Update test_doc.py

Author: jackalcooper

tests/test_doc.py

@@ -25,9 +25,14 @@
     "<400> InternalError.Algo.InvalidParameter: Temperature should be in [0.0, 2.0]"
 )
 ERR_MSG_PARTIAL_THINKING_CONFLICT = "<400> InternalError.Algo.InvalidParameter: Partial mode is not supported when enable_thinking is true"
-# R1 不支持 enable_thinking 的报错 (注意：表格中该报错包含 Python 字典的字符串表示，需严格匹配引号)
+# R1 不支持 enable_thinking 的报错
 ERR_MSG_R1_THINKING = "Error code: 400 - {'code': 20015, 'message': 'Value error, current model does not support parameter `enable_thinking`.', 'data': None}"
 
+# [UPDATED] tool_choice 校验错误 (根据百炼返回更新)
+ERR_MSG_TOOL_CHOICE = (
+    '<400> InternalError.Algo.InvalidParameter: tool_choice is one of the strings that should be ["none", "auto"]'
+)
+
 # --- HELPERS ---
 
 
@@ -43,7 +48,7 @@ def assert_exact_error(
 ):
     """
     严格校验错误返回：
-    1. HTTP 状态码通常为 4xx 或 500 (根据表格，部分 4xx 业务错误可能返回 200 或 400，此处以解析 body 为主)
+    1. HTTP 状态码通常为 4xx 或 500
     2. JSON body 中的 code 字段
     3. JSON body 中的 message 字段 (逐字匹配)
     """
@@ -52,7 +57,7 @@ def assert_exact_error(
     except Exception:
         pytest.fail(f"Response is not valid JSON: {response.text}")
 
-    # 1. Check Error Code (e.g., 'InvalidParameter' or 'InternalError')
+    # 1. Check Error Code
     actual_code = data.get("code")
     assert (
         actual_code == expected_code_str
@@ -74,7 +79,6 @@ def test_invalid_parameter_type_top_p(self):
         """
         表格行: 4xx的报错请求
         Input: top_p = "a" (string)
-        Expected: InvalidParameter, <400> ... unable to parse string as a number
         """
         payload = {
             "model": "pre-siliconflow/deepseek-v3",
@@ -83,8 +87,6 @@ def test_invalid_parameter_type_top_p(self):
         }
         response = make_request(payload)
 
-        # 根据表格预期，HTTP Code 可能是 400 或 500，但我们主要校验 Body 内容
-        # 表格预期返回: code="InvalidParameter"
         assert_exact_error(
             response,
             expected_code_str="InvalidParameter",
@@ -94,8 +96,6 @@ def test_invalid_parameter_type_top_p(self):
     def test_invalid_parameter_range_top_p(self):
         """
         表格行: pre-siliconflow-deepseek-v3.1 top_p取值范围（0,1.0]
-        Input: top_p = 0
-        Expected: InvalidParameter, Range of top_p should be (0.0, 1.0]
         """
         payload = {
             "model": "pre-siliconflow/deepseek-v3.1",
@@ -113,8 +113,6 @@ def test_invalid_parameter_range_top_p(self):
     def test_invalid_parameter_range_temperature(self):
         """
         表格行: pre-siliconflow-deepseek-v3.1 取值范围 [0, 2)
-        Input: temperature = 2.1
-        Expected: InvalidParameter, Temperature should be in [0.0, 2.0]
         """
         payload = {
             "model": "pre-siliconflow/deepseek-v3.1",
@@ -132,8 +130,6 @@ def test_invalid_parameter_range_temperature(self):
     def test_conflict_prefix_and_thinking(self):
         """
         表格行: 前缀续写...思考模式下...会报4xx
-        Input: partial=True AND enable_thinking=True
-        Expected: InvalidParameter, Partial mode is not supported when enable_thinking is true
         """
         payload = {
             "model": "pre-siliconflow/deepseek-v3.2",
@@ -156,8 +152,6 @@ def test_conflict_prefix_and_thinking(self):
     def test_r1_enable_thinking_unsupported(self):
         """
         表格行: r1传了enable_thinking报错
-        Input: model=deepseek-r1, enable_thinking=True
-        Expected: InternalError, Error code: 400 - {'code': 20015...}
         """
         payload = {
             "model": "pre-siliconflow/deepseek-r1",
@@ -166,37 +160,27 @@ def test_r1_enable_thinking_unsupported(self):
         }
         response = make_request(payload)
 
