@pandi2352/gemini-ocr 2.0.0 → 4.0.0

package/README.md

@@ -112,6 +112,23 @@ console.log('Extracted Data:', result.entityResult);
 */
 ```
 
+```
+
+### 6. Realtime Progress Feedback
+Get granular updates on the processing stages.
+
+```typescript
+await processOCR({
+  input: ['./large_document.pdf'],
+  apiKey: process.env.GEMINI_API_KEY,
+  
+  onProgress: (stage, message) => {
+    // stage: 'upload' | 'generate_text' | 'enrich' | 'complete'
+    console.log(`[${stage}]: ${message}`);
+  }
+});
+```
+
 ---
 
 ## 🛠️ Configuration Options
@@ -120,11 +137,12 @@ console.log('Extracted Data:', result.entityResult);
 | :--- | :--- | :--- | :--- |
 | `input` | `Array<string \| Buffer \| Object>` | **Required** | Array of file paths, URLs, Buffers, or Base64 strings. |
 | `apiKey` | `string` | **Required** | Your Google Gemini API Key. |
-| `model` | `string` | `gemini-1.5-flash` | The AI model to use. |
+| `model` | `string` | `gemini-1.5-flash` | The AI model (use `gemini-1.5-flash-8b` for speed). |
 | `summarize` | `boolean` | `false` | Generate `metadata` (title, desc, thumbnail). |
 | `mindmap` | `boolean` | `false` | Generate Mermaid.js syntax for visual mapping. |
 | `extractEntities`| `boolean` | `false` | Enable structured field extraction. |
 | `entitySchema` | `string[]` | `auto` | Custom fields to extract (optional). |
+| `onProgress` | `(stage, userMsg) => void` | `undefined` | Callback for realtime progress updates. |
 
 ---
 

package/dist/index.js

@@ -52,15 +52,19 @@ async function processSingleFile(input, options) {
     const logger = new utils_1.Logger();
     const requestId = (0, utils_1.generateRequestId)();
     logger.log(`INIT: Processing file. RequestId: ${requestId}`);
+    if (options.onProgress)
+        options.onProgress('init', 'Initializing processing...');
     try {
         if (!options.apiKey)
             throw new Error('Gemini API key is required.');
         const gemini = new llm_1.GeminiClient(options.apiKey, logger);
-        const modelName = options.model || 'gemini-1.5-flash';
+        const modelName = options.model || 'gemini-2.5-flash';
         // Input Processing
         const inputHandler = new input_handler_1.InputHandler(logger);
         const normalized = await inputHandler.processInput(input);
         const mimeType = normalized.mimeType;
+        if (options.onProgress)
+            options.onProgress('upload', 'Processing input file...');
         // Strategy
         let strategy = 'MEDIA';
         if (mimeType === 'text/plain' || mimeType === 'text/csv' || normalized.extension === 'txt' || normalized.extension === 'csv') {
@@ -85,6 +89,8 @@ async function processSingleFile(input, options) {
         let analysisText = '';
         let extractedTextDocx = '';
         let fileUri;
+        if (options.onProgress)
+            options.onProgress('generate_text', 'Generating analysis...');
         if (strategy === 'TEXT') {
             const content = normalized.data.toString('utf-8');
             finalPrompt += `\n\nDOCUMENT CONTENT:\n${content}`;
@@ -136,48 +142,58 @@ async function processSingleFile(input, options) {
                 catch (e) { }
             }
         }
-        // Mindmap
+        // Parallel Processing for Advanced Features
+        // Optimization Note: Future version can combine these into a single "Mega-Prompt" to reduce HTTP round-trips.
+        if (options.onProgress && (options.mindmap || options.extractEntities)) {
+            options.onProgress('enrich', 'Generating mindmap/entities...');
+        }
+        const tasks = [];
         let mindmap = null;
+        let entityResult = null;
         if (options.mindmap) {
             const enrichPrompt = `${prompts_1.ENRICHMENT_PROMPT}\n\nCONTEXT:\n${mainAnalysis}`;
-            try {
-                const enrichRes = await gemini.generateContent(modelName, enrichPrompt);
-                const jsonPart = enrichRes.match(/\{[\s\S]*\}/);
-                const mermaidPart = enrichRes.match(/```mermaid\n([\s\S]*?)\n```/);
-                if (jsonPart) {
-                    const parsed = JSON.parse(jsonPart[0]);
-                    mindmap = parsed.mermaid || null;
-                }
-                else if (mermaidPart) {
-                    mindmap = mermaidPart[1];
+            tasks.push((async () => {
+                try {
+                    const enrichRes = await gemini.generateContent(modelName, enrichPrompt);
+                    const jsonPart = enrichRes.match(/\{[\s\S]*\}/);
+                    const mermaidPart = enrichRes.match(/```mermaid\n([\s\S]*?)\n```/);
+                    if (jsonPart) {
+                        const parsed = JSON.parse(jsonPart[0]);
+                        mindmap = parsed.mermaid || null;
+                    }
+                    else if (mermaidPart) {
+                        mindmap = mermaidPart[1];
+                    }
+                    else {
+                        mindmap = enrichRes.replace(/```/g, '');
+                    }
                 }
-                else {
-                    mindmap = enrichRes.replace(/```/g, '');
+                catch (e) {
+                    logger.log(`Mindmap failed: ${e}`);
                 }
-            }
-            catch (e) {
-                logger.log(`Mindmap failed: ${e}`);
-            }
+            })());
         }
-        // Entities
-        let entityResult = null;
         if (options.extractEntities) {
             let entityPromptStr = options.entitySchema
                 ? (0, prompts_1.generateEntityPrompt)(options.entitySchema)
                 : prompts_1.AUTO_ENTITY_EXTRACTION_PROMPT;
-            try {
-                const context = strategy === 'DOCX' ? extractedTextDocx : mainAnalysis;
-                const finalEntityPrompt = `${entityPromptStr}\n\nDATA CONTEXT:\n${context}`;
-                const res = await gemini.generateContent(modelName, finalEntityPrompt);
-                const json = res.match(/\{[\s\S]*\}/);
-                if (json) {
-                    entityResult = JSON.parse(json[0]);
+            tasks.push((async () => {
+                try {
+                    const context = strategy === 'DOCX' ? extractedTextDocx : mainAnalysis;
+                    const finalEntityPrompt = `${entityPromptStr}\n\nDATA CONTEXT:\n${context}`;
+                    const res = await gemini.generateContent(modelName, finalEntityPrompt);
+                    const json = res.match(/\{[\s\S]*\}/);
+                    if (json) {
+                        entityResult = JSON.parse(json[0]);
+                    }
                 }
-            }
-            catch (e) {
-                logger.log(`Entity extraction failed: ${e}`);
-            }
+                catch (e) {
+                    logger.log(`Entity extraction failed: ${e}`);
+                }
+            })());
         }
+        // Wait for all parallel tasks to complete
+        await Promise.all(tasks);
         // Page Count
         let pageCount = 1;
         if (mimeType === 'application/pdf') {
@@ -188,6 +204,8 @@ async function processSingleFile(input, options) {
             catch (e) { }
         }
         const endTime = new Date();
+        if (options.onProgress)
+            options.onProgress('complete', 'Processing complete.');
         // Success Result
         return {
             status: 'success',

package/dist/types.d.ts

@@ -14,6 +14,7 @@ export interface OCROptions {
     pageLimit?: number;
     entitySchema?: string[] | any;
     classify?: boolean;
+    onProgress?: (stage: string, message: string) => void;
 }
 export interface OCRTimings {
     startTime: string;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@pandi2352/gemini-ocr",
-  "version": "2.0.0",
+  "version": "4.0.0",
   "description": "A lightweight OCR processing wrapper using Google Gemini Vision models.",
   "publishConfig": {
     "access": "public"