Bugfix/workspace file upload and csv encoding

application/single_app/config.py

@@ -98,7 +98,7 @@
 EXECUTOR_TYPE = 'thread'
 EXECUTOR_MAX_WORKERS = 30
 SESSION_TYPE = 'filesystem'
-VERSION = "0.261.029"
+VERSION = "0.261.030"
 IS_DEVELOPMENT = is_development_env_enabled()
 
 # Opt-out for deployments where App Service Easy Auth is active but the platform

application/single_app/functions_content.py

@@ -696,7 +696,8 @@ def extract_content_with_extraction_engine(
 def extract_table_file(file_path, file_ext):
     try:
         if file_ext == '.csv':
-            df = pandas.read_csv(file_path)
+            from functions_tabular_csv_query import read_tabular_csv
+            df = read_tabular_csv(file_path)
         elif file_ext in ['.xls', '.xlsx', '.xlsm']:
             df = pandas.read_excel(file_path)
         else:

application/single_app/functions_documents.py

@@ -61,6 +61,7 @@
     get_model_endpoint_api_type,
     resolve_model_endpoint_request_model,
 )
+from functions_tabular_csv_query import read_tabular_csv
 from model_endpoint_clients import MODEL_ENDPOINT_PROTOCOL_AZURE_OPENAI, infer_model_endpoint_protocol
 import azure.cognitiveservices.speech as speechsdk
 
@@ -8418,7 +8419,7 @@ def _build_minimal_tabular_summary(temp_file_path, original_filename, file_ext):
     if file_ext == '.csv':
         column_summary = "Column discovery unavailable"
         try:
-            header_df = pandas.read_csv(temp_file_path, keep_default_na=False, dtype=str, nrows=0)
+            header_df = read_tabular_csv(temp_file_path, keep_default_na=False, dtype=str, nrows=0)
             compact_columns = _compact_tabular_columns(header_df.columns.tolist())
             if compact_columns:
                 column_summary = ", ".join(compact_columns)
@@ -8465,7 +8466,7 @@ def _build_tabular_schema_summary(temp_file_path, original_filename, file_ext):
     plugin_note = "This file is available for detailed analysis via the Tabular Processing plugin."
 
     if file_ext == '.csv':
-        df_preview = pandas.read_csv(
+        df_preview = read_tabular_csv(
             temp_file_path,
             keep_default_na=False,
             dtype=str,
@@ -8703,7 +8704,7 @@ def process_tabular(document_id, user_id, temp_file_path, original_filename, fil
     if total_chunks_saved == 0 and not enable_enhanced_citations:
         try:
             if file_ext == '.csv':
-                df = pandas.read_csv(
+                df = read_tabular_csv(
                     temp_file_path,
                     keep_default_na=False,
                     dtype=str

application/single_app/functions_tabular_csv_query.py

@@ -2,10 +2,50 @@
 """Shared bounded CSV query evaluation for foreground tools and durable exports."""
 
 import ast
+import io
+import os
 
 import pandas
 
 
+TABULAR_CSV_ENCODINGS = ('utf-8-sig', 'utf-8', 'cp1252', 'latin-1')
+
+
+def read_tabular_csv(source, **read_options):
+    """Read a CSV after resolving UTF-8 and common ANSI encodings."""
+    if isinstance(source, (str, os.PathLike)):
+        with open(source, 'rb') as source_file:
+            source_bytes = source_file.read()
+    elif hasattr(source, 'read'):
+        current_position = source.tell() if hasattr(source, 'tell') else None
+        if hasattr(source, 'seek'):
+            source.seek(0)
+        source_value = source.read()
+        if current_position is not None and hasattr(source, 'seek'):
+            source.seek(current_position)
+        source_bytes = source_value.encode('utf-8') if isinstance(source_value, str) else source_value
+    else:
+        source_bytes = source
+
+    if not isinstance(source_bytes, bytes):
+        raise TypeError('CSV source must be a path, byte stream, or bytes')
+
+    for encoding in TABULAR_CSV_ENCODINGS:
+        try:
+            decoded_csv = source_bytes.decode(encoding)
+            return pandas.read_csv(io.StringIO(decoded_csv), **read_options)
+        except UnicodeDecodeError:
+            continue
+
+    raise UnicodeDecodeError(
+        'tabular_csv',
+        source_bytes,
+        0,
+        len(source_bytes),
+        'Unable to decode CSV using supported encodings',
+    )
+
+
 TABULAR_ROW_LOCAL_QUERY_AST_NODES = (
     ast.Expression,
     ast.BoolOp,
@@ -240,8 +280,7 @@ def validate_tabular_csv_query_expression(query_expression):
 def detect_tabular_csv_numeric_columns(csv_stream, source_chunk_rows, tabular_plugin):
     """Find columns that pandas can convert to numeric across every bounded chunk."""
     numeric_columns = None
-    csv_stream.seek(0)
-    for source_chunk in pandas.read_csv(
+    for source_chunk in read_tabular_csv(
         csv_stream,
         keep_default_na=False,
         dtype=str,
@@ -255,7 +294,6 @@ def detect_tabular_csv_numeric_columns(csv_stream, source_chunk_rows, tabular_pl
                 pandas.to_numeric(source_chunk[column_name])
             except (TypeError, ValueError):
                 numeric_columns.discard(column_name)
-    csv_stream.seek(0)
     return numeric_columns or set()
 
 
@@ -279,7 +317,6 @@ def iter_tabular_csv_query_rows(
     )
     parsed_return_columns = tabular_plugin._parse_optional_column_list_argument(return_columns)
 
-    csv_stream.seek(0)
     read_options = {
         'keep_default_na': False,
         'dtype': str,
@@ -289,7 +326,7 @@ def iter_tabular_csv_query_rows(
         read_options['skiprows'] = lambda row_index: 0 < row_index <= start_source_row
 
     source_row_offset = start_source_row
-    for source_chunk in pandas.read_csv(csv_stream, **read_options):
+    for source_chunk in read_tabular_csv(csv_stream, **read_options):
         source_chunk = tabular_plugin._normalize_dataframe_columns(source_chunk)
         source_chunk.index = range(source_row_offset, source_row_offset + len(source_chunk))
         source_row_offset += len(source_chunk)

application/single_app/route_backend_chats.py

@@ -9534,6 +9534,50 @@ def summarize_tabular_invocation_errors(invocations):
     return unique_errors
 
 
+def build_tabular_invocation_failure_signature(invocation):
+    """Build a stable signature for detecting repeated equivalent tool failures."""
+    error_message = get_tabular_invocation_error_message(invocation)
+    if not error_message:
+        return None
+
+    parameters = getattr(invocation, 'parameters', {}) or {}
+    comparable_parameters = {
+        str(parameter_name): parameter_value
+        for parameter_name, parameter_value in parameters.items()
+        if parameter_name not in {'user_id', 'conversation_id'}
+    }
+    normalized_error = re.sub(r'\s+', ' ', str(error_message).strip()).casefold()
+    return (
+        str(getattr(invocation, 'function_name', '') or '').strip(),
+        json.dumps(comparable_parameters, sort_keys=True, default=str),
+        normalized_error,
+    )
+
+
+def get_repeated_tabular_invocation_failures(invocations, minimum_repeats=2):
+    """Return repeated equivalent failures with their function and safe error text."""
+    failures_by_signature = {}
+    for invocation in invocations or []:
+        signature = build_tabular_invocation_failure_signature(invocation)
+        if signature is None:
+            continue
+        failures_by_signature.setdefault(signature, []).append(invocation)
+
+    repeated_failures = []
+    for (function_name, _parameters, normalized_error), matching_invocations in failures_by_signature.items():
+        if len(matching_invocations) < minimum_repeats:
+            continue
+        error_message = get_tabular_invocation_error_message(matching_invocations[0])
+        repeated_failures.append({
+            'function_name': function_name,
+            'count': len(matching_invocations),
+            'error_message': error_message,
+            'normalized_error': normalized_error,
+        })
+
+    return repeated_failures
+
+
 def summarize_tabular_discovery_invocations(invocations, max_sheet_names=6):
     """Return compact workbook-discovery summaries for retry prompts."""
     discovery_summaries = []
@@ -12730,6 +12774,42 @@ def build_system_prompt(force_tool_use=False, tool_error_messages=None,
                 else:
                     successful_schema_summary_invocations.append(invocation)
 
+            repeated_failures = get_repeated_tabular_invocation_failures(
+                failed_analytical_invocations + failed_schema_summary_invocations,
+            )
+            repeated_failure_feedback_messages = []
+            if repeated_failures:
+                repeated_failure = repeated_failures[0]
+                safe_repeated_failure_error = sanitize_plugin_invocation_value(
+                    repeated_failure['error_message']
+                )
+                repeated_failure_feedback_messages.append(
+                    f"The tool call {repeated_failure['function_name']} produced the same error repeatedly. Do not repeat that exact call. Change the filename, sheet, column, arguments, or use a different analytical function that addresses the user's question."
+                )
+                log_event(
+                    '[TABULAR_SK_ANALYSIS] Repeated equivalent tool failure detected; routing away from the failed call',
+                    extra={
+                        'function_name': repeated_failure['function_name'],
+                        'repeat_count': repeated_failure['count'],
+                        'error_message': repeated_failure['error_message'],
+                        'attempt_number': attempt_number,
+                    },
+                    level=logging.ERROR,
+                )
+                await emit_tabular_analysis_lifecycle_thought(
+                    thought_callback,
+                    f"Tabular tool {repeated_failure['function_name']} failed repeatedly",
+                    detail=(
+                        f"Failed {repeated_failure['count']} times with the same error: "
+                        f"{safe_repeated_failure_error}"
+                    ),
+                    title='Tabular analysis needs a different query path',
+                    state='running',
+                    phase='retry',
+                    attempt_number=attempt_number,
+                    attempt_count=3,
+                )
+
             if synthesis_exception is not None:
                 raw_tool_fallback = None
                 if not schema_summary_mode:
@@ -12785,6 +12865,7 @@ def build_system_prompt(force_tool_use=False, tool_error_messages=None,
 
