Cache compiled output-schema validators on ClientSession (#3134)

Co-authored-by: Max Isbey <224885523+maxisbey@users.noreply.github.com>

Author: jlowin

src/mcp/client/session.py

@@ -1,12 +1,13 @@
 from __future__ import annotations
 
+import json
 import logging
 from collections.abc import Callable, Mapping, Sequence
 from dataclasses import dataclass
 from functools import reduce
 from operator import or_
 from types import TracebackType
-from typing import Annotated, Any, Final, Literal, Protocol, TypeAlias, cast, overload
+from typing import TYPE_CHECKING, Annotated, Any, Final, Literal, Protocol, TypeAlias, cast, overload
 
 import anyio
 import anyio.abc
@@ -56,6 +57,11 @@
 from mcp.shared.subscriptions import SUBSCRIPTION_ID_META_KEY, event_from_wire
 from mcp.shared.transport_context import TransportContext
 
+if TYPE_CHECKING:
+    # `jsonschema` is imported lazily inside `validate_tool_result`: pulling it (and its
+    # `attrs`/`referencing` tree) in at module scope costs every client that never validates.
+    from jsonschema.protocols import Validator
+
 DEFAULT_CLIENT_INFO = types.Implementation(name="mcp", version="0.1.0")
 DISCOVER_TIMEOUT_SECONDS = 10.0
 _NOTIFICATION_QUEUE_SIZE: Final = 256
@@ -70,6 +76,17 @@ def _clamp_inbound_ttl(raw: dict[str, Any]) -> None:
         raw["ttlMs"] = 0
 
 
+def _same_schema(a: dict[str, Any] | None, b: dict[str, Any] | None) -> bool:
+    """JSON equality for two output schemas.
+
+    Python `==` is not JSON equality: it conflates `True`/`1` and `False`/`0`, which JSON
+    Schema keeps distinct (`const: true` vs `const: 1`). Canonical serialization compares as
+    JSON does; where it is stricter (`1` vs `1.0`), erring toward "changed" only costs a
+    recompile, never a stale validator.
+    """
+    return json.dumps(a, sort_keys=True) == json.dumps(b, sort_keys=True)
+
+
 def _preconnect_stamp(data: dict[str, Any], opts: CallOptions) -> None:
     # initialize/discover forbid cancellation; other pre-handshake requests (lowlevel
     # ClientSession callers may skip the handshake entirely) keep the courtesy cancel.
@@ -378,6 +395,9 @@ def __init__(
         self._logging_callback = logging_callback or _default_logging_callback
         self._message_handler = message_handler or _default_message_handler
         self._tool_output_schemas: dict[str, dict[str, Any] | None] = {}
+        # Compiled output-schema validators, derived from `_tool_output_schemas` and owned by
+        # `_absorb_tool_listing`, which evicts a tool's entry whenever its schema changes.
+        self._tool_output_validators: dict[str, Validator] = {}
         self._x_mcp_header_maps: dict[str, dict[tuple[str, ...], str]] = {}
         self._initialize_result: types.InitializeResult | None = None
         self._discover_result: types.DiscoverResult | None = None
@@ -1062,16 +1082,49 @@ async def validate_tool_result(self, name: str, result: types.CallToolResult) ->
             logger.warning(f"Tool {name} not listed by server, cannot validate any structured content")
 
         if output_schema is not None:
-            from jsonschema import SchemaError, ValidationError, validate
+            from jsonschema import exceptions as jsonschema_exceptions
 
