Merge branch 'feature/20251219_lxc' into develop

api/app/core/rag/app/naive.py

@@ -672,6 +672,7 @@ def chunk(filename, binary=None, from_page=0, to_page=100000,
         excel_parser = ExcelParser()
         if parser_config.get("html4excel"):
             sections = [(_, "") for _ in excel_parser.html(binary, 12) if _]
+            parser_config["chunk_token_num"] = 0
         else:
             sections = [(_, "") for _ in excel_parser(binary) if _]
         parser_config["chunk_token_num"] = 12800

api/app/core/rag/deepdoc/parser/excel_parser.py

@@ -5,6 +5,7 @@ from io import BytesIO
 
 import pandas as pd
 from openpyxl import Workbook, load_workbook
+from PIL import Image
 
 from app.core.rag.nlp import find_codec
 
@@ -28,7 +29,7 @@ class RAGExcelParser:
 
             try:
                 file_like_object.seek(0)
-                df = pd.read_csv(file_like_object)
+                df = pd.read_csv(file_like_object, on_bad_lines='skip')
                 return RAGExcelParser._dataframe_to_workbook(df)
 
             except Exception as e_csv:
@@ -42,14 +43,12 @@ class RAGExcelParser:
                 file_like_object.seek(0)
                 try:
                     dfs = pd.read_excel(file_like_object, sheet_name=None)
-                    if isinstance(dfs, dict):
-                        dfs = next(iter(dfs.values()))
                     return RAGExcelParser._dataframe_to_workbook(dfs)
                 except Exception as ex:
                     logging.info(f"pandas with default engine load error: {ex}, try calamine instead")
                     file_like_object.seek(0)
                     df = pd.read_excel(file_like_object, engine="calamine")
-                    print(df)
+                    print("lxc1")
                     return RAGExcelParser._dataframe_to_workbook(df)
             except Exception as e_pandas:
                 raise Exception(f"pandas.read_excel error: {e_pandas}, original openpyxl error: {e}")
@@ -68,7 +67,6 @@ class RAGExcelParser:
         # if contains multiple sheets use _dataframes_to_workbook
         if isinstance(df, dict) and len(df) > 1:
             return RAGExcelParser._dataframes_to_workbook(df)
-
         df = RAGExcelParser._clean_dataframe(df)
         wb = Workbook()
         ws = wb.active
@@ -80,15 +78,14 @@ class RAGExcelParser:
         for row_num, row in enumerate(df.values, 2):
             for col_num, value in enumerate(row, 1):
                 ws.cell(row=row_num, column=col_num, value=value)
-
         return wb
-    
+
     @staticmethod
     def _dataframes_to_workbook(dfs: dict):
         wb = Workbook()
         default_sheet = wb.active
         wb.remove(default_sheet)
-        
+
         for sheet_name, df in dfs.items():
             df = RAGExcelParser._clean_dataframe(df)
             ws = wb.create_sheet(title=sheet_name)
@@ -99,6 +96,52 @@ class RAGExcelParser:
                     ws.cell(row=row_num, column=col_num, value=value)
         return wb
 
+    @staticmethod
+    def _extract_images_from_worksheet(ws, sheetname=None):
+        """
+        Extract images from a worksheet and enrich them with vision-based descriptions.
+
+        Returns: List[dict]
+        """
+        images = getattr(ws, "_images", [])
+        if not images:
+            return []
+
+        raw_items = []
+
+        for img in images:
+            try:
+                img_bytes = img._data()
+                pil_img = Image.open(BytesIO(img_bytes)).convert("RGB")
+
+                anchor = img.anchor
+                if hasattr(anchor, "_from") and hasattr(anchor, "_to"):
+                    r1, c1 = anchor._from.row + 1, anchor._from.col + 1
+                    r2, c2 = anchor._to.row + 1, anchor._to.col + 1
+                    if r1 == r2 and c1 == c2:
+                        span = "single_cell"
+                    else:
+                        span = "multi_cell"
+                else:
+                    r1, c1 = anchor._from.row + 1, anchor._from.col + 1
+                    r2, c2 = r1, c1
+                    span = "single_cell"
+
+                item = {
+                    "sheet": sheetname or ws.title,
+                    "image": pil_img,
+                    "image_description": "",
+                    "row_from": r1,
+                    "col_from": c1,
+                    "row_to": r2,
+                    "col_to": c2,
+                    "span_type": span,
+                }
+                raw_items.append(item)
+            except Exception:
+                continue
+        return raw_items
+
     def html(self, fnm, chunk_rows=256):
         from html import escape
 
@@ -131,7 +174,7 @@ class RAGExcelParser:
                 tb = ""
                 tb += f"<table><caption>{sheetname}</caption>"
                 tb += tb_rows_0
-                for r in list(rows[1 + chunk_i * chunk_rows : min(1 + (chunk_i + 1) * chunk_rows, len(rows))]):
+                for r in list(rows[1 + chunk_i * chunk_rows: min(1 + (chunk_i + 1) * chunk_rows, len(rows))]):
                     tb += "<tr>"
                     for i, c in enumerate(r):
                         if c.value is None:
@@ -154,7 +197,7 @@ class RAGExcelParser:
         except Exception as e:
             logging.warning(f"Parse spreadsheet error: {e}, trying to interpret as CSV file")
             file_like_object.seek(0)
-            df = pd.read_csv(file_like_object)
+            df = pd.read_csv(file_like_object, on_bad_lines='skip')
         df = df.replace(r"^\s*$", "", regex=True)
         return df.to_markdown(index=False)
 
@@ -192,14 +235,14 @@ class RAGExcelParser:
         if fnm.split(".")[-1].lower().find("xls") >= 0:
             wb = RAGExcelParser._load_excel_to_workbook(BytesIO(binary))
             total = 0
-            
+
             for sheetname in wb.sheetnames:
-               try:
+                try:
-                   ws = wb[sheetname]
+                    ws = wb[sheetname]
-                   total += len(list(ws.rows))
+                    total += len(list(ws.rows))
-               except Exception as e:
+                except Exception as e:
-                   logging.warning(f"Skip sheet '{sheetname}' due to rows access error: {e}")
+                    logging.warning(f"Skip sheet '{sheetname}' due to rows access error: {e}")
-                   continue
+                    continue
             return total
 
         if fnm.split(".")[-1].lower() in ["csv", "txt"]: