typed path, path handling,

half the plan is done

src/wf_core/local_paths.py

@@ -3,36 +3,42 @@ from __future__ import annotations
 from collections.abc import Iterable, Mapping
 from typing import Any
 
+from wf_core.paths import LocalPath, PathResolutionError
+
 
 class LocalPathError(ValueError):
     """Raised when a node-local dotted path cannot be parsed or resolved."""
 
 
-def split_local_path(path: str) -> list[str]:
-    """Split one dotted node-local path, rejecting empty segments."""
-    if path == ".":
-        return []
-    parts = path.split(".")
-    if not path or any(not part for part in parts):
-        raise LocalPathError(f"invalid local path {path!r}")
-    return parts
+def _coerce_local_path(path: str | LocalPath) -> LocalPath:
+    try:
+        return path if isinstance(path, LocalPath) else LocalPath.parse(path)
+    except PathResolutionError as exc:
+        raise LocalPathError(str(exc)) from exc
 
 
-def get_local_value(payload: Mapping[str, Any], path: str) -> Any:
+def split_local_path(path: str | LocalPath) -> list[str]:
+    """Split one node-local path, accepting the new typed path object."""
+    return list(_coerce_local_path(path).parts)
+
+
+def get_local_value(payload: Mapping[str, Any], path: str | LocalPath) -> Any:
     """Resolve one node-local path from a nested mapping payload."""
-    if path == ".":
+    parsed = _coerce_local_path(path)
+    if not parsed.parts:
         return dict(payload)
     current: Any = payload
-    for part in split_local_path(path):
+    for part in parsed.parts:
         if not isinstance(current, Mapping) or part not in current:
-            raise LocalPathError(f"local path {path!r} could not be resolved")
+            raise LocalPathError(f"local path {str(parsed)!r} could not be resolved")
         current = current[part]
     return current
 
 
-def set_local_value(payload: dict[str, Any], path: str, value: Any) -> None:
+def set_local_value(payload: dict[str, Any], path: str | LocalPath, value: Any) -> None:
     """Write one value into a nested node-local mapping payload."""
-    parts = split_local_path(path)
+    parsed = _coerce_local_path(path)
+    parts = list(parsed.parts)
     if not parts:
         if not isinstance(value, Mapping):
             raise LocalPathError("root local path requires a mapping value")
@@ -43,12 +49,12 @@ def set_local_value(payload: dict[str, Any], path: str, value: Any) -> None:
     for part in parts[:-1]:
         next_value = current.setdefault(part, {})
         if not isinstance(next_value, dict):
-            raise LocalPathError(f"local path {path!r} overlaps an existing value")
+            raise LocalPathError(f"local path {str(parsed)!r} overlaps an existing value")
         current = next_value
     current[parts[-1]] = value
 
 
-def paths_overlap(left: str, right: str) -> bool:
+def paths_overlap(left: str | LocalPath, right: str | LocalPath) -> bool:
     """Return whether two dotted paths overlap by equality or ancestry."""
     left_parts = split_local_path(left)
     right_parts = split_local_path(right)
@@ -56,9 +62,9 @@ def paths_overlap(left: str, right: str) -> bool:
     return left_parts[:shortest] == right_parts[:shortest]
 
 
-def has_overlapping_paths(paths: Iterable[str]) -> bool:
+def has_overlapping_paths(paths: Iterable[str | LocalPath]) -> bool:
     """Return whether any pair of dotted paths overlaps."""
-    seen: list[str] = []
+    seen: list[str | LocalPath] = []
     for path in paths:
         if any(paths_overlap(path, prior) for prior in seen):
             return True

src/wf_core/models/steps.py

@@ -1,10 +1,46 @@
 from __future__ import annotations
 
+from collections.abc import Mapping
 from typing import Annotated, Literal
 
-from pydantic import BaseModel, ConfigDict, Field
+from pydantic import BaseModel, ConfigDict, Field, model_validator
 
 from wf_core.models.conditions import Condition
+from wf_core.paths import GraphSourcePath, LocalPath, StatePath
+
+
+class InputPathBinding(BaseModel):
+    """Map one workflow graph source path into one node-local input path."""
+
+    model_config = ConfigDict(extra="forbid")
+
+    target: LocalPath
+    path: GraphSourcePath
+
+
+class InputValueBinding(BaseModel):
+    """Map one static value into one node-local input path."""
+
+    model_config = ConfigDict(extra="forbid")
+
+    target: LocalPath
+    value: object
+
+
+InputBinding = Annotated[
+    InputPathBinding | InputValueBinding,
+    Field(union_mode="left_to_right"),
+]
+"""Canonical node input binding, distinguished by `path` vs `value` shape."""
+
+
+class OutputBinding(BaseModel):
+    """Map one node-local output path into one workflow state path."""
+
+    model_config = ConfigDict(extra="forbid")
+
+    source: LocalPath
+    target: StatePath
 
 
 class NodeUse(BaseModel):
@@ -14,32 +50,63 @@ class NodeUse(BaseModel):
     type: Literal["node"]
     node: str
     desc: str | None = None
-    in_map: dict[str, str] = Field(
-        default_factory=dict,
-        description=(
-            "Map graph source paths to node-local input paths. Keys are paths "
-            "such as input.text, state.user.name, or context.item; values are "
-            "input fields/paths inside the node payload."
-        ),
-    )
-    input_values: dict[str, object] = Field(
-        default_factory=dict,
-        description=(
-            "Static node-local input values keyed by destination input field/path. "
-            "Use this for graph-defined constants; use in_map only for graph paths."
-        ),
-    )
-    out_map: dict[str, str] = Field(
-        default_factory=dict,
-        description=(
-            "Map node-local output paths to workflow state destinations. Keys "
-            "are output fields/paths inside the node payload; values must be "
-            "state.* destination paths."
-        ),
-    )
+    input: list[InputBinding] = Field(default_factory=list)
+    output: list[OutputBinding] = Field(default_factory=list)
     retry: int | None = Field(default=None, ge=0)
     timeout_seconds: int | None = Field(default=None, gt=0)
 
+    @model_validator(mode="before")
+    @classmethod
+    def _coerce_deprecated_maps(cls, data: object) -> object:
+        """Normalize deprecated map fields into canonical parse-only bindings."""
+        if not isinstance(data, Mapping):
+            return data
+
+        old_fields = ("in_map", "input_values", "out_map")
+        has_canonical = "input" in data or "output" in data
+        present_old_fields = [field for field in old_fields if field in data]
+        if has_canonical and present_old_fields:
+            old_names = ", ".join(present_old_fields)
+            raise ValueError(
+                f"cannot mix canonical input/output with deprecated fields: {old_names}"
+            )
+
+        normalized = dict(data)
+        input_bindings = list(normalized.pop("input", []))
+        output_bindings = list(normalized.pop("output", []))
+
+        input_values = cls._deprecated_mapping(
+            normalized.pop("input_values", {}), field_name="input_values"
+        )
+        in_map = cls._deprecated_mapping(normalized.pop("in_map", {}), field_name="in_map")
+        out_map = cls._deprecated_mapping(
+            normalized.pop("out_map", {}), field_name="out_map"
+        )
+
+        input_bindings.extend(
+            {"target": target, "value": value}
+            for target, value in input_values.items()
+        )
+        input_bindings.extend(
+            {"target": target, "path": path} for path, target in in_map.items()
+        )
+        output_bindings.extend(
+            {"source": source, "target": target} for source, target in out_map.items()
+        )
+
+        normalized["input"] = input_bindings
+        normalized["output"] = output_bindings
+        return normalized
+
+    @staticmethod
+    def _deprecated_mapping(
+        value: object, *, field_name: str
+    ) -> Mapping[object, object]:
+        """Reject malformed deprecated map inputs before calling `.items()`."""
+        if not isinstance(value, Mapping):
+            raise ValueError(f"{field_name} must be a mapping")
+        return value
+
 
 class ConditionNode(BaseModel):
     """Control-flow step that routes through `true` or `false` outcomes."""

src/wf_core/paths.py

@@ -1,22 +1,252 @@
 from __future__ import annotations
 
+from dataclasses import dataclass
+import re
 from collections.abc import Mapping, MutableMapping
-from typing import Any
+from typing import Any, ClassVar, Literal
+
+from pydantic_core import core_schema
 
 
 class PathResolutionError(ValueError):
     pass
 
 
-def split_graph_path(path: str) -> tuple[str, list[str]]:
-    root, *parts = path.split(".")
-    if not root or not parts:
-        raise PathResolutionError(f"invalid path {path!r}")
+SEGMENT_RE = re.compile(r"^[A-Za-z_][A-Za-z0-9_]*$")
+GraphRoot = Literal["input", "state", "context"]
+
+
+def _validate_segment(segment: str, *, path_kind: str) -> str:
+    if not SEGMENT_RE.fullmatch(segment):
+        raise PathResolutionError(f"invalid {path_kind} segment {segment!r}")
+    return segment
+
+
+def _parse_fragments(*fragments: str, path_kind: str) -> tuple[str, ...]:
+    parts: list[str] = []
+    for fragment in fragments:
+        if not fragment:
+            raise PathResolutionError(f"invalid {path_kind} path")
+        fragment_parts = fragment.split(".")
+        if any(not part for part in fragment_parts):
+            raise PathResolutionError(f"invalid {path_kind} path {fragment!r}")
+        parts.extend(
+            _validate_segment(part, path_kind=path_kind) for part in fragment_parts
+        )
+    return tuple(parts)
+
+
+def _json_schema(pattern: str, description: str) -> dict[str, Any]:
+    return {"type": "string", "pattern": pattern, "description": description}
+
+
+@dataclass(frozen=True)
+class LocalPath:
+    """Node-local payload path. The root marker `.` means the whole payload."""
+
+    parts: tuple[str, ...]
+
+    _JSON_PATTERN: ClassVar[str] = (
+        r"^(\.|[A-Za-z_][A-Za-z0-9_]*(\.[A-Za-z_][A-Za-z0-9_]*)*)$"
+    )
+
+    def __post_init__(self) -> None:
+        object.__setattr__(
+            self,
+            "parts",
+            tuple(_validate_segment(part, path_kind="local") for part in self.parts),
+        )
+
+    @classmethod
+    def root(cls) -> LocalPath:
+        return cls(())
+
+    @classmethod
+    def of(cls, *fragments: str) -> LocalPath:
+        if not fragments:
+            return cls.root()
+        return cls(_parse_fragments(*fragments, path_kind="local"))
+
+    @classmethod
+    def parse(cls, raw: str) -> LocalPath:
+        if raw == ".":
+            return cls.root()
+        return cls.of(raw)
+
+    def __str__(self) -> str:
+        return "." if not self.parts else ".".join(self.parts)
+
+    @classmethod
+    def __get_pydantic_core_schema__(
+        cls, _source_type: object, _handler: object
+    ) -> core_schema.CoreSchema:
+        def validate(value: object) -> LocalPath:
+            if isinstance(value, cls):
+                return cls(value.parts)
+            if isinstance(value, str):
+                return cls.parse(value)
+            raise ValueError("expected local path string")
+
+        return core_schema.no_info_plain_validator_function(
+            validate,
+            serialization=core_schema.plain_serializer_function_ser_schema(
+                str, when_used="json"
+            ),
+        )
+
+    @classmethod
+    def __get_pydantic_json_schema__(
+        cls, _core_schema: core_schema.CoreSchema, _handler: object
+    ) -> dict[str, Any]:
+        return _json_schema(
+            cls._JSON_PATTERN,
+            "Node-local dotted path or root marker `.`.",
+        )
+
+
+@dataclass(frozen=True)
+class GraphSourcePath:
+    """Readable workflow graph path rooted at input, state, or context."""
+
+    root: GraphRoot
+    parts: tuple[str, ...] = ()
+
+    _ROOTS: ClassVar[set[str]] = {"input", "state", "context"}
+    _JSON_PATTERN: ClassVar[str] = (
+        r"^(input|state|context)(\.[A-Za-z_][A-Za-z0-9_]*)*$"
+    )
+
+    def __post_init__(self) -> None:
+        if self.root not in self._ROOTS:
+            raise PathResolutionError(f"unknown path root {self.root!r}")
+        object.__setattr__(
+            self,
+            "parts",
+            tuple(
+                _validate_segment(part, path_kind="graph source")
+                for part in self.parts
+            ),
+        )
+
+    @classmethod
+    def parse(cls, raw: str) -> GraphSourcePath:
+        root, *raw_parts = raw.split(".")
+        if root not in cls._ROOTS:
+            raise PathResolutionError(f"unknown path root {root!r}")
+        parts = tuple(
+            _validate_segment(part, path_kind="graph source") for part in raw_parts
+        )
+        return cls(root, parts)  # type: ignore[arg-type]
+
+    @classmethod
+    def input(cls, *fragments: str) -> GraphSourcePath:
+        return cls("input", _parse_fragments(*fragments, path_kind="graph source"))
+
+    @classmethod
+    def state(cls, *fragments: str) -> GraphSourcePath:
+        return cls("state", _parse_fragments(*fragments, path_kind="graph source"))
+
+    @classmethod
+    def context(cls, *fragments: str) -> GraphSourcePath:
+        return cls("context", _parse_fragments(*fragments, path_kind="graph source"))
+
+    def __str__(self) -> str:
+        return self.root if not self.parts else f"{self.root}.{'.'.join(self.parts)}"
+
+    @classmethod
+    def __get_pydantic_core_schema__(
+        cls, _source_type: object, _handler: object
+    ) -> core_schema.CoreSchema:
+        def validate(value: object) -> GraphSourcePath:
+            if isinstance(value, cls):
+                return cls(value.root, value.parts)
+            if isinstance(value, str):
+                return cls.parse(value)
+            raise ValueError("expected graph source path string")
+
+        return core_schema.no_info_plain_validator_function(
+            validate,
+            serialization=core_schema.plain_serializer_function_ser_schema(
+                str, when_used="json"
+            ),
+        )
+
+    @classmethod
+    def __get_pydantic_json_schema__(
+        cls, _core_schema: core_schema.CoreSchema, _handler: object
+    ) -> dict[str, Any]:
+        return _json_schema(
+            cls._JSON_PATTERN,
+            "Readable graph path rooted at input, state, or context.",
+        )
+
+
+@dataclass(frozen=True)
+class StatePath:
+    """Writable workflow state path. Bare `state` is intentionally invalid."""
+
+    parts: tuple[str, ...]
+
+    _JSON_PATTERN: ClassVar[str] = r"^state\.[A-Za-z_][A-Za-z0-9_]*(\.[A-Za-z_][A-Za-z0-9_]*)*$"
+
+    def __post_init__(self) -> None:
+        parts = tuple(_validate_segment(part, path_kind="state") for part in self.parts)
+        if not parts:
+            raise PathResolutionError("expected state path such as state.foo")
+        object.__setattr__(self, "parts", parts)
+
+    @classmethod
+    def of(cls, *fragments: str) -> StatePath:
+        parts = _parse_fragments(*fragments, path_kind="state")
+        if not parts:
+            raise PathResolutionError("expected state path such as state.foo")
+        return cls(parts)
+
+    @classmethod
+    def parse(cls, raw: str) -> StatePath:
+        parsed = GraphSourcePath.parse(raw)
+        if parsed.root != "state" or not parsed.parts:
+            raise PathResolutionError("expected state path such as state.foo")
+        return cls(parsed.parts)
+
+    def __str__(self) -> str:
+        return f"state.{'.'.join(self.parts)}"
+
+    @classmethod
+    def __get_pydantic_core_schema__(
+        cls, _source_type: object, _handler: object
+    ) -> core_schema.CoreSchema:
+        def validate(value: object) -> StatePath:
+            if isinstance(value, cls):
+                return cls(value.parts)
+            if isinstance(value, str):
+                return cls.parse(value)
+            raise ValueError("expected state path string")
+
+        return core_schema.no_info_plain_validator_function(
+            validate,
+            serialization=core_schema.plain_serializer_function_ser_schema(
+                str, when_used="json"
+            ),
+        )
+
+    @classmethod
+    def __get_pydantic_json_schema__(
+        cls, _core_schema: core_schema.CoreSchema, _handler: object
+    ) -> dict[str, Any]:
+        return _json_schema(cls._JSON_PATTERN, "Writable state path such as state.foo.")
+
+
+def split_graph_path(path: str | GraphSourcePath | StatePath) -> tuple[str, list[str]]:
+    if isinstance(path, StatePath):
+        return "state", list(path.parts)
+    parsed = path if isinstance(path, GraphSourcePath) else GraphSourcePath.parse(path)
+    root, parts = parsed.root, list(parsed.parts)
     return root, parts
 
 
 def is_valid_source_path(
-    path: str,
+    path: str | GraphSourcePath,
     state_root_fields: set[str],
     input_root_fields: set[str],
     *,
@@ -27,6 +257,11 @@ def is_valid_source_path(
     except PathResolutionError:
         return False
 
+    if not parts:
+        return root in {"input", "state", "context"} and (
+            root != "context" or allow_context
+        )
+
     field_name = parts[0]
     if allow_context and root == "context":
         return True
@@ -37,16 +272,16 @@ def is_valid_source_path(
     return False
 
 
-def is_valid_destination_path(path: str) -> bool:
+def is_valid_destination_path(path: str | StatePath) -> bool:
     try:
-        root, parts = split_graph_path(path)
+        StatePath.parse(str(path))
     except PathResolutionError:
         return False
-    return root == "state" and bool(parts)
+    return True
 
 
 def resolve_graph_path(
-    path: str,
+    path: str | GraphSourcePath,
     *,
     state: Mapping[str, Any],
     workflow_input: Mapping[str, Any],
@@ -72,7 +307,7 @@ def resolve_graph_path(
 
 
 def path_exists(
-    path: str,
+    path: str | GraphSourcePath,
     *,
     state: Mapping[str, Any],
     workflow_input: Mapping[str, Any],

src/wf_core/runtime/ops/nodes.py

@@ -8,13 +8,13 @@ from wf_core.errors import WorkflowExecutionError
 from wf_core.local_paths import LocalPathError, set_local_value
 from wf_core.models.results import NodeResult
 from wf_core.models.schemas import NodeDef
-from wf_core.models.steps import NodeUse
+from wf_core.models.steps import InputPathBinding, InputValueBinding, NodeUse
 from wf_core.models.workflow import Workflow
 from wf_core.run_state import RunState, RuntimeContext, StepExecutionResult
 from wf_core.runtime.ops.frames import frame_context_values
 from wf_core.runtime.ops.merges import ReducerDefinition
 from wf_core.runtime.ops.schemas import validate_payload_against_schema
-from wf_core.runtime.ops.state import apply_output_map
+from wf_core.runtime.ops.state import apply_output_bindings
 
 NodeHandler = Callable[[dict[str, Any], RuntimeContext], NodeResult | dict[str, Any]]
 AsyncNodeHandler = Callable[
@@ -33,20 +33,22 @@ def _resolve_node_execution(
     frame = run.current_frame()
     context_values = frame_context_values(frame)
     resolved_input: dict[str, Any] = {}
-    for destination_field, value in node.input_values.items():
-        try:
-            set_local_value(resolved_input, destination_field, value)
-        except LocalPathError as exc:
-            raise WorkflowExecutionError(str(exc)) from exc
-    for source_path, destination_field in node.in_map.items():
+    for binding in node.input:
+        if isinstance(binding, InputValueBinding):
+            value = binding.value
+        elif isinstance(binding, InputPathBinding):
             value = safe_resolve_path(
-            source_path,
+                str(binding.path),
                 state=run.state,
                 workflow_input=run.workflow_input,
                 context=context_values,
             )
+        else:
+            raise WorkflowExecutionError(
+                f"unsupported input binding for node {node.id!r}"
+            )
         try:
-            set_local_value(resolved_input, destination_field, value)
+            set_local_value(resolved_input, binding.target, value)
         except LocalPathError as exc:
             raise WorkflowExecutionError(str(exc)) from exc
     validate_payload_against_schema(
@@ -83,9 +85,9 @@ def _finalize_node_execution(
     validate_payload_against_schema(
         node_def.output_schema, result.output, f"node output for {node.id}"
     )
-    state_changes = apply_output_map(
+    state_changes = apply_output_bindings(
         workflow,
-        node,
+        node.output,
         result.output,
         run.state,
         reducers=reducers,

