@mlx-node/vlm 0.0.6 → 0.0.8

package/README.md

@@ -1,6 +1,6 @@
 # @mlx-node/vlm
 
-Vision-language models and document processing pipelines for Node.js on Apple Silicon. Extract text, tables, and structure from documents and images using PaddleOCR-VL and the PP-StructureV3 pipeline — all running locally on Metal GPU.
+Vision-language models and document processing pipelines for Node.js on Apple Silicon. Extract text, tables, and structure from documents and images using PaddleOCR-VL, Qianfan-OCR (InternVL), and the PP-StructureV3 pipeline — all running locally on Metal GPU.
 
 ## Requirements
 
@@ -13,6 +13,39 @@ Vision-language models and document processing pipelines for Node.js on Apple Si
 npm install @mlx-node/vlm
 ```
 
+## Multi-turn chat with VLMs
+
+`QianfanOCRModel` conforms to the same `ChatSession<M>` surface as the language models in `@mlx-node/lm`, so a single session handle drives both single-shot and multi-turn VLM conversations:
+
+```typescript
+import { ChatSession } from '@mlx-node/lm';
+import { QianfanOCRModel } from '@mlx-node/vlm';
+import { readFileSync } from 'node:fs';
+
+const model = await QianfanOCRModel.load('./models/Qianfan-VL');
+const session = new ChatSession(model, { system: 'You read documents precisely.' });
+
+// First turn with an image.
+const r1 = await session.send('Extract the text from this receipt.', {
+  images: [readFileSync('./receipt.jpg')],
+});
+console.log(r1.text);
+
+// Text-only follow-up reuses the same KV cache against the same image.
+const r2 = await session.send('What is the total price?');
+console.log(r2.text);
+
+// Swapping the image mid-session forcibly restarts the cache with the new image.
+const r3 = await session.send('And this one?', {
+  images: [readFileSync('./other-receipt.jpg')],
+});
+console.log(r3.text);
+```
+
+`loadSession()` from `@mlx-node/lm` cannot load Qianfan-OCR (that would introduce a circular package dependency), so construct the wrapper here with `QianfanOCRModel.load()` and pass it to `ChatSession` directly.
+
+The one-shot `VLModel.chat()` API shown below for PaddleOCR-VL is a single-turn OCR entry point and is intentionally kept out of the session surface.
+
 ## Quick Start
 
 ### Document Structure Analysis
@@ -257,7 +290,6 @@ const buffer = documentToXlsx(doc);
 
 | Export                      | Description                                      |
 | --------------------------- | ------------------------------------------------ |
-| `PADDLEOCR_VL_CONFIGS`      | Pre-defined PaddleOCR-VL 1.5 config              |
 | `createPaddleocrVlConfig()` | Default config factory                           |
 | `OutputFormat`              | Enum: `Raw`, `Plain`, `Markdown`, `Html`, `Json` |
 

package/dist/index.d.ts

@@ -32,10 +32,10 @@ export { TextDetModel, type TextBox, TextRecModel, type RecResult } from '@mlx-n