             if result.structured_content is None:
                 raise RuntimeError(f"Tool {name} has an output schema but did not return structured content")
-            try:
-                validate(result.structured_content, output_schema)
-            except ValidationError as e:
-                raise RuntimeError(f"Invalid structured content returned by tool {name}: {e}")
-            except SchemaError as e:  # pragma: no cover
-                raise RuntimeError(f"Invalid schema for tool {name}: {e}")  # pragma: no cover
+            validator = self._output_schema_validator(name, output_schema)
+            # `best_match` picks the same error the previous `jsonschema.validate()` call raised,
+            # so the message a caller sees is unchanged. It is untyped upstream.
+            errors = validator.iter_errors(result.structured_content)
+            error = cast(
+                "Exception | None",
+                jsonschema_exceptions.best_match(errors),  # pyright: ignore[reportUnknownMemberType]
+            )
+            if error is not None:
+                raise RuntimeError(f"Invalid structured content returned by tool {name}: {error}") from error
+
+    def _output_schema_validator(self, name: str, output_schema: dict[str, Any]) -> Validator:
+        """Compiled validator for the tool's cached output schema, built once per schema value.
+
+        Compiling is ~60x the cost of validating, so a one-shot `jsonschema.validate()` per
+        result dominates `call_tool`; the compiled validator is cached instead. It stays valid
+        because `_absorb_tool_listing` evicts a tool's validator whenever it absorbs a different
+        schema for that tool, so a cached entry always matches `output_schema`.
+
+        Raises:
+            RuntimeError: The schema is not a valid JSON Schema. Raised on every call, since a
+                failed compile is never cached.
+        """
+        from jsonschema import SchemaError
+        from jsonschema.validators import validator_for
+
+        if (validator := self._tool_output_validators.get(name)) is not None:
+            return validator
+
+        validator_cls = validator_for(output_schema)
+        try:
+            validator_cls.check_schema(output_schema)
+        except SchemaError as e:
+            raise RuntimeError(f"Invalid schema for tool {name}: {e}")
+        # jsonschema ships no `py.typed`, so pyright reads typeshed's stub, which declares
+        # `registry` as required (concrete validators default it); cast to a schema-only ctor.
+        validator = cast("Callable[[dict[str, Any]], Validator]", validator_cls)(output_schema)
+        self._tool_output_validators[name] = validator
+        return validator
 
     async def list_prompts(self, *, params: types.PaginatedRequestParams | None = None) -> types.ListPromptsResult:
         """Send a prompts/list request.
@@ -1200,8 +1253,14 @@ def _absorb_tool_listing(self, result: types.ListToolsResult, *, complete: bool)
                 kept.append(tool)
             result.tools = kept
 
-        # Cache tool output schemas for future validation; cursor pages only ever add.
+        # Cache tool output schemas for future validation; cursor pages only ever add. A
+        # changed schema evicts its compiled validator; an unchanged one (a re-listing, or the
+        # response cache re-absorbing a served hit) keeps it. Only validated tools pay the check.
         for tool in result.tools:
+            if tool.name in self._tool_output_validators and not _same_schema(
+                self._tool_output_schemas.get(tool.name), tool.output_schema
+            ):
+                del self._tool_output_validators[tool.name]
             self._tool_output_schemas[tool.name] = tool.output_schema
 
         if complete:
@@ -1210,6 +1269,7 @@ def _absorb_tool_listing(self, result: types.ListToolsResult, *, complete: bool)
             names = {tool.name for tool in result.tools}
             self._x_mcp_header_maps = {k: v for k, v in self._x_mcp_header_maps.items() if k in names}
             self._tool_output_schemas = {k: v for k, v in self._tool_output_schemas.items() if k in names}
+            self._tool_output_validators = {k: v for k, v in self._tool_output_validators.items() if k in names}
 
         return result
 

tests/client/test_client.py

@@ -657,12 +657,16 @@ async def test_a_complete_listing_prunes_per_tool_state_for_tools_it_no_longer_c
     with anyio.fail_after(5):
         async with Client(server) as client:
             await client.session.list_tools()
+            # Compile the retired tool's output-schema validator so its eviction is observable.
+            await client.session.validate_tool_result("retired", CallToolResult(content=[], structured_content={}))
             assert set(client.session._x_mcp_header_maps) == {"retired", "survivor"}
             assert set(client.session._tool_output_schemas) == {"retired", "survivor"}
+            assert set(client.session._tool_output_validators) == {"retired"}
 
