microsoft/semantic-kernel · error · NotImplementedError
Computer use tool is not implemented yet.
Error message
Computer use tool is not implemented yet.
What it means
Raised by configure_computer_use_tool(), which is a declared static method but unconditionally raises NotImplementedError. The Responses Agent exposes the method signature for API parity, but the computer-use tool is not yet wired up. This is intentional: callers get a clear signal rather than silent behavior.
Source
Thrown at python/semantic_kernel/agents/open_ai/openai_responses_agent.py:720
- The country field is a two-letter ISO country code, like US.
- The timezone field is an IANA timezone like America/Seattle.
Returns:
A WebSearchToolParam dictionary with any passed-in parameters.
"""
tool: WebSearchToolParam = {
"type": "web_search",
}
if context_size is not None:
tool["search_context_size"] = context_size
if user_location is not None:
tool["user_location"] = user_location
return tool
@staticmethod
def configure_computer_use_tool() -> ComputerToolParam:
"""Generate the tool definition for computer use."""
raise NotImplementedError("Computer use tool is not implemented yet.")
@staticmethod
def _generate_structured_output_response_format_schema(name: str, schema: dict) -> dict:
"""Mock function to simulate formatting the final schema with 'strict' = True."""
return {"type": "json_schema", "name": name, "schema": schema, "strict": True}
@staticmethod
def configure_response_format(
response_format: ResponseFormatUnion
| dict[Literal["type"], Literal["text", "json_object"]]
| dict[str, Any]
| type[BaseModel]
| type
| None = None,
) -> dict[str, Any] | None:
"""Form the response format.
{View on GitHub (pinned to c028a0c7dc)
Solutions
- Do not call configure_computer_use_tool(); it is not implemented in this version.
- Remove the computer-use tool from your tool list until the feature ships.
- Check the semantic_kernel changelog/release notes for when computer-use support lands.
- If you need computer use now, call the OpenAI Responses API directly outside the agent abstraction.
Example fix
// before tools = [OpenAIResponsesAgent.configure_computer_use_tool()] // after tools = [] # computer use unavailable; omit it
Defensive patterns
Strategy: type-guard
Validate before calling
# There is no valid call; guard against invoking it at all import semantic_kernel.agents.open_ai.openai_responses_agent as m assert not hasattr(m.OpenAIResponsesAgent, 'configure_computer_use_tool') or True # informational
Type guard
def computer_use_available() -> bool:
return False # not implemented in current version Try / catch
try:
tool = OpenAIResponsesAgent.configure_computer_use_tool()
except NotImplementedError:
tool = None # feature not shipped; skip computer use Prevention
- Do not call configure_computer_use_tool(); it always raises NotImplementedError.
- Gate any computer-use code path behind a feature flag until it ships.
- Track the semantic_kernel release notes for computer-use support.
When it happens
Trigger: Any call to OpenAIResponsesAgent.configure_computer_use_tool(). There is no input that succeeds — the method body is solely `raise NotImplementedError(...)`.
Common situations: Copy-pasting tool-configuration code from an OpenAI Assistants/Responses example that includes computer use, or enumerating available tools programmatically and invoking each configurator. Also hit when following docs that preview the API surface.
Related errors
- This method is not implemented in this test fixture.
- Streaming text contents not implemented.
- Streaming chat message contents not implemented.
- Refresh tokens not supported
- Unable to get channel keys. Channel type not configured.
AI-assisted analysis of microsoft/semantic-kernel@c028a0c7dc (2026-08-13).
Data as JSON: /api/errors/1584c6ace98c3f82.
Report an issue: GitHub.