From f85ee6c122e03c47da4a6174d81b1d463652e189 Mon Sep 17 00:00:00 2001
From: Folzo <folzo@hotmail.com>
Date: Sat, 11 Apr 2026 00:42:32 +0200
Subject: [PATCH] feat(kreuzberg): add multi-file processing and PaddleOCR
 backend

- Add extract_files() for processing multiple files one by one
- Support comma-separated and JSON array input
- Switch OCR backend from Tesseract to PaddleOCR (better accuracy)
- Update tool description
---
 effi-kreuzberg/TOOL-OPENWEBUI.md | 85 +++++++++++++++++++++++++++++---
 1 file changed, 78 insertions(+), 7 deletions(-)

diff --git a/effi-kreuzberg/TOOL-OPENWEBUI.md b/effi-kreuzberg/TOOL-OPENWEBUI.md
index aa9d0f3..0c0b017 100644
--- a/effi-kreuzberg/TOOL-OPENWEBUI.md
+++ b/effi-kreuzberg/TOOL-OPENWEBUI.md
@@ -4,7 +4,7 @@
 
 1. **Admin Panel** → **Tools** → **Create Tool**
 2. **Name**: `Kreuzberg OCR`
-3. **Description**: `Extract text from documents (PDF, images, scanned pages) using Kreuzberg OCR. Call `extract_file("filename")` on attached files.`
+3. **Description**: `Extract text from documents using Kreuzberg OCR (PaddleOCR backend). Single: extract_file("file") or Multiple: extract_files("file1,file2")`
 4. **Coller le code ci-dessous**
 5. **Save and Activate**
 
@@ -44,7 +44,7 @@ class Tools:
 
     def extract_file(self, filename: str) -> str:
         """
-        Extract text from a file using Kreuzberg OCR.
+        Extract text from a single file using Kreuzberg OCR.
         
         Args:
             filename: The filename shown in the chat (e.g., "document.pdf", "image.jpg")
@@ -53,6 +53,58 @@ class Tools:
         Returns:
             Extracted text content with markdown formatting
         """
+        return self._extract_single_file(filename)
+
+    def extract_files(self, filenames: str) -> str:
+        """
+        Extract text from multiple files, one at a time.
+        
+        Args:
+            filenames: Comma-separated list of filenames (e.g., "doc1.pdf,doc2.jpg")
+                      or JSON array string (e.g., '["doc1.pdf","doc2.jpg"]')
+        
+        Returns:
+            Combined extracted text from all files, processed sequentially
+        """
+        import json
+        
+        filenames = filenames.strip()
+        
+        if not filenames:
+            return "❌ No filenames provided. Usage: extract_files(\"file1.pdf,file2.jpg\")"
+        
+        if filenames.startswith('['):
+            try:
+                file_list = json.loads(filenames)
+            except json.JSONDecodeError:
+                return f"❌ Invalid JSON array: {filenames}"
+        else:
+            file_list = [f.strip() for f in filenames.split(',')]
+        
+        if not file_list:
+            return "❌ No files to process"
+        
+        if len(file_list) == 1:
+            return self._extract_single_file(file_list[0])
+        
+        results = []
+        total = len(file_list)
+        
+        for i, filename in enumerate(file_list, 1):
+            filename = filename.strip()
+            if not filename:
+                continue
+            
+            if total > 1:
+                results.append(f"--- Document {i}/{total}: `{filename}` ---")
+            
+            result = self._extract_single_file(filename)
+            results.append(result)
+        
+        return "\n\n".join(results)
+
+    def _extract_single_file(self, filename: str) -> str:
+        """Internal method to extract text from a single file"""
         try:
             file_path = self._find_file(filename)
             
@@ -89,6 +141,7 @@ class Tools:
                 files=files,
                 data=data,
                 timeout=300
+            )
             
             if response.status_code == 200:
                 result = response.json()
@@ -96,10 +149,8 @@ class Tools:
                 if 'document' in result:
                     content = result['document'].get('md_content', '')
                     if content and len(content.strip()) > 10:
-                        # Clean up the content
                         content = self._clean_content(content)
                         
-                        # Truncate if too long (LLM context limit)
                         if len(content) > 15000:
                             content = content[:15000] + "\n\n... (document truncated - too long for context)"
                         
@@ -290,7 +341,7 @@ class Tools:
 # OpenWebUI Tool Functions
 def extract_file(filename: str) -> str:
     """
-    Extract text from a file using Kreuzberg OCR.
+    Extract text from a single file using Kreuzberg OCR.
     
     Args:
         filename: The filename or UUID from the chat (e.g., "document.pdf", "abc123_document.pdf")
@@ -301,6 +352,20 @@ def extract_file(filename: str) -> str:
     tools = Tools()
     return tools.extract_file(filename)
 
+def extract_files(filenames: str) -> str:
+    """
+    Extract text from multiple files, one at a time.
+    
+    Args:
+        filenames: Comma-separated list (e.g., "doc1.pdf,doc2.jpg")
+                  or JSON array (e.g., '["doc1.pdf","doc2.jpg"]')
+    
+    Returns:
+        Combined extracted text from all files
+    """
+    tools = Tools()
+    return tools.extract_files(filenames)
+
 def list_recent_files(count: int = 10) -> str:
     """List recent uploaded files"""
     tools = Tools()
@@ -316,13 +381,19 @@ def search_files(pattern: str) -> str:
 
 After installing and activating the tool:
 
-### Extract text from a file:
+### Extract text from a single file:
 ```
 extract_file("document.pdf")
 extract_file("9af3d665-998b-4577-b893-340cb1c28385_document.pdf")
 extract_file("image.jpg")
 ```
 
+### Extract from multiple files (processed one by one):
+```
+extract_files("doc1.pdf,doc2.jpg")
+extract_files('["page1.jpg","page2.jpg","page3.jpg"]')
+```
+
 ### List recent files:
 ```
 list_recent_files()
@@ -337,4 +408,4 @@ search_files("polz")
 
 ---
 
-*Updated 11 April 2026 - Added image preprocessing (grayscale + contrast) for better OCR quality*
\ No newline at end of file
+*Updated 11 April 2026 - Added PaddleOCR backend + multi-file processing*
\ No newline at end of file