             await client.session.list_tools()
             assert set(client.session._x_mcp_header_maps) == {"survivor"}
             assert set(client.session._tool_output_schemas) == {"survivor"}
+            assert client.session._tool_output_validators == {}
 
 
 async def test_a_complete_listing_prunes_output_schemas_on_a_legacy_session_too() -> None:

tests/client/test_session_promotions.py

@@ -64,3 +64,64 @@ async def test_validate_tool_result_raises_on_schema_mismatch() -> None:
         # Stable SDK prefix only: the message tail is jsonschema text that shifts with the dependency.
         with pytest.raises(RuntimeError, match="Invalid structured content returned by tool t"):
             await client.session.validate_tool_result("t", CallToolResult(content=[], structured_content={"x": "no"}))
+
+
+@pytest.mark.anyio
+async def test_validate_tool_result_raises_on_an_unusable_output_schema() -> None:
+    """A schema that isn't valid JSON Schema is reported as such, on every call."""
+    server = _make_server({"type": "not-a-json-schema-type"})
+    async with Client(server) as client:
+        result = CallToolResult(content=[], structured_content={"x": 1})
+        for _ in range(2):
+            # Compiling is never cached on failure, so the second call raises like the first.
+            # Stable SDK prefix only: the message tail is jsonschema text that shifts with the dependency.
+            with pytest.raises(RuntimeError, match="Invalid schema for tool t"):
+                await client.session.validate_tool_result("t", result)
+
+
+@pytest.mark.anyio
+async def test_validate_tool_result_compiles_the_output_schema_once_per_tool() -> None:
+    """Regression guard: compiling dominates validating, so the validator must outlive one call."""
+    server = _make_server({"type": "object", "properties": {"x": {"type": "integer"}}, "required": ["x"]})
+    async with Client(server) as client:
+        result = CallToolResult(content=[], structured_content={"x": 1})
+        await client.session.validate_tool_result("t", result)
+        compiled = client.session._tool_output_validators["t"]
+        await client.session.validate_tool_result("t", result)
+        assert client.session._tool_output_validators["t"] is compiled
+
+
+@pytest.mark.anyio
+async def test_validate_tool_result_keeps_the_validator_across_a_relisting_of_the_same_schema() -> None:
+    """SDK-defined: an unchanged schema on a re-listing (or a re-absorbed cache hit) keeps its
+    compiled validator, so relisting between calls doesn't reintroduce the per-call compile."""
+    server = _make_server({"type": "object", "properties": {"x": {"type": "integer"}}, "required": ["x"]})
+    async with Client(server) as client:
+        result = CallToolResult(content=[], structured_content={"x": 1})
+        await client.session.validate_tool_result("t", result)
+        compiled = client.session._tool_output_validators["t"]
+
+        await client.session.list_tools()  # a second listing carrying an equal schema
+        await client.session.validate_tool_result("t", result)
+        assert client.session._tool_output_validators["t"] is compiled
+
+
+@pytest.mark.anyio
+async def test_validate_tool_result_recompiles_when_the_server_changes_the_schema() -> None:
+    """A relisted tool must not be validated against the schema it used to declare."""
+    schemas = [
+        {"type": "object", "properties": {"x": {"type": "integer"}}, "required": ["x"]},
+        {"type": "object", "properties": {"x": {"type": "string"}}, "required": ["x"]},
+    ]
+
+    async def on_list_tools(ctx: ServerRequestContext, params: PaginatedRequestParams | None) -> ListToolsResult:
+        return ListToolsResult(tools=[Tool(name="t", input_schema={"type": "object"}, output_schema=schemas.pop(0))])
+
+    server = Server("test-server", on_list_tools=on_list_tools)
+    async with Client(server) as client:
+        integer_result = CallToolResult(content=[], structured_content={"x": 1})
+        await client.session.validate_tool_result("t", integer_result)
+
+        await client.session.list_tools()
+        with pytest.raises(RuntimeError, match="Invalid structured content returned by tool t"):
+            await client.session.validate_tool_result("t", integer_result)