 export { DocOrientationModel, type OrientationResult, type ClassifyRotateResult, DocUnwarpModel, type UnwarpResult, } from '@mlx-node/core';
 export { StructureV3Pipeline, type StructureV3Config, type AnalyzeOptions, type StructuredElement, type StructuredDocument, type TextLine, } from './pipeline/structure-v3.js';
 export type { VisionConfig, TextConfig, ModelConfig, VlmChatConfig, VlmChatMessage, VlmBatchItem, } from '@mlx-node/core';
-export { PADDLEOCR_VL_CONFIGS, type PaddleOCRVLConfig } from './models/paddleocr-vl-configs.js';
+export { QianfanOCRModel } from './models/qianfan-ocr.js';
+export { createQianfanOcrConfig } from '@mlx-node/core';
+export type { QianfanOcrConfig, InternVisionConfig, Qwen3LmConfig } from '@mlx-node/core';
 export { VlmChatResult, type VLMChatResult } from '@mlx-node/core';
 export { parsePaddleResponse, parseVlmOutput, formatDocument, type ParsedDocument, type DocumentElement, type Table, type TableRow, type TableCell, type Paragraph, type ParserConfig, OutputFormat, } from '@mlx-node/core';
 export { documentToXlsx, saveToXlsx } from '@mlx-node/core';
-export { Qwen3Tokenizer as Tokenizer } from '@mlx-node/lm';
-export { MxArray, type DType } from '@mlx-node/core';
 //# sourceMappingURL=index.d.ts.map

package/dist/index.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../src/index.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;;;;;;;GA2BG;AAKH,OAAO,EAAE,OAAO,EAAE,uBAAuB,EAAE,MAAM,gBAAgB,CAAC;AAGlE,OAAO,EAAE,cAAc,EAAE,KAAK,aAAa,EAAE,MAAM,gBAAgB,CAAC;AAGpE,OAAO,EAAE,YAAY,EAAE,KAAK,OAAO,EAAE,YAAY,EAAE,KAAK,SAAS,EAAE,MAAM,gBAAgB,CAAC;AAG1F,OAAO,EACL,mBAAmB,EACnB,KAAK,iBAAiB,EACtB,KAAK,oBAAoB,EACzB,cAAc,EACd,KAAK,YAAY,GAClB,MAAM,gBAAgB,CAAC;AAGxB,OAAO,EACL,mBAAmB,EACnB,KAAK,iBAAiB,EACtB,KAAK,cAAc,EACnB,KAAK,iBAAiB,EACtB,KAAK,kBAAkB,EACvB,KAAK,QAAQ,GACd,MAAM,4BAA4B,CAAC;AAGpC,YAAY,EACV,YAAY,EACZ,UAAU,EACV,WAAW,EACX,aAAa,EACb,cAAc,EACd,YAAY,GACb,MAAM,gBAAgB,CAAC;AAGxB,OAAO,EAAE,oBAAoB,EAAE,KAAK,iBAAiB,EAAE,MAAM,kCAAkC,CAAC;AAGhG,OAAO,EAAE,aAAa,EAAE,KAAK,aAAa,EAAE,MAAM,gBAAgB,CAAC;AAGnE,OAAO,EACL,mBAAmB,EACnB,cAAc,EACd,cAAc,EACd,KAAK,cAAc,EACnB,KAAK,eAAe,EACpB,KAAK,KAAK,EACV,KAAK,QAAQ,EACb,KAAK,SAAS,EACd,KAAK,SAAS,EACd,KAAK,YAAY,EACjB,YAAY,GACb,MAAM,gBAAgB,CAAC;AAGxB,OAAO,EAAE,cAAc,EAAE,UAAU,EAAE,MAAM,gBAAgB,CAAC;AAG5D,OAAO,EAAE,cAAc,IAAI,SAAS,EAAE,MAAM,cAAc,CAAC;AAC3D,OAAO,EAAE,OAAO,EAAE,KAAK,KAAK,EAAE,MAAM,gBAAgB,CAAC"}
+{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../src/index.