src/wf_core/runtime/ops/state.py

@@ -1,15 +1,17 @@
 from __future__ import annotations
 
-from collections.abc import Mapping
+from collections.abc import Mapping, Sequence
+from copy import deepcopy
 from typing import Any
 
 from wf_core.errors import WorkflowExecutionError
 from wf_core.local_paths import LocalPathError, get_local_value, has_overlapping_paths
 from wf_core.models.reducers import ReducerRef
-from wf_core.models.steps import NodeUse
+from wf_core.models.steps import NodeUse, OutputBinding
 from wf_core.models.workflow import Workflow
 from wf_core.paths import (
     PathResolutionError,
+    StatePath,
     get_nested_value,
     set_nested_value,
     split_graph_path,
@@ -24,14 +26,67 @@ def apply_output_map(
     state: dict[str, Any],
     reducers: Mapping[str, ReducerDefinition] | None = None,
 ) -> dict[str, Any]:
-    return apply_mapped_state(
+    """Compatibility wrapper for callers that still invoke the old helper."""
+    try:
+        return apply_output_bindings(
             workflow,
+            node.output,
             node_output,
-        node.out_map,
             state,
             reducers=reducers,
-        missing_field_message=f"node {node.id!r} did not return required mapped field {{field}}",
+            missing_field_message=(
+                f"node {node.id!r} did not return required mapped field {{field}}"
+            ),
         )
+    except AttributeError as exc:
+        raise WorkflowExecutionError(
+            "apply_output_map requires NodeUse.output canonical bindings"
+        ) from exc
+
+
+def apply_output_bindings(
+    workflow: Workflow,
+    bindings: Sequence[OutputBinding],
+    node_output: dict[str, Any],
+    state: dict[str, Any],
+    *,
+    reducers: Mapping[str, ReducerDefinition] | None = None,
+    missing_field_message: str = "node output did not include required field {field}",
+) -> dict[str, Any]:
+    """Prepare and commit one atomic state patch from canonical output bindings."""
+    if has_overlapping_paths(str(binding.target) for binding in bindings):
+        raise WorkflowExecutionError(
+            "mapped state patch has overlapping destination paths"
+        )
+
+    resolved_patch: dict[StatePath, Any] = {}
+    for binding in bindings:
+        try:
+            value = get_local_value(node_output, binding.source)
+        except LocalPathError:
+            raise WorkflowExecutionError(
+                missing_field_message.format(field=repr(str(binding.source)))
+            ) from None
+        resolved_patch[binding.target] = value
+
+    prepared_patch: dict[StatePath, tuple[list[str], Any]] = {}
+    for destination_path, value in resolved_patch.items():
+        key_path, merged_value = prepare_state_value(
+            workflow,
+            state,
+            destination_path,
+            value,
+            reducers=reducers,
+        )
+        prepared_patch[destination_path] = (key_path, merged_value)
+
+    # Stage writes on a copy so commit-time path errors cannot partially mutate state.
+    staged_state = deepcopy(state)
+    for _destination_path, (key_path, merged_value) in prepared_patch.items():
+        safe_set_nested_value(staged_state, key_path, merged_value)
+    state.clear()
+    state.update(staged_state)
+    return {str(path): value for path, value in resolved_patch.items()}
 
 
 def apply_mapped_state(
@@ -43,30 +98,18 @@ def apply_mapped_state(
     reducers: Mapping[str, ReducerDefinition] | None = None,
     missing_field_message: str,
 ) -> dict[str, Any]:
-    if has_overlapping_paths(mapping.values()):
-        raise WorkflowExecutionError(
-            "mapped state patch has overlapping destination paths"
-        )
-
-    patch: dict[str, Any] = {}
-    for source_field, destination_path in mapping.items():
-        try:
-            value = get_local_value(source_data, source_field)
-        except LocalPathError:
-            raise WorkflowExecutionError(
-                missing_field_message.format(field=repr(source_field))
-            ) from None
-        patch[destination_path] = value
-
-    for destination_path, value in patch.items():
-        write_state_value(
+    bindings = [
+        OutputBinding.model_validate({"source": source, "target": target})
+        for source, target in mapping.items()
+    ]
+    return apply_output_bindings(
         workflow,
+        bindings,
+        source_data,
         state,
-            destination_path,
-            value,
         reducers=reducers,
+        missing_field_message=missing_field_message,
     )
-    return dict(patch)
 
