBAAE,MAAM,GAAG,CAAA;gBACtC,MAAM,KAAK,GAAG,GAAG,YAAY,KAAK,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,IAAI,KAAK,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAA;gBACjE,MAAM,IAAI,QAAQ,CAAC,8BAA8B,OAAO,EAAE,EAAE,EAAE,KAAK,EAAE,OAAO,EAAE,CAAC,CAAA;YACjF,CAAC;QACH,CAAC;KACF,CAAA;AACH,CAAC"}

package/dist/pdf-visual/captioners/shared.d.ts

@@ -0,0 +1,44 @@
+import { type DeviceType, RawImage } from '@huggingface/transformers';
+/**
+ * Build the `from_pretrained` option objects (processor + model) with the
+ * pinned dtype and resolved device. transformers.js declares `dtype` as a
+ * literal union; cast through `unknown` to widen-to-string-then-back.
+ */
+export declare function buildModelLoadOptions(resolvedDevice: string): {
+    dtypeOpt: {
+        dtype: 'q4';
+    };
+    modelOpt: {
+        dtype: 'q4';
+        device: DeviceType;
+    };
+};
+/** The processor + model pair produced by a profile's load callback. */
+export interface LoadedModel {
+    processor: unknown;
+    model: unknown;
+}
+/**
+ * Lazy model loader shared by both profiles. Encapsulates the
+ * pending→ok/failed state machine and the identical load-failure wrapping
+ * (`Captioner load failed (modelName=..., device=...)`). The per-profile
+ * `load` callback owns the model-class choice and receives the shared option
+ * objects. On first `ensureLoaded()` the model loads; subsequent calls return
+ * the cached pair; a prior failure re-throws the same wrapped error.
+ */
+export declare function createModelLoader(modelName: string, resolvedDevice: string, load: (opts: ReturnType<typeof buildModelLoadOptions>) => Promise<LoadedModel>): {
+    ensureLoaded: () => Promise<LoadedModel>;
+};
+/**
+ * Decode PNG bytes to a `RawImage`. `Blob` accepts `Uint8Array` directly (the
+ * renderer returns `Uint8Array` from `Pixmap.asPNG()`), but the `BlobPart`
+ * type omits `Uint8Array<ArrayBufferLike>` due to SharedArrayBuffer subtyping;
+ * cast through `unknown`. Profiles needing a fixed input size resize the result.
+ */
+export declare function decodePngToRawImage(pngBytes: Uint8Array): Promise<RawImage>;
+/**
+ * Apply the post-generation processing rules. Returns the final caption or
+ * `null` when the result is empty after stripping.
+ */
+export declare function postProcess(decoded: string): string | null;
+//# sourceMappingURL=shared.d.ts.map

package/dist/pdf-visual/captioners/shared.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"shared.d.ts","sourceRoot":"","sources":["../../../src/pdf-visual/captioners/shared.ts"],"names":[],"mappings":"AAaA,OAAO,EAAE,KAAK,UAAU,EAAE,QAAQ,EAAE,MAAM,2BAA2B,CAAA;AAWrE;;;;GAIG;AACH,wBAAgB,qBAAqB,CAAC,cAAc,EAAE,MAAM,GAAG;IAC7D,QAAQ,EAAE;QAAE,KAAK,EAAE,IAAI,CAAA;KAAE,CAAA;IACzB,QAAQ,EAAE;QAAE,KAAK,EAAE,IAAI,CAAC;QAAC,MAAM,EAAE,UAAU,CAAA;KAAE,CAAA;CAC9C,CAOA;AAED,wEAAwE;AACxE,MAAM,WAAW,WAAW;IAC1B,SAAS,EAAE,OAAO,CAAA;IAClB,KAAK,EAAE,OAAO,CAAA;CACf;AAED;;;;;;;GAOG;AACH,wBAAgB,iBAAiB,CAC/B,SAAS,EAAE,MAAM,EACjB,cAAc,EAAE,MAAM,EACtB,IAAI,EAAE,CAAC,IAAI,EAAE,UAAU,CAAC,OAAO,qBAAqB,CAAC,KAAK,OAAO,CAAC,WAAW,CAAC,GAC7E;IAAE,YAAY,EAAE,MAAM,OAAO,CAAC,WAAW,CAAC,CAAA;CAAE,CAuB9C;AAED;;;;;GAKG;AACH,wBAAsB,mBAAmB,CAAC,QAAQ,EAAE,UAAU,GAAG,OAAO,CAAC,QAAQ,CAAC,CAGjF;AAwBD;;;GAGG;AACH,wBAAgB,WAAW,CAAC,OAAO,EAAE,MAAM,GAAG,MAAM,GAAG,IAAI,CAK1D"}