@heripo/pdf-parser 0.1.4 → 0.1.6

package/dist/vlm-models.cjs

@@ -0,0 +1,147 @@
+"use strict";
+var __defProp = Object.defineProperty;
+var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
+var __getOwnPropNames = Object.getOwnPropertyNames;
+var __hasOwnProp = Object.prototype.hasOwnProperty;
+var __export = (target, all) => {
+  for (var name in all)
+    __defProp(target, name, { get: all[name], enumerable: true });
+};
+var __copyProps = (to, from, except, desc) => {
+  if (from && typeof from === "object" || typeof from === "function") {
+    for (let key of __getOwnPropNames(from))
+      if (!__hasOwnProp.call(to, key) && key !== except)
+        __defProp(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable });
+  }
+  return to;
+};
+var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod);
+
+// src/config/vlm-models.ts
+var vlm_models_exports = {};
+__export(vlm_models_exports, {
+  DEFAULT_VLM_MODEL: () => DEFAULT_VLM_MODEL,
+  VLM_MODELS: () => VLM_MODELS,
+  resolveVlmModel: () => resolveVlmModel
+});
+module.exports = __toCommonJS(vlm_models_exports);
+var VLM_MODELS = {
+  // ── DocTags models (specialized document structure output) ──────────
+  "granite-docling-258M-mlx": {
+    repo_id: "ibm-granite/granite-docling-258M-mlx",
+    inference_framework: "mlx",
+    response_format: "doctags",
+    transformers_model_type: "automodel-vision2seq",
+    description: "Granite Docling 258M (MLX, Apple Silicon optimized, ~6s/page)"
+  },
+  "granite-docling-258M": {
+    repo_id: "ibm-granite/granite-docling-258M",
+    inference_framework: "transformers",
+    response_format: "doctags",
+    transformers_model_type: "automodel-vision2seq",
+    description: "Granite Docling 258M (Transformers, cross-platform)"
+  },
+  "smoldocling-256M-mlx": {
+    repo_id: "docling-project/SmolDocling-256M-preview-mlx-bf16",
+    inference_framework: "mlx",
+    response_format: "doctags",
+    transformers_model_type: "automodel-vision2seq",
+    description: "SmolDocling 256M (MLX, fastest option)"
+  },
+  "smoldocling-256M": {
+    repo_id: "docling-project/SmolDocling-256M-preview",
+    inference_framework: "transformers",
+    response_format: "doctags",
+    transformers_model_type: "automodel-vision2seq",
+    description: "SmolDocling 256M (Transformers)"
+  },
+  // ── Markdown models (general-purpose vision LLMs) ──────────────────
+  "granite-vision-2B": {
+    repo_id: "ibm-granite/granite-vision-3.2-2b",
+    inference_framework: "transformers",
+    response_format: "markdown",
+    transformers_model_type: "automodel-vision2seq",
+    description: "Granite Vision 3.2 2B (IBM, higher accuracy)"
+  },
+  "qwen25-vl-3B-mlx": {
+    repo_id: "mlx-community/Qwen2.5-VL-3B-Instruct-bf16",
+    inference_framework: "mlx",
+    response_format: "markdown",
+    transformers_model_type: "automodel-vision2seq",
+    description: "Qwen 2.5 VL 3B (MLX, multilingual, good KCJ support)"
+  },
+  phi4: {
+    repo_id: "microsoft/Phi-4-multimodal-instruct",
+    inference_framework: "transformers",
+    response_format: "markdown",
+    transformers_model_type: "automodel",
+    description: "Phi-4 Multimodal (Microsoft, CausalLM)"
+  },
+  "pixtral-12B-mlx": {
+    repo_id: "mlx-community/pixtral-12b-bf16",
+    inference_framework: "mlx",
+    response_format: "markdown",
+    transformers_model_type: "automodel-vision2seq",
+    description: "Pixtral 12B (MLX, Mistral, high accuracy)"
+  },
+  "pixtral-12B": {
+    repo_id: "mistral-community/pixtral-12b",
+    inference_framework: "transformers",
+    response_format: "markdown",
+    transformers_model_type: "automodel-vision2seq",
+    description: "Pixtral 12B (Transformers, Mistral)"
+  },
+  got2: {
+    repo_id: "stepfun-ai/GOT-OCR-2.0-hf",
+    inference_framework: "transformers",
+    response_format: "markdown",
+    transformers_model_type: "automodel-vision2seq",
+    description: "GOT-OCR 2.0 (StepFun, OCR-specialized)"
+  },
+  "gemma3-12B-mlx": {
+    repo_id: "mlx-community/gemma-3-12b-it-bf16",
+    inference_framework: "mlx",
+    response_format: "markdown",
+    transformers_model_type: "automodel-vision2seq",
+    description: "Gemma 3 12B (MLX, Google)"
+  },
+  "gemma3-27B-mlx": {
+    repo_id: "mlx-community/gemma-3-27b-it-bf16",
+    inference_framework: "mlx",
+    response_format: "markdown",
+    transformers_model_type: "automodel-vision2seq",
+    description: "Gemma 3 27B (MLX, Google, highest accuracy)"
+  },
+  dolphin: {
+    repo_id: "ByteDance/Dolphin",
+    inference_framework: "transformers",
+    response_format: "markdown",
+    transformers_model_type: "automodel-vision2seq",
+    description: "Dolphin (ByteDance, document-oriented)"
+  }
+};
+var DEFAULT_VLM_MODEL = "granite-docling-258M-mlx";
+function resolveVlmModel(model) {
+  if (typeof model === "string") {
+    const preset = VLM_MODELS[model];
+    if (!preset) {
+      throw new Error(
+        `Unknown VLM model preset: "${model}". Available presets: ${Object.keys(VLM_MODELS).join(", ")}`
+      );
+    }
+    return {
+      repo_id: preset.repo_id,
+      inference_framework: preset.inference_framework,
+      response_format: preset.response_format,
+      transformers_model_type: preset.transformers_model_type
+    };
+  }
+  return model;
+}
+// Annotate the CommonJS export names for ESM import in node:
+0 && (module.exports = {
+  DEFAULT_VLM_MODEL,
+  VLM_MODELS,
+  resolveVlmModel
+});
+//# sourceMappingURL=vlm-models.cjs.map

