Update app.py
Browse files
app.py
CHANGED
|
@@ -162,16 +162,20 @@ def get_ticket_documents_iterator(file_path, file_extension, filename, chunksize
|
|
| 162 |
|
| 163 |
if file_extension in ['.xlsx', '.xls']:
|
| 164 |
try:
|
| 165 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 166 |
total_rows_processed = 0
|
| 167 |
-
|
| 168 |
-
|
| 169 |
-
|
| 170 |
|
| 171 |
chunk_documents = []
|
| 172 |
for index, row in df_chunk.iterrows():
|
| 173 |
ticket_content_parts = []
|
| 174 |
-
ticket_id = f"Row {
|
| 175 |
|
| 176 |
for col_name, value in row.items():
|
| 177 |
if pd.notna(value) and str(value).strip() != '':
|
|
@@ -186,11 +190,11 @@ def get_ticket_documents_iterator(file_path, file_extension, filename, chunksize
|
|
| 186 |
else:
|
| 187 |
ticket_content = "\n".join(ticket_content_parts)
|
| 188 |
|
| 189 |
-
metadata = {"source": filename, "ticket_row_number":
|
| 190 |
metadata.update(row.to_dict()) # Include all original row data in metadata
|
| 191 |
|
| 192 |
chunk_documents.append({"page_content": ticket_content, "metadata": metadata})
|
| 193 |
-
total_rows_processed += 1
|
| 194 |
|
| 195 |
yield chunk_documents, total_rows_processed # Yield chunk and running total
|
| 196 |
|
|
@@ -359,8 +363,24 @@ def parse_markdown_table_to_df(markdown_string):
|
|
| 359 |
break
|
| 360 |
|
| 361 |
if header_line is None or separator_line_index == -1:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 362 |
return pd.DataFrame(), "No valid Markdown table header or separator found."
|
| 363 |
|
|
|
|
| 364 |
data_lines = []
|
| 365 |
for line in lines[separator_line_index + 1:]:
|
| 366 |
if line.strip().startswith('|') and '|' in line.strip():
|
|
@@ -596,25 +616,43 @@ def perform_ticket_analysis(file, user_query, selected_timezone, progress=gr.Pro
|
|
| 596 |
category_count = len(category_specific_tickets_df)
|
| 597 |
yield from append_log_message(f"Analyzing category: '{category}' with {category_count} tickets.")
|
| 598 |
|
|
|
|
| 599 |
detail_prompt = f"""
|
| 600 |
-
You are an expert customer support ticket analyst.
|
| 601 |
-
|
| 602 |
-
Your
|
| 603 |
-
|
| 604 |
-
**
|
| 605 |
-
|
| 606 |
-
|
| 607 |
-
|
| 608 |
-
|
| 609 |
-
|
| 610 |
-
|
| 611 |
-
|
| 612 |
-
|
| 613 |
-
|
| 614 |
-
|
| 615 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 616 |
{category_tickets_content}
|
| 617 |
"""
|
|
|
|
| 618 |
|
| 619 |
try:
|
| 620 |
yield from append_log_message(f"Calling LLM for detailed analysis of category '{category}'...")