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;;;;;;;GA2BG;AAKH,OAAO,EAAE,OAAO,EAAE,uBAAuB,EAAE,MAAM,gBAAgB,CAAC;AAGlE,OAAO,EAAE,cAAc,EAAE,KAAK,aAAa,EAAE,MAAM,gBAAgB,CAAC;AAGpE,OAAO,EAAE,YAAY,EAAE,KAAK,OAAO,EAAE,YAAY,EAAE,KAAK,SAAS,EAAE,MAAM,gBAAgB,CAAC;AAG1F,OAAO,EACL,mBAAmB,EACnB,KAAK,iBAAiB,EACtB,KAAK,oBAAoB,EACzB,cAAc,EACd,KAAK,YAAY,GAClB,MAAM,gBAAgB,CAAC;AAGxB,OAAO,EACL,mBAAmB,EACnB,KAAK,iBAAiB,EACtB,KAAK,cAAc,EACnB,KAAK,iBAAiB,EACtB,KAAK,kBAAkB,EACvB,KAAK,QAAQ,GACd,MAAM,4BAA4B,CAAC;AAGpC,YAAY,EACV,YAAY,EACZ,UAAU,EACV,WAAW,EACX,aAAa,EACb,cAAc,EACd,YAAY,GACb,MAAM,gBAAgB,CAAC;AAGxB,OAAO,EAAE,eAAe,EAAE,MAAM,yBAAyB,CAAC;AAC1D,OAAO,EAAE,sBAAsB,EAAE,MAAM,gBAAgB,CAAC;AACxD,YAAY,EAAE,gBAAgB,EAAE,kBAAkB,EAAE,aAAa,EAAE,MAAM,gBAAgB,CAAC;AAG1F,OAAO,EAAE,aAAa,EAAE,KAAK,aAAa,EAAE,MAAM,gBAAgB,CAAC;AAGnE,OAAO,EACL,mBAAmB,EACnB,cAAc,EACd,cAAc,EACd,KAAK,cAAc,EACnB,KAAK,eAAe,EACpB,KAAK,KAAK,EACV,KAAK,QAAQ,EACb,KAAK,SAAS,EACd,KAAK,SAAS,EACd,KAAK,YAAY,EACjB,YAAY,GACb,MAAM,gBAAgB,CAAC;AAGxB,OAAO,EAAE,cAAc,EAAE,UAAU,EAAE,MAAM,gBAAgB,CAAC"}

package/dist/index.js

@@ -37,14 +37,12 @@ export { TextDetModel, TextRecModel } from '@mlx-node/core';
 export { DocOrientationModel, DocUnwarpModel, } from '@mlx-node/core';
 // Document understanding pipeline (PP-StructureV3)
 export { StructureV3Pipeline, } from './pipeline/structure-v3.js';
-// Model-specific configs
-export { PADDLEOCR_VL_CONFIGS } from './models/paddleocr-vl-configs.js';
+// Qianfan-OCR model (InternVL architecture)
+export { QianfanOCRModel } from './models/qianfan-ocr.js';
+export { createQianfanOcrConfig } from '@mlx-node/core';
 // Chat result type
 export { VlmChatResult } from '@mlx-node/core';
 // Output parsing and formatting (Rust implementation)
 export { parsePaddleResponse, parseVlmOutput, formatDocument, OutputFormat, } from '@mlx-node/core';
 // XLSX export (Rust implementation)
 export { documentToXlsx, saveToXlsx } from '@mlx-node/core';
-// Re-export shared utilities
-export { Qwen3Tokenizer as Tokenizer } from '@mlx-node/lm';
-export { MxArray } from '@mlx-node/core';

package/dist/models/qianfan-ocr.d.ts

@@ -0,0 +1,34 @@
+import { QianfanOCRModel as QianfanOCRModelNative } from '@mlx-node/core';
+declare const QianfanOCRModel_base: {
+    new (config: import("@mlx-node/core").QianfanOcrConfig): import("@mlx-node/lm").StreamingInstance<typeof QianfanOCRModelNative, {
+        readonly recordModelPath: false;
+    }>;
+    load(modelPath: string): Promise<import("@mlx-node/lm").StreamingInstance<typeof QianfanOCRModelNative, {
+        readonly recordModelPath: false;
+    }>>;
+};
+/**
+ * Qianfan-OCR Vision-Language Model wrapper.
+ *
+ * Built from the shared {@link makeStreamingModel} factory in
+ * `@mlx-node/lm` — the empty `extends` inherits the AsyncGenerator
+ * session-streaming overrides (`chatStreamSessionStart` /
+ * `chatStreamSessionContinue` / `chatStreamSessionContinueTool`) so the
+ * wrapper structurally satisfies `SessionCapableModel` and can be
+ * passed to `ChatSession<QianfanOCRModel>`. Importing the factory from
+ * `@mlx-node/lm` is one-directional (vlm → lm), matching the existing
+ * `_runChatStream` dependency, so it introduces no circular dependency.