 
 def write_state_value(
@@ -77,6 +120,25 @@ def write_state_value(
     *,
     reducers: Mapping[str, ReducerDefinition] | None = None,
 ) -> None:
+    key_path, merged_value = prepare_state_value(
+        workflow,
+        state,
+        destination_path,
+        value,
+        reducers=reducers,
+    )
+    safe_set_nested_value(state, key_path, merged_value)
+
+
+def prepare_state_value(
+    workflow: Workflow,
+    state: dict[str, Any],
+    destination_path: str | StatePath,
+    value: Any,
+    *,
+    reducers: Mapping[str, ReducerDefinition] | None = None,
+) -> tuple[list[str], Any]:
+    """Resolve reducer output for a state write without mutating state."""
     try:
         root, parts = split_graph_path(destination_path)
     except PathResolutionError as exc:
@@ -98,10 +160,10 @@ def write_state_value(
         reducer=reducer,
         current_value=current_value,
         incoming_value=value,
-        destination_path=destination_path,
+        destination_path=str(destination_path),
         reducers=reducers,
     )
-    safe_set_nested_value(state, key_path, merged_value)
+    return key_path, merged_value
 
 
 def project_output(workflow: Workflow, state: dict[str, Any]) -> dict[str, Any]:

src/wf_core/validation/steps.py

@@ -11,9 +11,21 @@ from wf_core.models.conditions import (
 )
 from wf_core.local_paths import LocalPathError, has_overlapping_paths, split_local_path
 from wf_core.models.schemas import NodeDef
-from wf_core.models.steps import ConditionNode, ForeachNode, InterruptNode, NodeUse
+from wf_core.models.steps import (
+    ConditionNode,
+    ForeachNode,
+    InputPathBinding,
+    InterruptNode,
+    NodeUse,
+)
 from wf_core.models.workflow import Workflow
-from wf_core.paths import is_valid_destination_path, is_valid_source_path
+from wf_core.paths import (
+    LocalPath,
+    PathResolutionError,
+    StatePath,
+    is_valid_destination_path,
+    is_valid_source_path,
+)
 from wf_core.validation.issues import ValidationIssueCode, ValidationReport
 
 
@@ -36,76 +48,66 @@ def validate_node_use(
     input_fields = set(node_def.input_schema.properties)
     output_fields = set(node_def.output_schema.properties)
     state_fields = set(workflow.state_schema.fields)
+    state_root_fields = {field.split(".", maxsplit=1)[0] for field in state_fields}
     input_root_fields = set(workflow.input_schema.properties)
 
-    for destination_field in node.input_values:
-        destination_root = _local_root(destination_field)
+    input_targets = []
+    for input_index, binding in enumerate(node.input):
+        input_targets.append(binding.target)
+        destination_root = _local_root(binding.target)
         if destination_root is None or (
             destination_root != "." and destination_root not in input_fields
         ):
             report.add(
                 ValidationIssueCode.INVALID_NODE_INPUT_FIELD,
-                f"nodes[{index}].input_values[{destination_field!r}]",
-                f"destination field {destination_field!r} is not declared in node input schema",
+                f"nodes[{index}].input[{input_index}].target",
+                f"destination field {str(binding.target)!r} is not declared in node input schema",
             )
 