                     if failed_schema_summary_invocations:
                         previous_tool_error_messages = summarize_tabular_invocation_errors(failed_schema_summary_invocations)
+                        previous_execution_gap_messages = repeated_failure_feedback_messages
                         log_event(
                             f"[TABULAR_SK_ANALYSIS] Attempt {attempt_number} used workbook schema tool(s) but all returned errors; retrying",
                             extra={
@@ -12890,7 +12971,7 @@ def build_system_prompt(force_tool_use=False, tool_error_messages=None,
 
                     if failed_analytical_invocations:
                         previous_tool_error_messages = summarize_tabular_invocation_errors(failed_analytical_invocations)
-                        previous_execution_gap_messages = []
+                        previous_execution_gap_messages = repeated_failure_feedback_messages
                         retry_sheet_overrides = get_tabular_retry_sheet_overrides(failed_analytical_invocations)
                         for workbook_name, override_payload in retry_sheet_overrides.items():
                             blob_location = workbook_blob_locations.get(workbook_name)

application/single_app/route_backend_documents.py

@@ -444,7 +444,8 @@ def get_file_content():
                     file_ext = os.path.splitext(filename)[1].lower()
                     if file_ext == '.csv':
                         import pandas
-                        df = pandas.read_csv(io.BytesIO(blob_data))
+                        from functions_tabular_csv_query import read_tabular_csv
+                        df = read_tabular_csv(io.BytesIO(blob_data))
                         combined_content = df.to_csv(index=False)
                     elif file_ext in ['.xlsx', '.xlsm']:
                         import pandas

application/single_app/route_enhanced_citations.py

@@ -929,7 +929,8 @@ def get_enhanced_citation_tabular_preview():
             selected_sheet = None
             sheet_names = []
             if ext == 'csv':
-                df = pandas.read_csv(io.BytesIO(data), keep_default_na=False, dtype=str, nrows=nrows_limit)
+                from functions_tabular_csv_query import read_tabular_csv
+                df = read_tabular_csv(io.BytesIO(data), keep_default_na=False, dtype=str, nrows=nrows_limit)
             elif ext in ('xlsx', 'xlsm'):
                 excel_file = pandas.ExcelFile(io.BytesIO(data), engine='openpyxl')
                 sheet_names = list(excel_file.sheet_names)

application/single_app/route_frontend_chats.py

@@ -868,7 +868,14 @@ def chats():
         user_groups_raw = []
         try:
             user_groups_raw = get_user_groups(user_id)
-            user_groups_simple = [{'id': g['id'], 'name': g.get('name', 'Unnamed')} for g in user_groups_raw]
+            user_groups_simple = [
+                {
+                    'id': group['id'],
+                    'name': group.get('name', 'Unnamed'),
+                    'userRole': get_user_role_in_group(group, user_id),
+                }
+                for group in user_groups_raw
+            ]
         except Exception as e:
             logger.warning(f"Failed to load user groups for chats page: {e}")
 

application/single_app/semantic_kernel_plugins/tabular_processing_plugin.py

@@ -27,6 +27,7 @@
 from functions_authentication import get_current_user_id
 from functions_tabular_csv_query import (
     iter_tabular_csv_query_rows,
+    read_tabular_csv,
     validate_tabular_csv_query_expression,
 )
 from functions_group import find_group_by_id, get_user_role_in_group
@@ -3113,7 +3114,7 @@ def _read_tabular_blob_to_dataframe(
 
         name_lower = blob_name.lower()
         if name_lower.endswith('.csv'):
-            df = pandas.read_csv(io.BytesIO(data), keep_default_na=False, dtype=str)
+            df = read_tabular_csv(io.BytesIO(data), keep_default_na=False, dtype=str)
         elif name_lower.endswith('.xlsx') or name_lower.endswith('.xlsm'):
             df = pandas.read_excel(
                 io.BytesIO(data),

docs/explanation/fixes/CHAT_GROUP_UPLOAD_OWNER_ROLE_FIX.md

@@ -0,0 +1,23 @@
+# Chat Group Upload Owner Role Fix
+
+## Issue
+
+Owners of group-scoped conversations were told that no group workspace was available for upload, even though owners are permitted to upload group documents.
+
+## Root Cause
+
+The chat page passed group IDs and names to the upload JavaScript but omitted each user's resolved group role. The client therefore treated owners as users without a group upload role before the upload request was sent.
+
+## Version
+
+Fixed/Implemented in version: **0.261.029**
+
+The application version was updated in `application/single_app/config.py`.
+
+## Technical Details
+
+The chat bootstrap payload now includes `userRole` for each group. This allows the existing `Owner`, `Admin`, and `DocumentManager` upload allowlist to operate correctly. Server-side group membership, status, and upload-role validation remain authoritative.
+
+## Validation
+
+The group upload handoff functional test includes a contract assertion that the resolved role is passed to the frontend. The full existing test currently has two unrelated stale assertions for older frontend/search contracts; the route syntax and version/role contract were validated separately.

docs/explanation/fixes/TABULAR_CSV_ANSI_ENCODING_AND_RETRY_FIX.md

@@ -0,0 +1,23 @@
+# Tabular CSV ANSI Encoding and Retry Fix
+
+## Issue
+
+ANSI-encoded CSV files could upload successfully while producing unreadable metadata and failed `search_rows` calls. Repeated tool failures could also consume a long automatic-invocation sequence before the outer tabular retry logic completed, without clearly reporting the failure in the live analysis stream.
+
+## Root Cause
+
+CSV readers passed raw files directly to pandas, which defaults to UTF-8 and does not reliably decode Windows-1252 content. Tabular analysis had no targeted circuit breaker for equivalent failed calls, so the model could repeat the same invalid request while consuming its automatic-invocation budget.
+
+## Implemented in version: **0.261.030**
+
+## Technical Details
+
+- Added shared CSV decoding that prefers UTF-8, then UTF-8 with BOM, Windows-1252, and Latin-1.
+- Applied the decoder to tabular metadata, indexing, citation previews, foreground queries, and durable row replay.
+- Restored the broader 20-call automatic-invocation budget so complex analysis is not cut short by a universal cap.
+- Detect repeated equivalent failures by function, arguments, and normalized error, then route the next model pass away from the failed call shape.
+- Emit an explicit tabular retry lifecycle thought and server-side failure event when repeated failures are detected.
+
+## Validation
+
+The regression test in `functional_tests/test_tabular_csv_ansi_encoding.py` verifies Windows-1252 characters are preserved, the 20-call budget remains available for complex analysis, and repeated-failure routing is present.

docs/explanation/release_notes.md

@@ -2,6 +2,25 @@
 
 For feature-focused and fix-focused drill-downs by version, see [Features by Version](https://github.com/microsoft/simplechat/tree/main/docs/explanation/features) and [Fixes by Version](https://github.com/microsoft/simplechat/tree/main/docs/explanation/fixes).
 
+### **(v0.261.030)**
+
+#### Bug Fixes
+
+*   **ANSI-Encoded CSV Files Are Now Read Correctly**
+    *   CSV metadata extraction, indexing, citations, row searches, and durable tabular replay now support UTF-8, UTF-8 with BOM, Windows-1252, and Latin-1 files, preserving characters that previously made uploaded content appear unreadable.
+    *   Tabular analysis retains the broader automatic-invocation budget for complex questions while detecting repeated equivalent failures and routing the next model pass to a different call shape instead of repeating the same error.
+    *   Repeated tabular tool failures now emit an explicit retry lifecycle thought and server-side diagnostic event, making the failure visible while recovery continues.
+    *   (Ref: `functions_tabular_csv_query.py`, `functions_documents.py`, `tabular_processing_plugin.py`, `route_backend_chats.py`, [Tabular CSV ANSI Encoding and Retry Fix](fixes/TABULAR_CSV_ANSI_ENCODING_AND_RETRY_FIX.md))
+
+### **(v0.261.029)**
+
+#### Bug Fixes
+
+*   **Group Chat Uploads Now Recognize Workspace Owners**
+    *   Fixed group-scoped chat uploads incorrectly reporting that no group workspace was available when the signed-in user was the group owner.
+    *   Chat bootstrap data now includes each group's resolved user role, allowing owners, admins, and document managers to use the existing group upload permissions while preserving server-side authorization checks.
+    *   (Ref: `route_frontend_chats.py`, `chat-input-actions.js`, [Chat Group Upload Owner Role Fix](fixes/CHAT_GROUP_UPLOAD_OWNER_ROLE_FIX.md))
+
 ### **(v0.261.028)**
 
 Tracking: [#1489](https://github.com/microsoft/simplechat/issues/1489); implementation: [PR #1488](https://github.com/microsoft/simplechat/pull/1488).