package/dist/vlm-models.cjs.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../src/config/vlm-models.ts"],"sourcesContent":["import type { VlmModelLocal } from 'docling-sdk';\n\n/**\n * VLM model preset with description\n */\nexport interface VlmModelPreset {\n  repo_id: string;\n  inference_framework: 'mlx' | 'transformers';\n  response_format: 'doctags' | 'markdown';\n  transformers_model_type: 'automodel-vision2seq' | 'automodel';\n  description: string;\n}\n\n/**\n * Available VLM model presets\n *\n * Based on Docling's official VLM model specs:\n * https://docling-project.github.io/docling/usage/vision_models/#available-local-models\n *\n * Users can select a preset key or provide a custom VlmModelLocal object.\n */\nexport const VLM_MODELS: Record<string, VlmModelPreset> = {\n  // ── DocTags models (specialized document structure output) ──────────\n\n  'granite-docling-258M-mlx': {\n    repo_id: 'ibm-granite/granite-docling-258M-mlx',\n    inference_framework: 'mlx',\n    response_format: 'doctags',\n    transformers_model_type: 'automodel-vision2seq',\n    description:\n      'Granite Docling 258M (MLX, Apple Silicon optimized, ~6s/page)',\n  },\n  'granite-docling-258M': {\n    repo_id: 'ibm-granite/granite-docling-258M',\n    inference_framework: 'transformers',\n    response_format: 'doctags',\n    transformers_model_type: 'automodel-vision2seq',\n    description: 'Granite Docling 258M (Transformers, cross-platform)',\n  },\n  'smoldocling-256M-mlx': {\n    repo_id: 'docling-project/SmolDocling-256M-preview-mlx-bf16',\n    inference_framework: 'mlx',\n    response_format: 'doctags',\n    transformers_model_type: 'automodel-vision2seq',\n    description: 'SmolDocling 256M (MLX, fastest option)',\n  },\n  'smoldocling-256M': {\n    repo_id: 'docling-project/SmolDocling-256M-preview',\n    inference_framework: 'transformers',\n    response_format: 'doctags',\n    transformers_model_type: 'automodel-vision2seq',\n    description: 'SmolDocling 256M (Transformers)',\n  },\n\n  // ── Markdown models (general-purpose vision LLMs) ──────────────────\n\n  'granite-vision-2B': {\n    repo_id: 'ibm-granite/granite-vision-3.2-2b',\n    inference_framework: 'transformers',\n    response_format: 'markdown',\n    transformers_model_type: 'automodel-vision2seq',\n    description: 'Granite Vision 3.2 2B (IBM, higher accuracy)',\n  },\n  'qwen25-vl-3B-mlx': {\n    repo_id: 'mlx-community/Qwen2.5-VL-3B-Instruct-bf16',\n    inference_framework: 'mlx',\n    response_format: 'markdown',\n    transformers_model_type: 'automodel-vision2seq',\n    description: 'Qwen 2.5 VL 3B (MLX, multilingual, good KCJ support)',\n  },\n  phi4: {\n    repo_id: 'microsoft/Phi-4-multimodal-instruct',\n    inference_framework: 'transformers',\n    response_format: 'markdown',\n    transformers_model_type: 'automodel',\n    description: 'Phi-4 Multimodal (Microsoft, CausalLM)',\n  },\n  'pixtral-12B-mlx': {\n    repo_id: 'mlx-community/pixtral-12b-bf16',\n    inference_framework: 'mlx',\n    response_format: 'markdown',\n    transformers_model_type: 'automodel-vision2seq',\n    description: 'Pixtral 12B (MLX, Mistral, high accuracy)',\n  },\n  'pixtral-12B': {\n    repo_id: 'mistral-community/pixtral-12b',\n    inference_framework: 'transformers',\n    response_format: 'markdown',\n    transformers_model_type: 'automodel-vision2seq',\n    description: 'Pixtral 12B (Transformers, Mistral)',\n  },\n  got2: {\n    repo_id: 'stepfun-ai/GOT-OCR-2.0-hf',\n    inference_framework: 'transformers',\n    response_format: 'markdown',\n    transformers_model_type: 'automodel-vision2seq',\n    description: 'GOT-OCR 2.0 (StepFun, OCR-specialized)',\n  },\n  'gemma3-12B-mlx': {\n    repo_id: 'mlx-community/gemma-3-12b-it-bf16',\n    inference_framework: 'mlx',\n    response_format: 'markdown',\n    transformers_model_type: 'automodel-vision2seq',\n    description: 'Gemma 3 12B (MLX, Google)',\n  },\n  'gemma3-27B-mlx': {\n    repo_id: 'mlx-community/gemma-3-27b-it-bf16',\n    inference_framework: 'mlx',\n    response_format: 'markdown',\n    transformers_model_type: 'automodel-vision2seq',\n    description: 'Gemma 3 27B (MLX, Google, highest accuracy)',\n  },\n  dolphin: {\n    repo_id: 'ByteDance/Dolphin',\n    inference_framework: 'transformers',\n    response_format: 'markdown',\n    transformers_model_type: 'automodel-vision2seq',\n    description: 'Dolphin (ByteDance, document-oriented)',\n  },\n} as const;\n\n/**\n * Default VLM model preset key\n */\nexport const DEFAULT_VLM_MODEL = 'granite-docling-258M-mlx';\n\n/**\n * Resolve a VLM model from a preset key or custom VlmModelLocal object.\n *\n * When using a preset key, only required fields are populated.\n * Optional fields (prompt, scale, extra_generation_config) use Docling defaults.\n */\nexport function resolveVlmModel(model: string | VlmModelLocal): VlmModelLocal {\n  if (typeof model === 'string') {\n    const preset = VLM_MODELS[model];\n    if (!preset) {\n      throw new Error(\n        `Unknown VLM model preset: \"${model}\". Available presets: ${Object.keys(VLM_MODELS).join(', ')}`,\n      );\n    }\n    return {\n      repo_id: preset.repo_id,\n      inference_framework: preset.inference_framework,\n      response_format: preset.response_format,\n      transformers_model_type: preset.transformers_model_type,\n    } as VlmModelLocal;\n  }\n  return model;\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAqBO,IAAM,aAA6C;AAAA;AAAA,EAGxD,4BAA4B;AAAA,IAC1B,SAAS;AAAA,IACT,qBAAqB;AAAA,IACrB,iBAAiB;AAAA,IACjB,yBAAyB;AAAA,IACzB,aACE;AAAA,EACJ;AAAA,EACA,wBAAwB;AAAA,IACtB,SAAS;AAAA,IACT,qBAAqB;AAAA,IACrB,iBAAiB;AAAA,IACjB,yBAAyB;AAAA,IACzB,aAAa;AAAA,EACf;AAAA,EACA,wBAAwB;AAAA,IACtB,SAAS;AAAA,IACT,qBAAqB;AAAA,IACrB,iBAAiB;AAAA,IACjB,yBAAyB;AAAA,IACzB,aAAa;AAAA,EACf;AAAA,EACA,oBAAoB;AAAA,IAClB,SAAS;AAAA,IACT,qBAAqB;AAAA,IACrB,iBAAiB;AAAA,IACjB,yBAAyB;AAAA,IACzB,aAAa;AAAA,EACf;AAAA;AAAA,EAIA,qBAAqB;AAAA,IACnB,SAAS;AAAA,IACT,qBAAqB;AAAA,IACrB,iBAAiB;AAAA,IACjB,yBAAyB;AAAA,IACzB,aAAa;AAAA,EACf;AAAA,EACA,oBAAoB;AAAA,IAClB,SAAS;AAAA,IACT,qBAAqB;AAAA,IACrB,iBAAiB;AAAA,IACjB,yBAAyB;AAAA,IACzB,aAAa;AAAA,EACf;AAAA,EACA,MAAM;AAAA,IACJ,SAAS;AAAA,IACT,qBAAqB;AAAA,IACrB,iBAAiB;AAAA,IACjB,yBAAyB;AAAA,IACzB,aAAa;AAAA,EACf;AAAA,EACA,mBAAmB;AAAA,IACjB,SAAS;AAAA,IACT,qBAAqB;AAAA,IACrB,iBAAiB;AAAA,IACjB,yBAAyB;AAAA,IACzB,aAAa;AAAA,EACf;AAAA,EACA,eAAe;AAAA,IACb,SAAS;AAAA,IACT,qBAAqB;AAAA,IACrB,iBAAiB;AAAA,IACjB,yBAAyB;AAAA,IACzB,aAAa;AAAA,EACf;AAAA,EACA,MAAM;AAAA,IACJ,SAAS;AAAA,IACT,qBAAqB;AAAA,IACrB,iBAAiB;AAAA,IACjB,yBAAyB;AAAA,IACzB,aAAa;AAAA,EACf;AAAA,EACA,kBAAkB;AAAA,IAChB,SAAS;AAAA,IACT,qBAAqB;AAAA,IACrB,iBAAiB;AAAA,IACjB,yBAAyB;AAAA,IACzB,aAAa;AAAA,EACf;AAAA,EACA,kBAAkB;AAAA,IAChB,SAAS;AAAA,IACT,qBAAqB;AAAA,IACrB,iBAAiB;AAAA,IACjB,yBAAyB;AAAA,IACzB,aAAa;AAAA,EACf;AAAA,EACA,SAAS;AAAA,IACP,SAAS;AAAA,IACT,qBAAqB;AAAA,IACrB,iBAAiB;AAAA,IACjB,yBAAyB;AAAA,IACzB,aAAa;AAAA,EACf;AACF;AAKO,IAAM,oBAAoB;AAQ1B,SAAS,gBAAgB,OAA8C;AAC5E,MAAI,OAAO,UAAU,UAAU;AAC7B,UAAM,SAAS,WAAW,KAAK;AAC/B,QAAI,CAAC,QAAQ;AACX,YAAM,IAAI;AAAA,QACR,8BAA8B,KAAK,yBAAyB,OAAO,KAAK,UAAU,EAAE,KAAK,IAAI,CAAC;AAAA,MAChG;AAAA,IACF;AACA,WAAO;AAAA,MACL,SAAS,OAAO;AAAA,MAChB,qBAAqB,OAAO;AAAA,MAC5B,iBAAiB,OAAO;AAAA,MACxB,yBAAyB,OAAO;AAAA,IAClC;AAAA,EACF;AACA,SAAO;AACT;","names":[]}

package/dist/vlm-models.d.cts

@@ -0,0 +1,34 @@
+import { VlmModelLocal } from 'docling-sdk';
+
+/**
+ * VLM model preset with description
+ */
+interface VlmModelPreset {
+    repo_id: string;
+    inference_framework: 'mlx' | 'transformers';
+    response_format: 'doctags' | 'markdown';
+    transformers_model_type: 'automodel-vision2seq' | 'automodel';
+    description: string;
+}
+/**
+ * Available VLM model presets
+ *
+ * Based on Docling's official VLM model specs:
+ * https://docling-project.github.io/docling/usage/vision_models/#available-local-models
+ *
+ * Users can select a preset key or provide a custom VlmModelLocal object.
+ */
+declare const VLM_MODELS: Record<string, VlmModelPreset>;
+/**
+ * Default VLM model preset key
+ */
+declare const DEFAULT_VLM_MODEL = "granite-docling-258M-mlx";
+/**
+ * Resolve a VLM model from a preset key or custom VlmModelLocal object.
+ *
+ * When using a preset key, only required fields are populated.
+ * Optional fields (prompt, scale, extra_generation_config) use Docling defaults.
+ */
+declare function resolveVlmModel(model: string | VlmModelLocal): VlmModelLocal;
+
+export { DEFAULT_VLM_MODEL, VLM_MODELS, type VlmModelPreset, resolveVlmModel };

