Upsert secrets to the Arcade Engine. Args: engine_url: The base URL of the Arcade Engine secrets: Set of secret keys to upse

Wait for HTTP server to become healthy. Args: port: Server port timeout: Maximum time to wait in seconds process: Op

Convert WireTypeInfo to ValueSchema, including nested properties.

Write capture results to a JSON file.

Async context manager entry.

Async context manager exit.

Return a complete string representation of the rubric configuration.

Union a field's type with "null" so a null value validates. A nested field can be required yet nullable (``str | None`` with no default): it must

Apply middleware chain to a request.

Normalize a sampling message to a plain dict. ``Sampling.create_message`` hands this layer ``SamplingMessage`` pydantic objects (and/

Recursively: every object subschema is closed and lists ALL its properties as required.

Extract and validate Bearer token from Authorization header. Args: request: Starlette request object Returns:

Build the adapter chain for error handling. Args: adapters: User-provided list of error adapters auth_provider: The auth pro

Build the ``_meta.arcade`` payload that accompanies a tool error ``CallToolResult``. Schema (camelCase on the wire, matching ``examples/m

Build the Arcade MCP gateway URL.

Build the _meta.arcade structure from a tool definition. The structure of _meta.arcade mirrors Arcade format when possible.

Build developer message with additional details from Slack API error.

Build the handler chain for the specific message type.

Build RFC 9728 Protected Resource Metadata Returns: Dictionary containing resource metadata per RFC 9728

Map a ValueSchema to a JSON Schema ``outputSchema``. Per the MCP specification, ``outputSchema.type`` MUST be ``"object"`` because ``structur

Run capture mode with Anthropic client.

Check for missing tool secrets and log warnings. This method is called during server startup to provide early feedback about

Check if client has a capability.

Check if request has correct Content-Type.

Check that the client supports the requested elicitation mode. Servers MUST NOT send elicitation requests with modes that are not sup

Check transport restrictions for tools requiring auth or secrets. Tools requiring authorization or secrets are blocked on unauthenticated HTT

Run cleanup_expired() on a fixed interval until cancelled.

Configure MCP SDK logging to suppress misleading messages.

Shared helper for clients that use the ``mcpServers`` JSON key. Used by Claude Desktop, Windsurf, and Amazon Q which all share the same confi

Convert an MCP tool reference to a NamedExpectedToolCall (NEW in this PR).

Create a CORS preflight response for OPTIONS requests. Returns: Response with 204 status and CORS headers

Create a cost matrix for the assignment problem. Args: actual_tool_calls: A list of tuples of actual tool calls.

Create a mapping of authorization server URLs to their JWKSTokenValidator instances. Args: entries: List of authorization server

Detect legacy (API key) credentials in the credentials file.

Decode and verify the provided JWT token. Uses jwt.decode for signature verification and payload parsing, then performs time-based cl

Get default server instructions.

Display evaluation results to a Rich console.

Helper to build a syntactically-valid cursor for a non-existent task.

Ensure the URL path ends with '/mcp' (without duplicating). Preserves query strings and fragments.

Execute a tool and return a CallToolResult. The caller is responsible for passing a ``tool_context`` that has already cleared scope/t

Execute a tool in the background for task-augmented tools/call. ``tool_context`` is pre-populated by ``_handle_call_tool`` via ``_che

Check if a tool result contains an authorization URL (OAuth required).

Extract client ID from decoded token. Args: decoded: Decoded token claims Returns: Client identifier or "unk

Extract a trivial set of "entities" — emails & capitalised words. Mock.

Return the error message or a fallback.

Extract traceparent and otel request flags from _meta.

Extract (name, args) pairs from CallbackInvocationEvent entries in trace. pctx reports callback ids as "namespace__snake_case_name" (e.g. "tools_

Return the first traceId found in resource_spans, or None.

Extract field descriptions from TypedDict docstrings. TypedDict classes typically have field descriptions as docstrings after each field.

Extract TypeScript code from LLM response (JSON or markdown fallback).

Extract and validate user_id from decoded token. Args: decoded: Decoded token claims Returns: User ID from '

Fetch JWKS with caching. Returns: JWKS dictionary containing public keys Raises: AuthenticationError: If JWK

Find the arcade command (either direct or via uv run). When using uv run from temp dirs (e.g. configure tests), use ``--project`` ins

Locate the LangChain instrumentor's span for the active tool invocation.

Walk a nested structure and return human-readable descriptions of any non-JSON-native keys or values. JSON-native: str, int, float, bool, Non

Find the nearest ancestor directory containing pyproject.toml. This is used as a default boundary for upward directory traversal to prevent a

Find the signing key from JWKS that matches the token's kid. Args: jwks: JSON Web Key Set token: JWT token R

Format comparative evaluation results showing tracks side-by-side.

Format comparative evaluation results with tabbed track view.

Format comparative evaluation results.

Format comparative evaluation results showing tracks side-by-side.

Format multi-model comparative evaluation grouped by case first.

Format multi-model comparative evaluation grouped by case first.

Format multi-model comparative evaluation grouped by case first.

Format multi-model comparative evaluation grouped by case first.

Format single-model comparative evaluation results.

Format single-model comparative evaluation results.

Format single-model comparative evaluation results.

Format single-model comparative evaluation results.

Format evaluation results with color-coded matches and scores. Args: evaluation: An EvaluationResult object containing the evaluatio

Format inner error details into a readable string.

Format multi-model evaluation results with comparison tables.

Format multi-model capture results with track sections.

Format multi-model evaluation results with comparison tables.

Format multi-model capture results with track tabs.

Format multi-model evaluation results with comparison structure.

Format multi-model capture results with track-aware structure.

Format evaluation results with multi-model comparison tables.

Format multi-model capture results with track sections.

Format regular (non-comparative) evaluation results.

Format regular (non-comparative) evaluation results.

Format regular (non-comparative) evaluation results.

Format regular (non-comparative) evaluation results.