Make OpenAIResponsesModelSettings.openai_builtin_tools work again (#2520)

DouweM · web-flow · commit 3bc8e4359362 · 2025-08-12T12:06:39.000-06:00
diff --git a/pydantic_ai_slim/pydantic_ai/models/anthropic.py b/pydantic_ai_slim/pydantic_ai/models/anthropic.py
@@ -23,6 +23,7 @@
 from .. import ModelHTTPError, UnexpectedModelBehavior, _utils, usage
 from .._run_context import RunContext
 from .._utils import guard_tool_call_id as _guard_tool_call_id
+from ..exceptions import UserError
 from ..messages import (
     BinaryContent,
     BuiltinToolCallPart,
@@ -367,6 +368,10 @@ def _get_builtin_tools(self, model_request_parameters: ModelRequestParameters) -
                 )
             elif isinstance(tool, CodeExecutionTool):  # pragma: no branch
                 tools.append(BetaCodeExecutionTool20250522Param(name='code_execution', type='code_execution_20250522'))
+            else:  # pragma: no cover
+                raise UserError(
+                    f'`{tool.__class__.__name__}` is not supported by `AnthropicModel`. If it should be, please file an issue.'
+                )
         return tools
 
     async def _map_message(self, messages: list[ModelMessage]) -> tuple[str, list[BetaMessageParam]]:  # noqa: C901
diff --git a/pydantic_ai_slim/pydantic_ai/models/google.py b/pydantic_ai_slim/pydantic_ai/models/google.py
@@ -221,6 +221,10 @@ def _get_tools(self, model_request_parameters: ModelRequestParameters) -> list[T
                 tools.append(ToolDict(google_search=GoogleSearchDict()))
             elif isinstance(tool, CodeExecutionTool):  # pragma: no branch
                 tools.append(ToolDict(code_execution=ToolCodeExecutionDict()))
+            else:  # pragma: no cover
+                raise UserError(
+                    f'`{tool.__class__.__name__}` is not supported by `GoogleModel`. If it should be, please file an issue.'
+                )
         return tools or None
 
     def _get_tool_config(
diff --git a/pydantic_ai_slim/pydantic_ai/models/groq.py b/pydantic_ai_slim/pydantic_ai/models/groq.py
@@ -13,7 +13,7 @@
 from .._run_context import RunContext
 from .._thinking_part import split_content_into_text_and_thinking
 from .._utils import generate_tool_call_id, guard_tool_call_id as _guard_tool_call_id, number_to_datetime
-from ..builtin_tools import CodeExecutionTool, WebSearchTool
+from ..builtin_tools import WebSearchTool
 from ..exceptions import UserError
 from ..messages import (
     BinaryContent,
@@ -318,8 +318,10 @@ def _get_builtin_tools(
             if isinstance(tool, WebSearchTool):
                 if not GroqModelProfile.from_profile(self.profile).groq_always_has_web_search_builtin_tool:
                     raise UserError('`WebSearchTool` is not supported by Groq')  # pragma: no cover
-            elif isinstance(tool, CodeExecutionTool):  # pragma: no branch
-                raise UserError('`CodeExecutionTool` is not supported by Groq')
+            else:
+                raise UserError(
+                    f'`{tool.__class__.__name__}` is not supported by `GroqModel`. If it should be, please file an issue.'
+                )
         return tools
 
     def _map_messages(self, messages: list[ModelMessage]) -> list[chat.ChatCompletionMessageParam]:
diff --git a/pydantic_ai_slim/pydantic_ai/models/openai.py b/pydantic_ai_slim/pydantic_ai/models/openai.py
@@ -462,8 +462,10 @@ def _get_web_search_options(self, model_request_parameters: ModelRequestParamete
                         ),
                     )
                 return WebSearchOptions(search_context_size=tool.search_context_size)
-            elif isinstance(tool, CodeExecutionTool):  # pragma: no branch
-                raise UserError('`CodeExecutionTool` is not supported by OpenAI')
+            else:
+                raise UserError(
+                    f'`{tool.__class__.__name__}` is not supported by `OpenAIModel`. If it should be, please file an issue.'
+                )
 
     async def _map_messages(self, messages: list[ModelMessage]) -> list[chat.ChatCompletionMessageParam]:
         """Just maps a `pydantic_ai.Message` to a `openai.types.ChatCompletionMessageParam`."""
@@ -632,14 +634,6 @@ class OpenAIResponsesModel(Model):
     The [OpenAI Responses API](https://platform.openai.com/docs/api-reference/responses) is the
     new API for OpenAI models.
 