+ *
+ * Qianfan-OCR is a VLM (InternViT + Qwen3 language model). The continue
+ * path cannot splice new vision features into a live KV cache — image
+ * changes always require a fresh session start, which the high-level
+ * `ChatSession` wrapper handles via its `lastImagesKey` check.
+ *
+ * Unlike the LM families, Qianfan-OCR does NOT record its model path
+ * (`recordModelPath: false`) and therefore exposes no
+ * `applyChatTemplate` — preserving its historical surface.
+ */
+export declare class QianfanOCRModel extends QianfanOCRModel_base {
+}
+export {};
+//# sourceMappingURL=qianfan-ocr.d.ts.map

package/dist/models/qianfan-ocr.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"qianfan-ocr.d.ts","sourceRoot":"","sources":["../../src/models/qianfan-ocr.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,eAAe,IAAI,qBAAqB,EAAE,MAAM,gBAAgB,CAAC;;;;;;;;;AAI1E;;;;;;;;;;;;;;;;;;;;GAoBG;AACH,qBAAa,eAAgB,SAAQ,oBAAqE;CAAG"}

package/dist/models/qianfan-ocr.js

@@ -0,0 +1,38 @@
+import { QianfanOCRModel as QianfanOCRModelNative } from '@mlx-node/core';
+import { makeStreamingModel } from '@mlx-node/lm';
+/**
+ * Qianfan-OCR Vision-Language Model wrapper.
+ *
+ * Built from the shared {@link makeStreamingModel} factory in
+ * `@mlx-node/lm` — the empty `extends` inherits the AsyncGenerator
+ * session-streaming overrides (`chatStreamSessionStart` /
+ * `chatStreamSessionContinue` / `chatStreamSessionContinueTool`) so the
+ * wrapper structurally satisfies `SessionCapableModel` and can be
+ * passed to `ChatSession<QianfanOCRModel>`. Importing the factory from
+ * `@mlx-node/lm` is one-directional (vlm → lm), matching the existing
+ * `_runChatStream` dependency, so it introduces no circular dependency.
+ *
+ * Qianfan-OCR is a VLM (InternViT + Qwen3 language model). The continue
+ * path cannot splice new vision features into a live KV cache — image
+ * changes always require a fresh session start, which the high-level
+ * `ChatSession` wrapper handles via its `lastImagesKey` check.
+ *
+ * Unlike the LM families, Qianfan-OCR does NOT record its model path
+ * (`recordModelPath: false`) and therefore exposes no
+ * `applyChatTemplate` — preserving its historical surface.
+ */
+export class QianfanOCRModel extends makeStreamingModel(QianfanOCRModelNative, { recordModelPath: false }) {
+}
+// -------------------------------------------------------------------
+// Compile-time conformance check
+// -------------------------------------------------------------------
+//
+// Ensures the wrapper structurally satisfies `SessionCapableModel` so
+// `ChatSession<QianfanOCRModel>` will type-check in downstream code.
+// The assignment is compile-only — the `null as unknown as T`
+// placeholder never runs.
+function _assertSessionCapable() {
+    const _qianfan = null;
+    void _qianfan;
+}
+void _assertSessionCapable;

package/dist/pipeline/structure-v3.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"structure-v3.d.ts","sourceRoot":"","sources":["../../src/pipeline/structure-v3.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;;;;;;GA0BG;AAiBH,0DAA0D;AAC1D,MAAM,WAAW,iBAAiB;IAChC,6CAA6C;IAC7C,eAAe,EAAE,MAAM,CAAC;IACxB,sDAAsD;IACtD,gBAAgB,EAAE,MAAM,CAAC;IACzB,wDAAwD;IACxD,gBAAgB,EAAE,MAAM,CAAC;IACzB,kEAAkE;IAClE,QAAQ,EAAE,MAAM,CAAC;IACjB,wEAAwE;IACxE,uBAAuB,CAAC,EAAE,MAAM,CAAC;IACjC,uDAAuD;IACvD,kBAAkB,CAAC,EAAE,MAAM,CAAC;CAC7B;AAED,qCAAqC;AACrC,MAAM,WAAW,cAAc;IAC7B,2DAA2D;IAC3D,eAAe,CAAC,EAAE,MAAM,CAAC;IACzB,yDAAyD;IACzD,gBAAgB,CAAC,EAAE,MAAM,CAAC;IAC1B,0EAA0E;IAC1E,cAAc,CAAC,EAAE,OAAO,CAAC;IACzB,yFAAyF;IACzF,yBAAyB,CAAC,EAAE,OAAO,CAAC;IACpC,wEAAwE;IACxE,eAAe,CAAC,EAAE,OAAO,CAAC;CAC3B;AAED,sDAAsD;AACtD,MAAM,WAAW,QAAQ;IACvB,iEAAiE;IACjE,IAAI,EAAE,MAAM,EAAE,CAAC;IACf,sBAAsB;IACtB,IAAI,EAAE,MAAM,CAAC;IACb,6BAA6B;IAC7B,KAAK,EAAE,MAAM,CAAC;CACf;AAED,6DAA6D;AAC7D,MAAM,WAAW,iBAAiB;IAChC,yCAAyC;IACzC,KAAK,EAAE,MAAM,CAAC;IACd,2BAA2B;IAC3B,KAAK,EAAE,MAAM,CAAC;IACd,kEAAkE;IAClE,IAAI,EAAE,MAAM,EAAE,CAAC;IACf,0BAA0B;IAC1B,KAAK,EAAE,MAAM,CAAC;IACd,8BAA8B;IAC9B,IAAI,EAAE,MAAM,CAAC;IACb,wDAAwD;IACxD,KAAK,CAAC,EAAE,QAAQ,EAAE,CAAC;CACpB;AAED,mCAAmC;AACnC,MAAM,WAAW,kBAAkB;IACjC,2CAA2C;IAC3C,QAAQ,EAAE,iBAAiB,EAAE,CAAC;IAC9B,gCAAgC;IAChC,QAAQ,EAAE,MAAM,CAAC;CAClB;AAqCD;;;;;GAKG;AACH,qBAAa,mBAAmB;IAC9B,OAAO,CAAC,MAAM,CAAiB;IAC/B,OAAO,CAAC,OAAO,CAAe;IAC9B,OAAO,CAAC,OAAO,CAAe;IAC9B,OAAO,CAAC,cAAc,CAA6B;IACnD,OAAO,CAAC,SAAS,CAAwB;IAEzC,OAAO;IAcP;;OAEG;IACH,MAAM,CAAC,IAAI,CAAC,MAAM,EAAE,iBAAiB,GAAG,mBAAmB;IAa3D;;;;;;OAMG;IACG,OAAO,CAAC,SAAS,EAAE,UAAU,GAAG,MAAM,EAAE,OAAO,GAAE,cAAmB,GAAG,OAAO,CAAC,kBAAkB,CAAC;IAgGxG;;;;OAIG;IACG,QAAQ,CAAC,SAAS,EAAE,MAAM,EAAE,gBAAgB,CAAC,EAAE,MAAM,GAAG,OAAO,CAAC,QAAQ,EAAE,CAAC;IAIjF;;;;;OAKG;YACW,SAAS;IAwDvB;;OAEG;YACW,WAAW;CAW1B"}
