whatsapp-pi 1.0.51 → 1.0.52

package/README.md

@@ -27,7 +27,7 @@ Pi is a powerful agentic AI coding assistant that operates in your terminal. Thi
 - **Media Support**: 
   - **Vision Analysis**: Automatically forwards WhatsApp images to Pi for analysis.
   - **Audio Transcription**: Transcribes voice notes when Whisper is installed.
-  - **Document Handling**: Downloads and stores documents (PDF, text) for agent access.
+  - **Document Handling**: Downloads and stores documents (PDF, text) for agent access; PDFs include a bounded text preview when readable.
 
 ## Prerequisites
 
@@ -36,10 +36,8 @@ To enable audio transcription features:
 python -m pip install -U openai-whisper
 ```
 
-To enable PDF reading capabilities (required for the agent to process documents):
-- **Linux**: `sudo apt-get install poppler-utils`
-- **macOS**: `brew install poppler`
-- **Windows**: Install `poppler` (e.g., via Scoop) and add to PATH.
+PDF documents are parsed locally and do not require extra system utilities.
+If a PDF cannot be parsed automatically, it is still saved and forwarded with a clear fallback notice.
 
 ## Quick Start
 
@@ -164,7 +162,7 @@ npm test
 - **Group-Only Mode**: Use `--whatsapp-group <jid>` to bind Pi to a single WhatsApp group. The group must also be present in Allowed Groups.
 - **Recents Store**: Recent conversations and message history are persisted in `~/.pi/whatsapp-pi/recents/recents.json`.
 - **Message Detail / Reply**: Open a message from history to inspect full content and reply with `R`.
-- **Media Support**: Images are forwarded for vision analysis, audio is transcribed with Whisper, and documents are saved under `./.pi-data/whatsapp/documents/`.
+- **Media Support**: Images are forwarded for vision analysis, audio is transcribed with Whisper, and PDFs are saved under `./.pi-data/whatsapp/documents/` with local text preview when available.
 - **Session Handling**: Saved state, allow list, and startup reconnects are restored automatically when available.
 - **Intelligent Message Filtering**: Messages ending with `π` are ignored to prevent bot loops.
 - **Storage Management**: Persistent data lives under `.pi-data/` plus the recents store in the user home directory.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "whatsapp-pi",
-  "version": "1.0.51",
+  "version": "1.0.52",
   "type": "module",
   "description": "WhatsApp integration extension for Pi",
   "main": "whatsapp-pi.ts",
@@ -32,6 +32,7 @@
     "typecheck": "tsc --noEmit"
   },
   "dependencies": {
+    "@llamaindex/liteparse": "^1.5.3",
     "baileys": "^6.7.21",
     "pino": "^10.3.1",
     "qrcode-terminal": "^0.12.0"

package/src/i18n.ts

@@ -197,6 +197,8 @@ const fallback = {
     "incoming.media.documentSize": "Size: {size}",
     "incoming.media.documentLocation": "Location: {relativePath}",
     "incoming.media.documentDescription": "Description: {caption}",
+    "incoming.media.documentPdfPreviewHeading": "PDF text preview:",
+    "incoming.media.documentPdfFallbackNotice": "PDF text was not extracted automatically. The file is saved at the path above.",
     "incoming.media.documentDownloadFailed": "[WhatsApp-Pi] Failed to download document:",
     "incoming.media.documentDownloadFailedText": "[Document: {fileName} (download failed)]",
     "audio.emptyTranscription": "[Empty transcription]",
@@ -375,6 +377,8 @@ const translations: Record<Locale, Partial<Record<Key, string>>> = {
         "incoming.media.documentSize": "Tamanho: {size}",
         "incoming.media.documentLocation": "Local: {relativePath}",
         "incoming.media.documentDescription": "Descrição: {caption}",
+        "incoming.media.documentPdfPreviewHeading": "Prévia do texto do PDF:",
+        "incoming.media.documentPdfFallbackNotice": "O texto do PDF não pôde ser extraído automaticamente. O arquivo foi salvo no caminho acima.",
         "incoming.media.documentDownloadFailed": "[WhatsApp-Pi] Falha ao baixar documento:",
