Fix none approach streaming passthrough for tool-calling clients

optillm/server.py

@@ -203,24 +203,38 @@ def count_reasoning_tokens(text: str, tokenizer=None) -> int:
 
 plugin_approaches = {}
 
+def _is_multimodal_content(content: list) -> bool:
+    for item in content:
+        if not isinstance(item, dict):
+            continue
+        item_type = item.get('type')
+        if item_type and item_type != 'text':
+            return True
+        if 'image_url' in item or 'image' in item:
+            return True
+    return False
+
 def normalize_message_content(messages):
     """
-    Ensure all message content fields are strings, not lists.
-    Some models don't handle list-format content correctly.
+    Flatten text-only list content to strings for models that expect plain text.
+
+    Multimodal content (image_url, etc.) is preserved as a list so vision-capable
+    upstream models receive images intact.
     """
     normalized_messages = []
     for message in messages:
         normalized_message = message.copy()
         content = message.get('content', '')
 
-        # Convert list content to string if needed
         if isinstance(content, list):
-            # Extract text content from the list
-            text_content = ' '.join(
-                item.get('text', '') for item in content
-                if isinstance(item, dict) and item.get('type') == 'text'
-            )
-            normalized_message['content'] = text_content
+            if _is_multimodal_content(content):
+                normalized_message['content'] = content
+            else:
+                text_content = ' '.join(
+                    item.get('text', '') for item in content
+                    if isinstance(item, dict) and item.get('type') == 'text'
+                )
+                normalized_message['content'] = text_content
 
         normalized_messages.append(normalized_message)
 
@@ -548,6 +562,67 @@ def execute_n_times(n: int, approaches, operation: str, system_prompt: str, init
         return responses[0], total_tokens
     return responses, total_tokens
 
+def promote_tool_calls_to_first_choice(response_dict):
+    """Merge tool_calls from a non-zero choice into choices[0].
+
+    Some OpenAI-compatible providers return assistant text in choices[0] and
+    tool_calls in a later choice. Clients that only read choices[0] miss tools.
+    """
+    if not isinstance(response_dict, dict):
+        return response_dict
+
+    choices = response_dict.get('choices')
+    if not choices or len(choices) < 2:
+        return response_dict
+
+    tool_idx = None
+    for idx, choice in enumerate(choices):
+        message = choice.get('message') or {}
+        if message.get('tool_calls'):
+            tool_idx = idx
+            break
+
+    if tool_idx is None or tool_idx == 0:
+        return response_dict
+
+    promoted = dict(response_dict)
+    text_choice = dict(choices[0])
+    tool_choice = choices[tool_idx]
+    merged_message = dict(text_choice.get('message') or {})
+    tool_message = tool_choice.get('message') or {}
+
+    if tool_message.get('tool_calls'):
+        merged_message['tool_calls'] = tool_message['tool_calls']
+
+    text_choice['message'] = merged_message
+    text_choice['finish_reason'] = tool_choice.get('finish_reason') or 'tool_calls'
+    promoted['choices'] = [text_choice]
+    return promoted
+
+def generate_stream_passthrough(messages, client, model, kwargs, request_id=None):
+    """Stream upstream SSE chunks verbatim for transparent none-approach proxying."""
+    if model.startswith('none-'):
+        model = model[5:]
+
+    passthrough_kwargs = kwargs.copy() if kwargs else {}
+    passthrough_kwargs['stream'] = True
+    normalized_messages = normalize_message_content(messages)
+
+    stream = client.chat.completions.create(
+        model=model,
+        messages=normalized_messages,
+        **passthrough_kwargs
+    )
+
+    for chunk in stream:
+        chunk_dict = chunk.model_dump() if hasattr(chunk, 'model_dump') else chunk
+        yield "data: " + json.dumps(chunk_dict) + "\n\n"
+
+    yield "data: [DONE]\n\n"
+
+    if request_id:
+        logger.info(f'Request {request_id}: Completed (streaming passthrough)')
+
 def generate_streaming_response(final_response, model):
     # Generate a unique response ID
     response_id = f"chatcmpl-{int(time.time()*1000)}"
@@ -821,24 +896,35 @@ def proxy():
         contains_none = any(approach == 'none' for approach in approaches)
 
         if operation == 'SINGLE' and approaches[0] == 'none':
-            # Pass through the request including the n parameter
-            result, completion_tokens = execute_single_approach(approaches[0], system_prompt, initial_query, client, model, request_config, request_id)
+            passthrough_kwargs = request_config.copy() if request_config else {}
+
+            if stream:
+                if request_id:
+                    logger.info(f'Request {request_id}: Starting streaming passthrough')
+                return Response(
+                    generate_stream_passthrough(messages, client, model, passthrough_kwargs, request_id),
+                    content_type='text/event-stream'
+                )
+
+            passthrough_kwargs.pop('stream', None)
+            result = none_approach(
+                original_messages=messages,
+                client=client,
+                model=model,
+                request_id=request_id,
+                **passthrough_kwargs
+            )
+            result = promote_tool_calls_to_first_choice(result)
 
             logger.debug(f'Direct proxy response: {result}')
 
-            # Log the final response and finalize conversation logging
             if conversation_logger and request_id:
                 conversation_logger.log_final_response(request_id, result)
                 conversation_logger.finalize_conversation(request_id)
 
-            if stream:
-                if request_id:
-                    logger.info(f'Request {request_id}: Completed (streaming response)')
-                return Response(generate_streaming_response(extract_contents(result), model), content_type='text/event-stream')
-            else :
-                if request_id:
-                    logger.info(f'Request {request_id}: Completed')
-                return jsonify(result), 200
+            if request_id:
+                logger.info(f'Request {request_id}: Completed')
+            return jsonify(result), 200
 
         elif operation == 'AND' or operation == 'OR':
             if contains_none: