feat: Add markdown table parsing for XLSX output.

agent/component/message.py

@@ -232,39 +232,66 @@ class Message(ComponentBase):
             elif self._param.output_format == "xlsx":
                 import pandas as pd
                 from io import BytesIO
-                from datetime import datetime
+                import re
 
-                # Get all conversation history from canvas
+                # Try to parse markdown table from the content
-                history = getattr(self._canvas, 'history', [])
+                df = None
                 
-                # Build rows from conversation history
+                if isinstance(content, str):
-                rows = []
+                    # Extract markdown table from content
-                for role, msg in history:
+                    # Pattern: lines starting with | and containing |
-                    if isinstance(msg, dict):
+                    lines = content.strip().split('\n')
-                        msg_content = msg.get("content", str(msg))
+                    table_lines = []
-                    else:
+                    in_table = False
-                        msg_content = str(msg) if msg else ""
-                    rows.append({
-                        "Role": role,
-                        "Content": msg_content,
-                    })
                     
-                # Add current message if not already in history
+                    for line in lines:
-                if content and (not rows or rows[-1].get("Content") != content):
+                        line = line.strip()
-                    rows.append({
+                        if line.startswith('|') and '|' in line[1:]:
-                        "Role": "assistant",
+                            in_table = True
-                        "Content": content,
+                            # Skip separator line (|---|---| or |:---:|:---:| etc.)
-                    })
+                            # Check if line only contains |, -, :, and whitespace
+                            cleaned = line.replace(' ', '').replace('|', '').replace('-', '').replace(':', '')
+                            if cleaned == '':
+                                continue  # Skip separator line
+                            table_lines.append(line)
+                        elif in_table and not line.startswith('|'):
+                            # End of table
+                            break
                     
-                if rows:
+                    if table_lines:
-                    df = pd.DataFrame(rows)
+                        # Parse the markdown table
-                else:
+                        rows = []
-                    df = pd.DataFrame({"Role": ["assistant"], "Content": [content if content else ""]})
+                        headers = None
+                        
+                        for line in table_lines:
+                            # Split by | and clean up
+                            cells = [cell.strip() for cell in line.split('|')]
+                            # Remove empty first and last elements from split
+                            cells = [c for c in cells if c]
+                            
+                            if headers is None:
+                                headers = cells
+                            else:
+                                rows.append(cells)
+                        
+                        if headers and rows:
+                            # Ensure all rows have same number of columns as headers
+                            normalized_rows = []
+                            for row in rows:
+                                while len(row) < len(headers):
+                                    row.append('')
+                                normalized_rows.append(row[:len(headers)])
+                            
+                            df = pd.DataFrame(normalized_rows, columns=headers)
+                
+                # Fallback: if no table found, create single column with content
+                if df is None or df.empty:
+                    df = pd.DataFrame({"Content": [content if content else ""]})
 
                 # Write to Excel
                 excel_io = BytesIO()
                 with pd.ExcelWriter(excel_io, engine='openpyxl') as writer:
-                    df.to_excel(writer, sheet_name="Conversation", index=False)
+                    df.to_excel(writer, sheet_name="Data", index=False)
                 
                 excel_io.seek(0)
                 binary_content = excel_io.read()