+{"version":3,"file":"structure-v3.d.ts","sourceRoot":"","sources":["../../src/pipeline/structure-v3.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;;;;;;GA0BG;AAiBH,0DAA0D;AAC1D,MAAM,WAAW,iBAAiB;IAChC,6CAA6C;IAC7C,eAAe,EAAE,MAAM,CAAC;IACxB,sDAAsD;IACtD,gBAAgB,EAAE,MAAM,CAAC;IACzB,wDAAwD;IACxD,gBAAgB,EAAE,MAAM,CAAC;IACzB,kEAAkE;IAClE,QAAQ,EAAE,MAAM,CAAC;IACjB,wEAAwE;IACxE,uBAAuB,CAAC,EAAE,MAAM,CAAC;IACjC,uDAAuD;IACvD,kBAAkB,CAAC,EAAE,MAAM,CAAC;CAC7B;AAED,qCAAqC;AACrC,MAAM,WAAW,cAAc;IAC7B,2DAA2D;IAC3D,eAAe,CAAC,EAAE,MAAM,CAAC;IACzB,yDAAyD;IACzD,gBAAgB,CAAC,EAAE,MAAM,CAAC;IAC1B,0EAA0E;IAC1E,cAAc,CAAC,EAAE,OAAO,CAAC;IACzB,yFAAyF;IACzF,yBAAyB,CAAC,EAAE,OAAO,CAAC;IACpC,wEAAwE;IACxE,eAAe,CAAC,EAAE,OAAO,CAAC;CAC3B;AAED,sDAAsD;AACtD,MAAM,WAAW,QAAQ;IACvB,iEAAiE;IACjE,IAAI,EAAE,MAAM,EAAE,CAAC;IACf,sBAAsB;IACtB,IAAI,EAAE,MAAM,CAAC;IACb,6BAA6B;IAC7B,KAAK,EAAE,MAAM,CAAC;CACf;AAED,6DAA6D;AAC7D,MAAM,WAAW,iBAAiB;IAChC,yCAAyC;IACzC,KAAK,EAAE,MAAM,CAAC;IACd,2BAA2B;IAC3B,KAAK,EAAE,MAAM,CAAC;IACd,kEAAkE;IAClE,IAAI,EAAE,MAAM,EAAE,CAAC;IACf,0BAA0B;IAC1B,KAAK,EAAE,MAAM,CAAC;IACd,8BAA8B;IAC9B,IAAI,EAAE,MAAM,CAAC;IACb,wDAAwD;IACxD,KAAK,CAAC,EAAE,QAAQ,EAAE,CAAC;CACpB;AAED,mCAAmC;AACnC,MAAM,WAAW,kBAAkB;IACjC,2CAA2C;IAC3C,QAAQ,EAAE,iBAAiB,EAAE,CAAC;IAC9B,gCAAgC;IAChC,QAAQ,EAAE,MAAM,CAAC;CAClB;AAqCD;;;;;GAKG;AACH,qBAAa,mBAAmB;IAC9B,OAAO,CAAC,MAAM,CAAiB;IAC/B,OAAO,CAAC,OAAO,CAAe;IAC9B,OAAO,CAAC,OAAO,CAAe;IAC9B,OAAO,CAAC,cAAc,CAA6B;IACnD,OAAO,CAAC,SAAS,CAAwB;IAEzC,OAAO,eAYN;IAED;;OAEG;IACH,MAAM,CAAC,IAAI,CAAC,MAAM,EAAE,iBAAiB,GAAG,mBAAmB,CAW1D;IAED;;;;;;OAMG;IACG,OAAO,CAAC,SAAS,EAAE,UAAU,GAAG,MAAM,EAAE,OAAO,GAAE,cAAmB,GAAG,OAAO,CAAC,kBAAkB,CAAC,CA8FvG;IAED;;;;OAIG;IACG,QAAQ,CAAC,SAAS,EAAE,MAAM,EAAE,gBAAgB,CAAC,EAAE,MAAM,GAAG,OAAO,CAAC,QAAQ,EAAE,CAAC,CAEhF;IAED;;;;;OAKG;YACW,SAAS;IAwDvB;;OAEG;YACW,WAAW;CAW1B"}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@mlx-node/vlm",
-  "version": "0.0.6",
+  "version": "0.0.8",
   "homepage": "https://github.com/mlx-node/mlx-node",
   "bugs": {
     "url": "https://github.com/mlx-node/mlx-node/issues"
@@ -28,11 +28,11 @@
     "test": "vite test run"
   },
   "dependencies": {
-    "@mlx-node/core": "0.0.6",
-    "@mlx-node/lm": "0.0.6",
-    "@napi-rs/image": "^1.12.0"
+    "@mlx-node/core": "0.0.8",
+    "@mlx-node/lm": "0.0.8",
+    "@napi-rs/image": "^1.13.0"
   },
   "devDependencies": {
-    "@types/node": "@types/node@25.5.0"
+    "@types/node": "@types/node@24.12.2"
   }
 }

package/dist/models/index.d.ts

@@ -1,5 +0,0 @@
-/**
- * Model exports
- */
-export { PADDLEOCR_VL_CONFIGS, type PaddleOCRVLConfig } from './paddleocr-vl-configs.js';
-//# sourceMappingURL=index.d.ts.map