-    The Responses API has built-in tools, that you can use instead of building your own:
-
-    - [Web search](https://platform.openai.com/docs/guides/tools-web-search)
-    - [File search](https://platform.openai.com/docs/guides/tools-file-search)
-    - [Computer use](https://platform.openai.com/docs/guides/tools-computer-use)
-
-    Use the `openai_builtin_tools` setting to add these tools to your model.
-
     If you are interested in the differences between the Responses API and the Chat Completions API,
     see the [OpenAI API docs](https://platform.openai.com/docs/guides/responses-vs-chat-completions).
     """
@@ -780,8 +774,11 @@ async def _responses_create(
         model_settings: OpenAIResponsesModelSettings,
         model_request_parameters: ModelRequestParameters,
     ) -> responses.Response | AsyncStream[responses.ResponseStreamEvent]:
-        tools = self._get_tools(model_request_parameters)
-        tools = self._get_builtin_tools(model_request_parameters) + tools
+        tools = (
+            self._get_builtin_tools(model_request_parameters)
+            + list(model_settings.get('openai_builtin_tools', []))
+            + self._get_tools(model_request_parameters)
+        )
 
         if not tools:
             tool_choice: Literal['none', 'required', 'auto'] | None = None
@@ -880,6 +877,10 @@ def _get_builtin_tools(self, model_request_parameters: ModelRequestParameters) -
                 tools.append(web_search_tool)
             elif isinstance(tool, CodeExecutionTool):  # pragma: no branch
                 tools.append({'type': 'code_interpreter', 'container': {'type': 'auto'}})
+            else:
+                raise UserError(  # pragma: no cover
+                    f'`{tool.__class__.__name__}` is not supported by `OpenAIResponsesModel`. If it should be, please file an issue.'
+                )
         return tools
 
     def _map_tool_definition(self, f: ToolDefinition) -> responses.FunctionToolParam:
diff --git a/tests/models/cassettes/test_openai_responses/test_openai_responses_model_builtin_tools.yaml b/tests/models/cassettes/test_openai_responses/test_openai_responses_model_builtin_tools.yaml
@@ -8,7 +8,7 @@ interactions:
       connection:
       - keep-alive
       content-length:
-      - '217'
+      - '199'
       content-type:
       - application/json
       host:
@@ -18,7 +18,6 @@ interactions:
       input:
       - content: Give me the best news about LLMs from the last 24 hours. Be short.
         role: user
-      instructions: ''
       model: gpt-4o
       stream: false
       tool_choice: auto
@@ -32,67 +31,77 @@ interactions:
       connection:
       - keep-alive
       content-length:
-      - '3210'
+      - '2994'
       content-type:
       - application/json
       openai-organization:
       - pydantic-28gund
       openai-processing-ms:
-      - '2843'
+      - '8814'
+      openai-project:
+      - proj_dKobscVY9YJxeEaDJen54e3d
       openai-version:
       - '2020-10-01'
       strict-transport-security:
       - max-age=31536000; includeSubDomains; preload
       transfer-encoding:
       - chunked
     parsed_body:
-      created_at: 1743506361
+      background: false
+      created_at: 1755020432
       error: null
-      id: resp_67ebcbb93728819197f923ff16e98bce04f5055a2a33abc3
+      id: resp_689b7c90010c8196ac0efd68b021490f07450cfc2d48b975
       incomplete_details: null
-      instructions: ''
+      instructions: null
       max_output_tokens: null
+      max_tool_calls: null
       metadata: {}
       model: gpt-4o-2024-08-06
       object: response
       output:
-      - id: ws_67ebcbb9ab4481918bebf63b66dbb67c04f5055a2a33abc3
+      - action:
+          query: latest news on LLMs October 2023
+          type: search
+        id: ws_689b7c90de788196b18c6d68e7894def07450cfc2d48b975
         status: completed
         type: web_search_call
       - content:
         - annotations:
-          - end_index: 571
-            start_index: 404
-            title: OpenAI plans to release open-weight language model in coming months
+          - end_index: 364
+            start_index: 273
+            title: OpenAI's big GPT-5 launch gets bumpy
             type: url_citation
-            url: https://www.reuters.com/technology/artificial-intelligence/openai-plans-release-open-weight-language-model-coming-months-2025-03-31/?utm_source=openai
-          - end_index: 846
-            start_index: 625
-            title: OpenAI plans to release open-weight language model in coming months
+            url: https://www.axios.com/2025/08/12/gpt-5-bumpy-launch-openai?utm_source=openai
+          - end_index: 533
+            start_index: 417
+            title: OpenAI's big GPT-5 launch gets bumpy
             type: url_citation
-            url: https://www.reuters.com/technology/artificial-intelligence/openai-plans-release-open-weight-language-model-coming-months-2025-03-31/?utm_source=openai
-          text: "In the past 24 hours, OpenAI announced plans to release its first open-weight language model with reasoning
-            capabilities since GPT-2. This model will allow developers to fine-tune it for specific applications without needing
-            the original training data. To gather feedback and refine the model, OpenAI will host developer events starting
-            in San Francisco and expanding to Europe and Asia-Pacific regions. ([reuters.com](https://www.reuters.com/technology/artificial-intelligence/openai-plans-release-open-weight-language-model-coming-months-2025-03-31/?utm_source=openai))\n\n\n##
-            OpenAI to Release Open-Weight Language Model:\n- [OpenAI plans to release open-weight language model in coming
-            months](https://www.reuters.com/technology/artificial-intelligence/openai-plans-release-open-weight-language-model-coming-months-2025-03-31/?utm_source=openai) "
+            url: https://www.axios.com/2025/08/12/gpt-5-bumpy-launch-openai?utm_source=openai
+          logprobs: []
+          text: "OpenAI's recent launch of GPT-5 has faced mixed reactions. Despite strong benchmark performance and early
+            praise, users have reported issues like errors in basic math and geography. CEO Sam Altman has acknowledged these
+            concerns and assured that improvements are underway. ([axios.com](https://www.axios.com/2025/08/12/gpt-5-bumpy-launch-openai?utm_source=openai))\n\n\n##
+            OpenAI's GPT-5 Launch Faces Mixed Reactions:\n- [OpenAI's big GPT-5 launch gets bumpy](https://www.axios.com/2025/08/12/gpt-5-bumpy-launch-openai?utm_source=openai) "
           type: output_text
-        id: msg_67ebcbbaf988819192d44919020b82e704f5055a2a33abc3
+        id: msg_689b7c951b7481968513d007e75151fd07450cfc2d48b975
         role: assistant
         status: completed
         type: message
       parallel_tool_calls: true
       previous_response_id: null
+      prompt_cache_key: null
       reasoning:
         effort: null
-        generate_summary: null
+        summary: null
+      safety_identifier: null
+      service_tier: default
       status: completed
       store: true
       temperature: 1.0
       text:
         format:
           type: text
+        verbosity: medium
       tool_choice: auto
       tools:
       - search_context_size: medium
@@ -103,16 +112,17 @@ interactions:
           region: null
           timezone: null
           type: approximate
+      top_logprobs: 0
       top_p: 1.0
       truncation: disabled
       usage:
         input_tokens: 320
         input_tokens_details:
           cached_tokens: 0
-        output_tokens: 200
+        output_tokens: 159
         output_tokens_details:
           reasoning_tokens: 0
-        total_tokens: 520
+        total_tokens: 479
       user: null
     status:
       code: 200
diff --git a/tests/models/test_openai_responses.py b/tests/models/test_openai_responses.py
@@ -428,23 +428,23 @@ async def test_openai_responses_model_builtin_tools(allow_model_requests: None,
                 parts=[
                     TextPart(
                         content="""\
-In the past 24 hours, OpenAI announced plans to release its first open-weight language model with reasoning capabilities since GPT-2. This model will allow developers to fine-tune it for specific applications without needing the original training data. To gather feedback and refine the model, OpenAI will host developer events starting in San Francisco and expanding to Europe and Asia-Pacific regions. ([reuters.com](https://www.reuters.com/technology/artificial-intelligence/openai-plans-release-open-weight-language-model-coming-months-2025-03-31/?utm_source=openai))
+OpenAI's recent launch of GPT-5 has faced mixed reactions. Despite strong benchmark performance and early praise, users have reported issues like errors in basic math and geography. CEO Sam Altman has acknowledged these concerns and assured that improvements are underway. ([axios.com](https://www.axios.com/2025/08/12/gpt-5-bumpy-launch-openai?utm_source=openai))
 
 
-## OpenAI to Release Open-Weight Language Model:
-- [OpenAI plans to release open-weight language model in coming months](https://www.reuters.com/technology/artificial-intelligence/openai-plans-release-open-weight-language-model-coming-months-2025-03-31/?utm_source=openai) \
+## OpenAI's GPT-5 Launch Faces Mixed Reactions:
+- [OpenAI's big GPT-5 launch gets bumpy](https://www.axios.com/2025/08/12/gpt-5-bumpy-launch-openai?utm_source=openai) \
 """
                     )
                 ],
                 usage=Usage(
                     request_tokens=320,
-                    response_tokens=200,
-                    total_tokens=520,
+                    response_tokens=159,
+                    total_tokens=479,
                     details={'reasoning_tokens': 0, 'cached_tokens': 0},
                 ),
                 model_name='gpt-4o-2024-08-06',
                 timestamp=IsDatetime(),
-                vendor_id='resp_67ebcbb93728819197f923ff16e98bce04f5055a2a33abc3',
+                vendor_id='resp_689b7c90010c8196ac0efd68b021490f07450cfc2d48b975',
             ),
         ]
     )