package/dist/vlm-models.d.ts

@@ -0,0 +1,34 @@
+import { VlmModelLocal } from 'docling-sdk';
+
+/**
+ * VLM model preset with description
+ */
+interface VlmModelPreset {
+    repo_id: string;
+    inference_framework: 'mlx' | 'transformers';
+    response_format: 'doctags' | 'markdown';
+    transformers_model_type: 'automodel-vision2seq' | 'automodel';
+    description: string;
+}
+/**
+ * Available VLM model presets
+ *
+ * Based on Docling's official VLM model specs:
+ * https://docling-project.github.io/docling/usage/vision_models/#available-local-models
+ *
+ * Users can select a preset key or provide a custom VlmModelLocal object.
+ */
+declare const VLM_MODELS: Record<string, VlmModelPreset>;
+/**
+ * Default VLM model preset key
+ */
+declare const DEFAULT_VLM_MODEL = "granite-docling-258M-mlx";
+/**
+ * Resolve a VLM model from a preset key or custom VlmModelLocal object.
+ *
+ * When using a preset key, only required fields are populated.
+ * Optional fields (prompt, scale, extra_generation_config) use Docling defaults.
+ */
+declare function resolveVlmModel(model: string | VlmModelLocal): VlmModelLocal;
+
+export { DEFAULT_VLM_MODEL, VLM_MODELS, type VlmModelPreset, resolveVlmModel };

package/dist/vlm-models.js

@@ -0,0 +1,12 @@
+import {
+  DEFAULT_VLM_MODEL,
+  VLM_MODELS,
+  resolveVlmModel
+} from "./chunk-WWNI354M.js";
+import "./chunk-VUNV25KB.js";
+export {
+  DEFAULT_VLM_MODEL,
+  VLM_MODELS,
+  resolveVlmModel
+};
+//# sourceMappingURL=vlm-models.js.map

package/dist/vlm-models.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":[],"sourcesContent":[],"mappings":"","names":[]}

package/package.json

@@ -2,7 +2,7 @@
   "name": "@heripo/pdf-parser",
   "private": false,
   "type": "module",
-  "version": "0.1.4",
+  "version": "0.1.6",
   "description": "PDF parsing library using Docling SDK with OCR support for macOS",
   "main": "dist/index.cjs",
   "module": "dist/index.js",
@@ -13,6 +13,12 @@
       "import": "./dist/index.js",
       "require": "./dist/index.cjs",
       "default": "./dist/index.js"
+    },
+    "./vlm-models": {
+      "types": "./dist/vlm-models.d.ts",
+      "import": "./dist/vlm-models.js",
+      "require": "./dist/vlm-models.cjs",
+      "default": "./dist/vlm-models.js"
     }
   },
   "files": [
@@ -47,7 +53,7 @@
     "macos"
   ],
   "engines": {
-    "node": ">=22"
+    "node": ">=24"
   },
   "os": [
     "darwin"
@@ -59,19 +65,19 @@
     "docling-sdk": "^1.3.6",
     "es-toolkit": "^1.44.0",
     "yauzl": "^3.2.0",
-    "@heripo/model": "0.1.4"
+    "@heripo/model": "0.1.6"
   },
   "devDependencies": {
     "@types/yauzl": "^2.10.3",
-    "@vitest/coverage-v8": "^3.2.4",
-    "@vitest/expect": "^3.2.4",
+    "@vitest/coverage-v8": "^4.0.18",
+    "@vitest/expect": "^4.0.18",
     "tsup": "^8.5.1",
-    "vitest": "^3.2.4",
+    "vitest": "^4.0.18",
     "@heripo/logger": "0.0.0",
-    "@heripo/shared": "0.0.0",
     "@heripo/tsconfig": "0.0.0",
-    "@heripo/vitest-config": "0.0.0",
-    "@heripo/tsup-config": "0.0.0"
+    "@heripo/shared": "0.0.0",
+    "@heripo/tsup-config": "0.0.0",
+    "@heripo/vitest-config": "0.0.0"
   },
   "scripts": {
     "clean": "rimraf dist",