+        # 注意：此处维持原逻辑，如果需要校验具体错误可放开注释
         assert response.status_code == 200
 
 
 class TestFunctionalFixes:
-    """
-    测试表格中提到的功能修复验证 (Verify Fixes) 和特定的字段格式检查
-    """
 
     def test_r1_usage_structure_no_text_tokens(self):
         """
         表格行: .usage.output_tokens_details 该路径下不应该返回 text_tokens 字段
-        Model: pre-siliconflow/deepseek-r1
-        Check: usage.output_tokens_details 应该只包含 reasoning_tokens
         """
         payload = {
             "model": "pre-siliconflow/deepseek-r1",
             "input": {"messages": [{"role": "user", "content": "你好"}]},
-            "parameters": {"max_tokens": 10},  # 限制输出以加快测试
+            "parameters": {"max_tokens": 10},
         }
         response = make_request(payload)
 
         assert (
             response.status_code == 200
         ), f"Request failed with {response.status_code}"
 
-        # 解析 SSE 流或直接读取 JSON (假设是非流式或读取最后一块)
-        # 这里简化处理：如果是流式，我们需要解析最后一个包含 usage 的块
-        # 为了测试稳定性，建议在此处强制是非流式请求，或者手动解析SSE
-        # 这里演示解析 JSON Body (如果接口支持非流式返回) 或者解析 SSE
-
-        # 简单起见，读取整个流并查找 usage
         content = response.text
         assert (
             '"text_tokens"' not in content
@@ -207,9 +191,10 @@ def test_r1_usage_structure_no_text_tokens(self):
 
     def test_tool_choice_invalid_error_mapping(self):
         """
+        [UPDATED]
         表格行: Error code: 400，_sse_http_status": 500
-        描述: 传递了不符合协议的 tool_choice (传递了type但没有function定义，或格式错误)
-        Expected: 返回明确的 InternalError 且包含 upstream 的 400 信息
+        描述: 传递了不符合协议的 tool_choice (如传递了不支持的对象格式或枚举值)
+        Expected: InvalidParameter (之前是 InternalError), 且包含明确的枚举值提示
         """
         payload = {
             "model": "pre-siliconflow/deepseek-r1",
@@ -220,7 +205,7 @@ def test_tool_choice_invalid_error_mapping(self):
             },
             "parameters": {
                 "result_format": "message",
-                # 错误的 tool_choice 格式，导致上游报错
+                # 错误的 tool_choice 格式 (百炼预期是 string: "none" 或 "auto")
                 "tool_choice": {"type": "get_current_weather"},
                 "tools": [
                     {
@@ -239,22 +224,17 @@ def test_tool_choice_invalid_error_mapping(self):
             },
         }
         response = make_request(payload)
-        data = response.json()
 
-        expected_msg_snippet = "Input should be 'none', 'auto' or 'required'"
-
-        assert data.get("code") == "InternalError"
-        # 验证错误信息是否透传了上游的详细校验失败信息
-        assert expected_msg_snippet in data.get(
-            "message"
-        ), f"Expected error message to contain '{expected_msg_snippet}', got: {data.get('message')}"
+        # 使用 assert_exact_error 统一校验 Code 和 Message
+        assert_exact_error(
+            response,
+            expected_code_str="InvalidParameter",  # 根据百炼返回更新 code
+            expected_message=ERR_MSG_TOOL_CHOICE,  # 使用上方定义的新常量
+        )
 
     def test_history_tool_call_fix(self):
         """
         表格行: 3.1和3.2 message中包含历史tool_call调用信息会报5xx -> 修复验证
-        Model: pre-siliconflow/deepseek-v3.2
-        Input: 包含 system, user, assistant(tool_calls), tool(result) 的完整历史
-        Expected: 200 OK (之前报 500)
         """
         payload = {
             "model": "pre-siliconflow/deepseek-v3.2",
@@ -282,7 +262,6 @@ def test_history_tool_call_fix(self):
                     },
                 ]
             },
-            # 确保不开启思考，避免干扰测试 tool history 功能
             "parameters": {"enable_thinking": False},
         }
 
@@ -294,9 +273,6 @@ def test_history_tool_call_fix(self):
     def test_prefix_completion_success(self):
         """
         表格行: 前缀续写，pre-siliconflow-deepseek-v3.2 报500 -> 修复验证
-        Model: pre-siliconflow/deepseek-v3.2
-        Scenario: Assistant 消息带 partial=True
-        Expected: 200 OK (只要不开启 enable_thinking)
         """
         payload = {
             "model": "pre-siliconflow/deepseek-v3.2",
@@ -306,14 +282,10 @@ def test_prefix_completion_success(self):
                     {"role": "assistant", "partial": True, "content": "你好，我是"},
                 ]
             },
-            # 明确关闭 thinking 以测试单纯的前缀续写功能
             "parameters": {"enable_thinking": False},
         }
 
         response = make_request(payload)
         assert (
             response.status_code == 200
         ), f"Prefix completion failed with {response.status_code}. Body: {response.text}"
-
-        # 可选：验证返回内容确实是以前缀开始的续写
-        # context = response.text...