|
|
@@ -633,12 +671,17 @@ def perform_ticket_analysis(file, user_query, selected_timezone, progress=gr.Pro
|
|
| 633 |
"Sub-Category": "N/A",
|
| 634 |
"Count": category_count,
|
| 635 |
"Issue Description (Summary)": category_specific_tickets_df['issue_summary_short'].iloc[0] if not category_specific_tickets_df.empty else "Error in summary generation/parsing for this category.",
|
| 636 |
-
"Root Cause (Inferred)": "Undetermined due to parsing error.",
|
| 637 |
"Potential Resolution Steps": "N/A",
|
| 638 |
"Future Prediction": "N/A",
|
| 639 |
"Proactive Measures": "N/A"
|
| 640 |
}
|
|
|
|
|
|
|
|
|
|
|
|
|
| 641 |
final_analysis_rows.append(default_row)
|
|
|
|
| 642 |
else:
|
| 643 |
detailed_row = detailed_df.iloc[0].to_dict()
|
| 644 |
detailed_row['Category'] = category # Ensure category is correctly assigned
|
|
@@ -656,6 +699,10 @@ def perform_ticket_analysis(file, user_query, selected_timezone, progress=gr.Pro
|
|
| 656 |
else: # Add 'Count' if missing from AI output
|
| 657 |
detailed_row['Count'] = category_count
|
| 658 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 659 |
final_analysis_rows.append(detailed_row)
|
| 660 |
|
| 661 |
except Exception as e:
|
|
@@ -729,7 +776,7 @@ def perform_ticket_analysis(file, user_query, selected_timezone, progress=gr.Pro
|
|
| 729 |
|
| 730 |
|
| 731 |
html_output = f"""
|
| 732 |
-
<div style="font-family: Arial, sans-serif; padding: 20px; background-color: #f9f9f9; border-radius: 10px; box-shadow: 0
|
| 733 |
<h1 align="center">🕵️♀️ Executive Summary: Customer Ticket Analysis 🕵️♀️</h1>
|
| 734 |
|
| 735 |
<div style="background-color: #E8EAF6; padding: 15px; border-radius: 8px; margin-bottom: 20px;">
|
|
@@ -742,7 +789,7 @@ def perform_ticket_analysis(file, user_query, selected_timezone, progress=gr.Pro
|
|
| 742 |
Total Original Tickets in File: <strong>{total_tickets_count}</strong>.<br>
|
| 743 |
Total Tickets Summarized in Report: <strong>{summarized_count}</strong>.<br>
|
| 744 |
Total Categories Identified: <strong>{len(unique_categories) if 'unique_categories' in locals() else 0}</strong>.
|
| 745 |
-
{"<p style='color:red; font-weight:bold;'>WARNING: The number of tickets summarized in the report does not match the total original tickets. This
|
| 746 |
</p>
|
| 747 |
</div>
|
| 748 |
"""
|
|
|
|
| 162 |
|
| 163 |
if file_extension in ['.xlsx', '.xls']:
|
| 164 |
try:
|
| 165 |
+
# Read the entire Excel file into a DataFrame first
|
| 166 |
+
df_full = pd.read_excel(file_path, engine='openpyxl')
|
| 167 |
+
if df_full.empty:
|
| 168 |
+
raise ValueError("The uploaded Excel file contains no data.")
|
| 169 |
+
|
| 170 |
total_rows_processed = 0
|
| 171 |
+
# Manually chunk the DataFrame
|
| 172 |
+
for i in range(0, len(df_full), chunksize):
|
| 173 |
+
df_chunk = df_full.iloc[i : i + chunksize]
|
| 174 |
|
| 175 |
chunk_documents = []
|
| 176 |
for index, row in df_chunk.iterrows():
|
| 177 |
ticket_content_parts = []
|
| 178 |
+
ticket_id = f"Row {index + 1}" # Default ticket ID based on original DataFrame index
|
| 179 |
|
| 180 |
for col_name, value in row.items():
|
| 181 |
if pd.notna(value) and str(value).strip() != '':
|
|
|
|
| 190 |
else:
|
| 191 |
ticket_content = "\n".join(ticket_content_parts)
|
| 192 |
|
| 193 |
+
metadata = {"source": filename, "ticket_row_number": index + 1, "ticket_id": ticket_id}
|
| 194 |
metadata.update(row.to_dict()) # Include all original row data in metadata
|
| 195 |
|
| 196 |
chunk_documents.append({"page_content": ticket_content, "metadata": metadata})
|
| 197 |
+
total_rows_processed += 1 # Update count based on actual processed rows in this chunk
|
| 198 |
|
| 199 |
yield chunk_documents, total_rows_processed # Yield chunk and running total
|
| 200 |
|
|
|
|
| 363 |
break
|
| 364 |
|
| 365 |
if header_line is None or separator_line_index == -1:
|
| 366 |
+
# If no header/separator found, assume the first non-empty line is a data row if it looks like one
|
| 367 |
+
if any(line.strip().startswith('|') and '|' in line.strip() for line in lines):
|
| 368 |
+
# Attempt to parse as if it's just a data row without a header.
|
| 369 |
+
# This is a fallback and might not be perfect.
|
| 370 |
+
potential_data_line = next((line for line in lines if line.strip().startswith('|') and '|' in line.strip()), None)
|
| 371 |
+
if potential_data_line:
|
| 372 |
+
# Create dummy headers if no header was provided
|
| 373 |
+
num_cols = len([v for v in potential_data_line.strip('|').split('|') if v])
|
| 374 |
+
headers = [f"Col{i+1}" for i in range(num_cols)]
|
| 375 |
+
data = [[v.strip() for v in potential_data_line.strip('|').split('|') if v]]
|
| 376 |
+
if len(data[0]) == len(headers):
|
| 377 |
+
df = pd.DataFrame(data, columns=headers)
|
| 378 |
+
df = df.dropna(axis=1, how='all')
|
| 379 |
+
df.columns = df.columns.str.strip()
|
| 380 |
+
return df, "No explicit Markdown table header or separator found, parsed as a single data row."
|
| 381 |
return pd.DataFrame(), "No valid Markdown table header or separator found."
|
| 382 |
|
| 383 |
+
|
| 384 |
data_lines = []
|
| 385 |
for line in lines[separator_line_index + 1:]:
|
| 386 |
if line.strip().startswith('|') and '|' in line.strip():
|
|
|
|
| 616 |
category_count = len(category_specific_tickets_df)
|
| 617 |
yield from append_log_message(f"Analyzing category: '{category}' with {category_count} tickets.")
|
| 618 |
|
| 619 |
+
# --- START: OPTIMIZED PROMPT FOR DETAILED ANALYSIS ---
|
| 620 |
detail_prompt = f"""
|
| 621 |
+
You are an expert customer support ticket analyst, specializing in NLP and data interpretation.
|
| 622 |
+
Deeply analyze the following raw customer support tickets, which all belong to the **'{category}'** category.
|
| 623 |
+
Your goal is to synthesize the information into a single, comprehensive markdown table row, providing highly contextualized and actionable insights for this specific category.
|
| 624 |
+
|
| 625 |
+
**Crucial Instructions:**
|
| 626 |
+
1. **Strict Output Format:** Your output MUST be a **single Markdown table row**. Do NOT include the header row, only the data row.
|
| 627 |
+
**Table Columns (MUST be in this exact order):** `Category`, `Sub-Category`, `Count`, `Issue Description (Summary)`, `Root Cause (Inferred)`, `Potential Resolution Steps`, `Future Prediction`, `Proactive Measures`.
|
| 628 |
+
2. **Category:** This MUST be '{category}'.
|
| 629 |
+
3. **Count:** This MUST be the **exact number of tickets provided for this category ({category_count})**.
|
| 630 |
+
4. **Issue Description (Summary):**
|
| 631 |
+
* Synthesize the **common themes and recurring patterns** observed across ALL tickets in this category.
|
| 632 |
+
* Identify and mention **key affected systems, user actions, or specific features** causing issues.
|
| 633 |
+
* Keep it concise (3-5 sentences), focusing on the most prevalent aspects.
|
| 634 |
+
5. **Sub-Category:**
|
| 635 |
+
* Based on internal patterns, identify one or two **most prominent, granular sub-categories**.
|
| 636 |
+
* If no clear sub-category emerges, state 'General {category}' or 'N/A'.
|
| 637 |
+
* Examples: For 'Technical Issue' -> 'Login Failure', 'Software Bug'; for 'Billing Inquiry' -> 'Incorrect Charge', 'Subscription Error'.
|
| 638 |
+
6. **Root Cause (Inferred):**
|
| 639 |
+
* Perform an **in-depth inference** of the **underlying systemic or causal factors** leading to these issues. Do not just restate the symptom.
|
| 640 |
+
* Think about why the problem exists. (e.g., "System misconfiguration", "User error due to lack of documentation", "Third-party integration failure", "Defect in recent software update").
|
| 641 |
+
* If truly undetermined from the data, state 'Undetermined from available data'.
|
| 642 |
+
7. **Potential Resolution Steps:**
|
| 643 |
+
* Suggest **actionable, generic steps** that would effectively resolve or mitigate the issues in this category.
|
| 644 |
+
* These should be **strategic resolutions**, not just individual ticket fixes. (e.g., "Develop new API endpoint", "Update user guides", "Patch database system", "Conduct user training").
|
| 645 |
+
8. **Future Prediction:**
|
| 646 |
+
* Based on the trends and nature of these tickets, predict **what might happen if the underlying issues are not addressed**.
|
| 647 |
+
* Consider potential increase in ticket volume, customer churn, or emergence of new, related problems. (e.g., "Increased call volume for billing disputes", "Decline in user retention for feature X", "Escalation to higher severity issues").
|
| 648 |
+
9. **Proactive Measures:**
|
| 649 |
+
* Suggest **specific, actionable steps to prevent recurrence or significantly reduce the volume** of these types of tickets in the future.
|
| 650 |
+
* Focus on **system improvements, process changes, documentation, or user education**. (e.g., "Implement automated system health checks", "Revise onboarding flow documentation", "Roll out a mandatory security update", "Train support staff on new refund process").
|
| 651 |
+
|
| 652 |
+
Context (Raw Customer Tickets for Category '{category}' - Total {category_count} tickets):
|
| 653 |
{category_tickets_content}
|
| 654 |
"""
|
| 655 |
+
# --- END: OPTIMIZED PROMPT FOR DETAILED ANALYSIS ---
|
| 656 |
|
| 657 |
try:
|
| 658 |
yield from append_log_message(f"Calling LLM for detailed analysis of category '{category}'...")
|
|
|
|
| 671 |
"Sub-Category": "N/A",
|
| 672 |
"Count": category_count,
|
| 673 |
"Issue Description (Summary)": category_specific_tickets_df['issue_summary_short'].iloc[0] if not category_specific_tickets_df.empty else "Error in summary generation/parsing for this category.",
|
| 674 |
+
"Root Cause (Inferred)": "Undetermined due to parsing error or insufficient data.",
|
| 675 |
"Potential Resolution Steps": "N/A",
|
| 676 |
"Future Prediction": "N/A",
|
| 677 |
"Proactive Measures": "N/A"
|
| 678 |
}
|
| 679 |
+
# Add a check to ensure all expected columns are present, add if missing
|
| 680 |
+
for col in ['Sub-Category', 'Root Cause (Inferred)', 'Potential Resolution Steps', 'Future Prediction', 'Proactive Measures']:
|
| 681 |
+
if col not in default_row:
|
| 682 |
+
default_row[col] = 'N/A'
|
| 683 |
final_analysis_rows.append(default_row)
|
| 684 |
+
|
| 685 |
else:
|
| 686 |
detailed_row = detailed_df.iloc[0].to_dict()
|
| 687 |
detailed_row['Category'] = category # Ensure category is correctly assigned
|
|
|
|
| 699 |
else: # Add 'Count' if missing from AI output
|
| 700 |
detailed_row['Count'] = category_count
|
| 701 |
|
| 702 |
+
# Ensure all required columns are present in the parsed row
|
| 703 |
+
for col in ['Sub-Category', 'Root Cause (Inferred)', 'Potential Resolution Steps', 'Future Prediction', 'Proactive Measures']:
|
| 704 |
+
if col not in detailed_row:
|
| 705 |
+
detailed_row[col] = 'N/A' # Default to N/A if LLM missed a column
|
| 706 |
final_analysis_rows.append(detailed_row)
|
| 707 |
|
| 708 |
except Exception as e:
|
|
|
|
| 776 |
|
| 777 |
|
| 778 |
html_output = f"""
|
| 779 |
+
<div style="font-family: Arial, sans-serif; padding: 20px; background-color: #f9f9f9; border-radius: 10px; box-shadow: 0 4px 8px rgba(0,0,0,0.1);">
|
| 780 |
<h1 align="center">🕵️♀️ Executive Summary: Customer Ticket Analysis 🕵️♀️</h1>
|
| 781 |
|
| 782 |
<div style="background-color: #E8EAF6; padding: 15px; border-radius: 8px; margin-bottom: 20px;">
|
|
|
|
| 789 |
Total Original Tickets in File: <strong>{total_tickets_count}</strong>.<br>
|
| 790 |
Total Tickets Summarized in Report: <strong>{summarized_count}</strong>.<br>
|
| 791 |
Total Categories Identified: <strong>{len(unique_categories) if 'unique_categories' in locals() else 0}</strong>.
|
| 792 |
+
{"<p style='color:red; font-weight:bold;'>WARNING: The number of tickets summarized in the report does not match the total original tickets. This might indicate a parsing or AI categorization issue.</p>" if summarized_count != total_tickets_count else ""}
|
| 793 |
</p>
|
| 794 |
</div>
|
| 795 |
"""
|