         "incoming.media.documentDownloadFailedText": "[Documento: {fileName} (download falhou)]",
     },

package/src/services/incoming-media.service.ts

@@ -1,6 +1,7 @@
 import { downloadContentFromMessage } from 'baileys';
 import { mkdir, writeFile } from 'node:fs/promises';
 import { join } from 'node:path';
+import { LiteParse } from '@llamaindex/liteparse';
 import { AudioService } from './audio.service.js';
 import type { IncomingResolution } from './incoming-message.resolver.js';
 import { WhatsAppPiLogger } from './whatsapp-pi.logger.js';
@@ -12,7 +13,11 @@ export interface ProcessedIncomingContent {
     imageMimeType?: string;
 }
 
+const PDF_PREVIEW_LIMIT = 1200;
+
 export class IncomingMediaService {
+    private readonly pdfParser = new LiteParse({ ocrEnabled: true });
+
     constructor(
         private readonly audioService: AudioService,
         private readonly logger = new WhatsAppPiLogger(false)
@@ -80,6 +85,15 @@ export class IncomingMediaService {
                 + t('incoming.media.documentSize', { size: this.formatFileSize(fileSize) }) + '\n'
                 + t('incoming.media.documentLocation', { relativePath });
 
+            if (this.isPdfDocument(fileName, mimeType)) {
+                const preview = await this.extractPdfPreview(buffer);
+                if (preview) {
+                    text += `\n\n${t('incoming.media.documentPdfPreviewHeading')}\n${preview}`;
+                } else {
+                    text += `\n\n${t('incoming.media.documentPdfFallbackNotice')}`;
+                }
+            }
+
             if (documentMessage.caption) {
                 text += `\n\n${t('incoming.media.documentDescription', { caption: documentMessage.caption })}`;
             }
@@ -91,6 +105,34 @@ export class IncomingMediaService {
         }
     }
 
+    private async extractPdfPreview(buffer: Buffer): Promise<string | null> {
+        try {
+            const result = await this.pdfParser.parse(buffer);
+            return this.formatPdfPreview(result.text);
+        } catch (error) {
+            this.logger.warn('[WhatsApp-Pi] PDF parsing failed, falling back to storage-only behavior.', error);
+            return null;
+        }
+    }
+
+    private formatPdfPreview(text: string | undefined | null): string | null {
+        const normalized = (text || '').replace(/\r\n/g, '\n').trim();
+        if (!normalized) {
+            return null;
+        }
+
+        if (normalized.length <= PDF_PREVIEW_LIMIT) {
+            return normalized;
+        }
+
+        return `${normalized.slice(0, PDF_PREVIEW_LIMIT)}…`;
+    }
+
+    private isPdfDocument(fileName: string, mimeType: string): boolean {
+        const normalizedMimeType = mimeType.toLowerCase().split(';')[0].trim();
+        return normalizedMimeType === 'application/pdf' || fileName.toLowerCase().endsWith('.pdf');
+    }
+
     private async downloadMessage(message: any, type: 'image' | 'document'): Promise<Buffer> {
         const stream = await downloadContentFromMessage(message, type);
         let buffer = Buffer.from([]);

package/whatsapp-pi.ts

@@ -183,17 +183,6 @@ export default function (pi: ExtensionAPI) {
         }
 
         ctx.ui.notify('WhatsApp: Session reset via /new is now fully supported.', 'info');
-
-        // Verify pdftotext availability for document support
-        try {
-            const { code } = await pi.exec('pdftotext', ['-v']);
-            if (code !== 0 && code !== 99) { // 99 is a common exit code for -v in some versions
-                throw new Error(`pdftotext returned code ${code}`);
-            }
-        } catch {
-            ctx.ui.notify('WhatsApp: pdftotext not found. PDF document support will be limited to storage only.', 'warning');
-            logger.warn('[WhatsApp-Pi] Warning: pdftotext not found in system PATH.');
-        }
     });
 
     // Track whether send_wa_message tool already sent a reply this turn