feat(vision): wire pipeline into multimodal indexer and add IVisionProvider implementations

- MultimodalIndexer constructor now accepts optional visionPipeline param
  that auto-wraps as PipelineVisionProvider for seamless integration
- LLMVisionProvider: wraps generateText() for simple cloud-only vision
- PipelineVisionProvider: wraps full VisionPipeline as IVisionProvider
- Export all vision types and classes from main barrel (src/index.ts)
- Backward-compatible: existing visionProvider param still works

Author: jddunn

src/index.ts

@@ -87,6 +87,22 @@ export * from './voice';
 export * from './speech';
 // Unified image generation providers
 export * from './core/images';
+// Unified vision pipeline (OCR + handwriting + document AI + CLIP + cloud)
+export { VisionPipeline, createVisionPipeline, LLMVisionProvider, PipelineVisionProvider } from './core/vision/index.js';
+export type {
+  VisionPipelineConfig,
+  VisionResult,
+  VisionStrategy,
+  VisionTier,
+  ContentCategory as VisionContentCategory,
+  TierResult as VisionTierResult,
+  TextRegion as VisionTextRegion,
+  DocumentLayout,
+  DocumentPage,
+  LayoutBlock,
+  VisionPreprocessingConfig,
+} from './core/vision/types.js';
+export type { LLMVisionProviderConfig } from './core/vision/providers/LLMVisionProvider.js';
 // Skills (SKILL.md prompt modules)
 export * from './skills';
 // Multilingual exports

src/rag/multimodal/MultimodalIndexer.ts

@@ -51,6 +51,7 @@ import type {
   ISpeechToTextProvider,
   MultimodalIndexerConfig,
 } from './types.js';
+import type { VisionPipeline } from '../../core/vision/VisionPipeline.js';
 
 // ---------------------------------------------------------------------------
 // Constants
@@ -157,26 +158,39 @@ export class MultimodalIndexer {
    * @param deps.embeddingManager - Manager for generating text embeddings.
    * @param deps.vectorStore - Vector store for document storage and search.
    * @param deps.visionProvider - Optional vision LLM for image description.
+   * @param deps.visionPipeline - Optional full vision pipeline with OCR, handwriting,
+   *   document understanding, CLIP embeddings, and cloud fallback. When provided,
+   *   it is wrapped as an {@link IVisionProvider} via {@link PipelineVisionProvider},
+   *   overriding any `visionProvider` passed alongside it.
    * @param deps.sttProvider - Optional STT provider for audio transcription.
    * @param deps.config - Optional configuration overrides.
    *
    * @throws {Error} If embeddingManager or vectorStore is missing.
    *
    * @example
    * ```typescript
+   * // With a simple vision LLM provider
    * const indexer = new MultimodalIndexer({
    *   embeddingManager,
    *   vectorStore,
    *   visionProvider: myVisionLLM,
    *   sttProvider: myWhisperService,
    *   config: { defaultCollection: 'knowledge' },
    * });
+   *
+   * // With the full vision pipeline (recommended)
+   * const indexer = new MultimodalIndexer({
+   *   embeddingManager,
+   *   vectorStore,
+   *   visionPipeline: myVisionPipeline,
+   * });
    * ```
    */
   constructor(deps: {
     embeddingManager: IEmbeddingManager;
     vectorStore: IVectorStore;
     visionProvider?: IVisionProvider;
+    visionPipeline?: VisionPipeline;
     sttProvider?: ISpeechToTextProvider;
     config?: MultimodalIndexerConfig;
   }) {
@@ -189,9 +203,21 @@ export class MultimodalIndexer {
 
     this._embeddingManager = deps.embeddingManager;
     this._vectorStore = deps.vectorStore;
-    this._visionProvider = deps.visionProvider;
     this._sttProvider = deps.sttProvider;
 
+    // If a full VisionPipeline is provided, wrap it as an IVisionProvider.
+    // This gives the indexer access to the progressive OCR + vision pipeline
+    // for image description, while maintaining backward compatibility with
+    // the simpler IVisionProvider interface.
+    if (deps.visionPipeline) {
+      // Lazy import to avoid circular dependency at module load time.
+      // PipelineVisionProvider is a thin adapter — safe to require synchronously.
+      const { PipelineVisionProvider } = require('../../core/vision/providers/PipelineVisionProvider.js');
+      this._visionProvider = new PipelineVisionProvider(deps.visionPipeline);
+    } else {
+      this._visionProvider = deps.visionProvider;
+    }
+
     this._config = {
       defaultCollection: deps.config?.defaultCollection ?? DEFAULT_COLLECTION,
       imageDescriptionPrompt: deps.config?.imageDescriptionPrompt ?? DEFAULT_IMAGE_DESCRIPTION_PROMPT,