paperless_to_openwebui.py aktualisiert

paperless_to_openwebui.py

@@ -880,6 +880,75 @@ def wait_for_openwebui():
 # MAIN
 # ============================================================
 def main():
+    def ocr_fix_only_pass():
+        """
+        Geht alle Paperless-Dokumente durch und korrigiert nur den OCR-Text.
+        Kein Upload nach OpenWebUI – das erledigt der nächste reguläre Sync.
+        """
+        log.info("=== OCR FIX ONLY MODE ===")
+        log.info("Fetching all documents from Paperless...")
+    
+        docs = []
+        page = 1
+        while True:
+            url = f"{PAPERLESS_URL}/api/documents/?page_size=100&page={page}"
+            r = requests.get(url, headers=paperless_headers, timeout=120)
+            r.raise_for_status()
+            data = r.json()
+            results = data.get("results", [])
+            if not results:
+                break
+            docs.extend(results)
+            if not data.get("next"):
+                break
+            page += 1
+    
+        total = len(docs)
+        log.info("Found %s documents to check.", total)
+    
+        fixed = 0
+        skipped = 0
+        errors = 0
+    
+        for i, doc in enumerate(docs):
+            doc_id = doc["id"]
+            title = doc.get("title", f"doc_{doc_id}")
+            content = doc.get("content", "")
+    
+            if not content or len(content.strip()) < 50:
+                skipped += 1
+                continue
+    
+            if not needs_ocr_fix(content):
+                skipped += 1
+                if (i + 1) % 100 == 0:
+                    log.info("  Progress: %d/%d (fixed=%d, skipped=%d)",
+                             i + 1, total, fixed, skipped)
+                continue
+    
+            log.info("OCR fix needed: #%s '%s'", doc_id, title)
+            try:
+                fixed_text = fix_ocr_chunked(content)
+                if fixed_text and fixed_text != content:
+                    if update_paperless_content(doc_id, fixed_text):
+                        fixed += 1
+                        log.info("  ✓ Fixed #%s (delta: %+d chars)",
+                                 doc_id, len(fixed_text) - len(content))
+                    else:
+                        errors += 1
+                else:
+                    skipped += 1
+                    log.debug("  No changes for #%s", doc_id)
+            except Exception as e:
+                log.error("  ✗ Error fixing #%s: %s", doc_id, e)
+                errors += 1
+    
+        log.info("=== OCR FIX COMPLETE ===")
+        log.info("  Total:   %d", total)
+        log.info("  Fixed:   %d", fixed)
+        log.info("  Skipped: %d", skipped)
+        log.info("  Errors:  %d", errors)
+        
     global KNOWLEDGE_ID
     init_db()
     wait_for_openwebui()
@@ -896,6 +965,15 @@ def calc_wait_time(target_time_str):
     return (target - n).total_seconds()
 
 if __name__ == "__main__":
+    OCR_FIX_ONLY = os.environ.get("OCR_FIX_ONLY", "false").lower() == "true"
+
+    if OCR_FIX_ONLY:
+        # Nur OCR fixen, kein Sync nach OpenWebUI
+        init_db()
+        ocr_fix_only_pass()
+        sys.exit(0)
+
+    # Normaler Betrieb
     RUN_AT = os.getenv("SYNC_TIME", "04:00")
     print(f"Container started, task will run daily at {RUN_AT}", flush=True)
     while True: