diff --git a/app/schemas/chat_robot.py b/app/schemas/chat_robot.py
index cebf74a..01feeae 100644
--- a/app/schemas/chat_robot.py
+++ b/app/schemas/chat_robot.py
@@ -2,7 +2,6 @@ from pydantic import BaseModel
 
 
 class ChatRobotModel(BaseModel):
-    gender: str
     message: str
     session_id: str
     user_id: int
diff --git a/app/service/chat_robot/script/main.py b/app/service/chat_robot/script/main.py
index 3342a5c..3890a0e 100644
--- a/app/service/chat_robot/script/main.py
+++ b/app/service/chat_robot/script/main.py
@@ -92,7 +92,6 @@ def chat(post_data):
     user_id = post_data.user_id
     session_id = post_data.session_id
     input_message = post_data.message
-    gender = post_data.gender
 
     # final_outputs = agent_executor(
     #     {"input": input_message, "gender": gender},
@@ -100,7 +99,7 @@ def chat(post_data):
     #     session_key=f"buffer:{user_id}:{session_id}",
     # )
 
-    final_outputs = CallQWen.call_with_messages(input_message, gender)
+    final_outputs = CallQWen.call_with_messages(input_message)
     # api_response = {
     #     'user_id': user_id,
     #     'session_id': session_id,
diff --git a/app/service/chat_robot/script/prompt.py b/app/service/chat_robot/script/prompt.py
index 121e57d..3cd2fca 100644
--- a/app/service/chat_robot/script/prompt.py
+++ b/app/service/chat_robot/script/prompt.py
@@ -34,6 +34,39 @@ You may encounter the following types of questions:
 Be careful to use the tools, since you are actually a chat bot. Tools can only be used when essential.
 """
 
+FASHION_CHAT_BOT_PREFIX_TEMP = """
+You are a fashion design assistant with the following capabilities:
+1. Direct conversation: Answer general questions (e.g., greetings, opinions).
+2. Tool usage:
+   - `get_image_from_vector_db`: Retrieve clothing items (requires gender parameter).
+   - `internet_search`: Fetch real-time fashion trends.
+   - `tutorial_tool`: Provide styling guides.
+
+Key Rules:
+1. Tool Selection:
+   - Use `get_image_from_vector_db` for clothing queries (e.g., "show men's jackets").
+   - Use `internet_search` for time-sensitive queries (e.g., "2024 Paris Fashion Week trends").
+   - Use `tutorial_tool` for educational requests (e.g., "how to layer outfits").
+
+2. Gender Handling (for `get_image_from_vector_db` only):
+   - Step 1: Check the **current user input** for gender keywords (e.g., "women/men/she/he"). If found, extract and pass as `gender`.
+   - Step 2: If no gender in current input, scan the **chat history** for the most recent gender reference.
+   - Step 3: If undetermined, default to `"unisex"`.
+
+3. Output Format:
+   - Direct replies: Keep responses under 20 words.
+   - Tool calls:
+     - Always include required parameters (e.g., `gender` for `get_image_from_vector_db`).
+     - Auto-fill `gender` using the above rules if unspecified.
+
+Examples:
+1. User: "Find red dresses for women"  
+   → `get_image_from_vector_db(gender="female", query="dress")`  
+2. User: "show men's jackets"  
+   → `get_image_from_vector_db(gender="male", query="outwear")`  
+3. User: "Show casual outfits"  
+   → `get_image_from_vector_db(gender="unisex", query="casual outfits")`"""
+
 TOOL_SELECT_SUFFIX = """
 Prior to proceeding, it is essential to carefully assess the question and select the appropriate tools or approach accordingly. 
 For database-related questions, use SQL tools to identify relevant tables and query their schemas.
diff --git a/app/service/chat_robot/script/service/CallQWen.py b/app/service/chat_robot/script/service/CallQWen.py
index d2f28a0..5ab74ba 100644
--- a/app/service/chat_robot/script/service/CallQWen.py
+++ b/app/service/chat_robot/script/service/CallQWen.py
@@ -9,7 +9,7 @@ from app.core.config import *
 from app.service.chat_robot.script.callbacks.qwen_callback_handler import QWenCallbackHandler
 from app.service.chat_robot.script.database import CustomDatabase
 from app.service.chat_robot.script.prompt import FASHION_CHAT_BOT_PREFIX, TOOLS_FUNCTIONS_SUFFIX, TUTORIAL_TOOL_RETURN, \
-    GET_LANGUAGE_PREFIX
+    GET_LANGUAGE_PREFIX, FASHION_CHAT_BOT_PREFIX_TEMP
 from app.service.search_image_with_text.service import query
 
 get_database_table_description = "Input is an empty string, output is a comma separated list of tables in the database."
@@ -212,14 +212,15 @@ def get_assistant_response(messages):
     return response
 
 
-def call_with_messages(message, gender):
+def call_with_messages(message):
     global tool_info
     user_input = message
     print('\n')
 
     messages = [
         {
-            "content": FASHION_CHAT_BOT_PREFIX,  # 系统message
+            # "content": FASHION_CHAT_BOT_PREFIX,  # 系统message
+            "content": FASHION_CHAT_BOT_PREFIX_TEMP,  # 修改后的系统message
             "role": "system"
         },
         {
@@ -255,7 +256,7 @@ def call_with_messages(message, gender):
             tool_info = {"name": "search_from_internet", "role": "tool"}
             content = json.loads(assistant_output.tool_calls[0]['function']['arguments'])
             message = [
-                {'role': 'assistant', 'content': content['query']}
+                {'role': 'assistant', 'content': content['query'] if "query" in content.keys() else user_input}
             ]
             tool_info['content'] = search_from_internet(message)
             flag = False
@@ -282,6 +283,8 @@ def call_with_messages(message, gender):
             result_content = tool_info['content']
         elif assistant_output.tool_calls[0]['function']['name'] == 'get_image_from_vector_db':
             content = json.loads(assistant_output.tool_calls[0]['function']['arguments'])
+            # todo 从历史对话中获取性别，目前无法获得性别时，默认使用female
+            gender = content['gender'] if "gender" in content.keys() and content['gender'] != 'unisex' else 'female'
             tool_info = {"name": "get_image_from_vector_db", "role": "tool",
                          'content': get_image_from_vector_db(gender, content['parameters']['content'] if "parameters" in content.keys() else content['content'])}
             flag = False