package/dist/models/index.d.ts.map

@@ -1 +0,0 @@
-{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../src/models/index.ts"],"names":[],"mappings":"AAAA;;GAEG;AACH,OAAO,EAAE,oBAAoB,EAAE,KAAK,iBAAiB,EAAE,MAAM,2BAA2B,CAAC"}

package/dist/models/index.js

@@ -1,4 +0,0 @@
-/**
- * Model exports
- */
-export { PADDLEOCR_VL_CONFIGS } from './paddleocr-vl-configs.js';

package/dist/models/paddleocr-vl-configs.d.ts

@@ -1,53 +0,0 @@
-/**
- * PaddleOCR-VL Model Configurations
- *
- * Based on the PaddleOCR-VL-1.5 model architecture.
- */
-export interface VisionConfig {
-    modelType: 'paddleocr_vl';
-    hiddenSize: number;
-    intermediateSize: number;
-    numHiddenLayers: number;
-    numAttentionHeads: number;
-    numChannels: number;
-    imageSize: number;
-    patchSize: number;
-    hiddenAct: string;
-    layerNormEps: number;
-    attentionDropout: number;
-    spatialMergeSize: number;
-}
-export interface TextConfig {
-    modelType: 'paddleocr_vl';
-    hiddenSize: number;
-    numHiddenLayers: number;
-    intermediateSize: number;
-    numAttentionHeads: number;
-    rmsNormEps: number;
-    vocabSize: number;
-    numKeyValueHeads: number;
-    maxPositionEmbeddings: number;
-    ropeTheta: number;
-    ropeTraditional: boolean;
-    useBias: boolean;
-    headDim: number;
-    mropeSection: [number, number, number];
-}
-export interface PaddleOCRVLConfig {
-    visionConfig: VisionConfig;
-    textConfig: TextConfig;
-    modelType: 'paddleocr_vl';
-    ignoreIndex: number;
-    imageTokenId: number;
-    videoTokenId: number;
-    visionStartTokenId: number;
-    visionEndTokenId: number;
-    eosTokenId: number;
-}
-/**
- * Available PaddleOCR-VL configurations
- */
-export declare const PADDLEOCR_VL_CONFIGS: {
-    readonly 'paddleocr-vl-1.5': PaddleOCRVLConfig;
-};
-//# sourceMappingURL=paddleocr-vl-configs.d.ts.map

package/dist/models/paddleocr-vl-configs.d.ts.map

@@ -1 +0,0 @@
-{"version":3,"file":"paddleocr-vl-configs.d.ts","sourceRoot":"","sources":["../../src/models/paddleocr-vl-configs.ts"],"names":[],"mappings":"AAAA;;;;GAIG;AAEH,MAAM,WAAW,YAAY;IAC3B,SAAS,EAAE,cAAc,CAAC;IAC1B,UAAU,EAAE,MAAM,CAAC;IACnB,gBAAgB,EAAE,MAAM,CAAC;IACzB,eAAe,EAAE,MAAM,CAAC;IACxB,iBAAiB,EAAE,MAAM,CAAC;IAC1B,WAAW,EAAE,MAAM,CAAC;IACpB,SAAS,EAAE,MAAM,CAAC;IAClB,SAAS,EAAE,MAAM,CAAC;IAClB,SAAS,EAAE,MAAM,CAAC;IAClB,YAAY,EAAE,MAAM,CAAC;IACrB,gBAAgB,EAAE,MAAM,CAAC;IACzB,gBAAgB,EAAE,MAAM,CAAC;CAC1B;AAED,MAAM,WAAW,UAAU;IACzB,SAAS,EAAE,cAAc,CAAC;IAC1B,UAAU,EAAE,MAAM,CAAC;IACnB,eAAe,EAAE,MAAM,CAAC;IACxB,gBAAgB,EAAE,MAAM,CAAC;IACzB,iBAAiB,EAAE,MAAM,CAAC;IAC1B,UAAU,EAAE,MAAM,CAAC;IACnB,SAAS,EAAE,MAAM,CAAC;IAClB,gBAAgB,EAAE,MAAM,CAAC;IACzB,qBAAqB,EAAE,MAAM,CAAC;IAC9B,SAAS,EAAE,MAAM,CAAC;IAClB,eAAe,EAAE,OAAO,CAAC;IACzB,OAAO,EAAE,OAAO,CAAC;IACjB,OAAO,EAAE,MAAM,CAAC;IAChB,YAAY,EAAE,CAAC,MAAM,EAAE,MAAM,EAAE,MAAM,CAAC,CAAC;CACxC;AAED,MAAM,WAAW,iBAAiB;IAChC,YAAY,EAAE,YAAY,CAAC;IAC3B,UAAU,EAAE,UAAU,CAAC;IACvB,SAAS,EAAE,cAAc,CAAC;IAC1B,WAAW,EAAE,MAAM,CAAC;IACpB,YAAY,EAAE,MAAM,CAAC;IACrB,YAAY,EAAE,MAAM,CAAC;IACrB,kBAAkB,EAAE,MAAM,CAAC;IAC3B,gBAAgB,EAAE,MAAM,CAAC;IACzB,UAAU,EAAE,MAAM,CAAC;CACpB;AA6CD;;GAEG;AACH,eAAO,MAAM,oBAAoB;;CAEvB,CAAC"}

package/dist/models/paddleocr-vl-configs.js

@@ -1,53 +0,0 @@
-/**
- * PaddleOCR-VL Model Configurations
- *
- * Based on the PaddleOCR-VL-1.5 model architecture.
- */
-/**
- * PaddleOCR-VL-1.5 default configuration
- */
-const PADDLEOCR_VL_1_5_CONFIG = {
-    modelType: 'paddleocr_vl',
-    ignoreIndex: -100,
-    imageTokenId: 100295,
-    videoTokenId: 100296,
-    visionStartTokenId: 101305,
-    visionEndTokenId: 101306,
-    eosTokenId: 2,
-    visionConfig: {
-        modelType: 'paddleocr_vl',
-        hiddenSize: 1152,
-        intermediateSize: 4304,
-        numHiddenLayers: 27,
-        numAttentionHeads: 16,
-        numChannels: 3,
-        imageSize: 384,
-        patchSize: 14,
-        hiddenAct: 'gelu_pytorch_tanh',
-        layerNormEps: 1e-6,
-        attentionDropout: 0.0,
-        spatialMergeSize: 2,
-    },
-    textConfig: {
-        modelType: 'paddleocr_vl',
-        hiddenSize: 1024,
-        numHiddenLayers: 18,
-        intermediateSize: 3072,
-        numAttentionHeads: 16,
-        rmsNormEps: 1e-5,
-        vocabSize: 103424,
-        numKeyValueHeads: 2,
-        maxPositionEmbeddings: 131072,
-        ropeTheta: 500000.0,
-        ropeTraditional: false,
-        useBias: false,
-        headDim: 128,
-        mropeSection: [16, 24, 24],
-    },
-};
-/**
- * Available PaddleOCR-VL configurations
- */
-export const PADDLEOCR_VL_CONFIGS = {
-    'paddleocr-vl-1.5': PADDLEOCR_VL_1_5_CONFIG,
-};