-    for source_path, destination_field in node.in_map.items():
-        destination_root = _local_root(destination_field)
-        if destination_root is None or (
-            destination_root != "." and destination_root not in input_fields
-        ):
-            report.add(
-                ValidationIssueCode.INVALID_NODE_INPUT_FIELD,
-                f"nodes[{index}].in_map[{source_path!r}]",
-                f"destination field {destination_field!r} is not declared in node input schema",
-            )
-        if not is_valid_source_path(
-            source_path, state_fields, input_root_fields, allow_context=True
+        if isinstance(binding, InputPathBinding) and not is_valid_source_path(
+            binding.path, state_root_fields, input_root_fields, allow_context=True
         ):
             report.add(
                 ValidationIssueCode.INVALID_SOURCE_PATH,
-                f"nodes[{index}].in_map[{source_path!r}]",
+                f"nodes[{index}].input[{input_index}].path",
                 "source path must start with input., state., or context. and reference a declared root field when applicable",
             )
 
-    if has_overlapping_paths(node.in_map.values()):
+    if has_overlapping_paths(input_targets):
         report.add(
             ValidationIssueCode.INVALID_NODE_INPUT_FIELD,
-            f"nodes[{index}].in_map",
-            "in_map has overlapping node-local input paths",
-        )
-    if has_overlapping_paths([*node.input_values, *node.in_map.values()]):
-        report.add(
-            ValidationIssueCode.INVALID_NODE_INPUT_FIELD,
-            f"nodes[{index}].input_values",
-            "static input_values overlap with path-based in_map destinations",
+            f"nodes[{index}].input",
+            "input has overlapping node-local input paths",
         )
 
-    for source_field, destination_path in node.out_map.items():
-        source_root = _local_root(source_field)
+    output_targets = []
+    for output_index, binding in enumerate(node.output):
+        output_targets.append(str(binding.target))
+        source_root = _local_root(binding.source)
         if source_root is None or (
             source_root != "." and source_root not in output_fields
         ):
             report.add(
                 ValidationIssueCode.INVALID_NODE_OUTPUT_FIELD,
-                f"nodes[{index}].out_map[{source_field!r}]",
-                f"source field {source_field!r} is not declared in node output schema",
+                f"nodes[{index}].output[{output_index}].source",
+                f"source field {str(binding.source)!r} is not declared in node output schema",
             )
-        if not is_valid_destination_path(destination_path):
+        destination_root = _state_destination_root(binding.target)
+        if destination_root is None or destination_root not in state_root_fields:
             report.add(
                 ValidationIssueCode.INVALID_DESTINATION_PATH,
-                f"nodes[{index}].out_map[{source_field!r}]",
-                "destination path must start with state.",
+                f"nodes[{index}].output[{output_index}].target",
+                "destination path must start with state. and reference a declared root field",
             )
-    if has_overlapping_paths(node.out_map.values()):
+    if has_overlapping_paths(output_targets):
         report.add(
             ValidationIssueCode.INVALID_DESTINATION_PATH,
-            f"nodes[{index}].out_map",
-            "out_map has overlapping state destination paths",
+            f"nodes[{index}].output",
+            "output has overlapping state destination paths",
         )
 
 
-def _local_root(path: str) -> str | None:
+def _local_root(path: str | LocalPath) -> str | None:
     try:
         parts = split_local_path(path)
     except LocalPathError:
@@ -113,6 +115,13 @@ def _local_root(path: str) -> str | None:
     return "." if not parts else parts[0]
 
 
+def _state_destination_root(path: object) -> str | None:
+    try:
+        return StatePath.parse(str(path)).parts[0]
+    except PathResolutionError:
+        return None
+
+
 def validate_condition_node(
     node: ConditionNode,
     index: int,

tests/core/test_atomic_state_patches.py

@@ -0,0 +1,185 @@
+from __future__ import annotations
+
+import pytest
+
+from wf_core import (
+    END,
+    Edge,
+    NodeDef,
+    NodeUse,
+    ReducerRef,
+    SchemaRef,
+    StateField,
+    StateSchema,
+    Workflow,
+    WorkflowExecutionError,
+)
+from wf_core.models.steps import OutputBinding
+from wf_core.runtime.engine import resume_workflow
+from wf_core.runtime.ops.runs import create_run_state
+from wf_core.runtime.ops.state import apply_output_bindings
+
+
+def test_output_bindings_commit_patch_atomically_when_source_is_missing() -> None:
+    workflow = _workflow()
+    state = {"person": {"name": "old"}}
+
+    with pytest.raises(WorkflowExecutionError, match="missing"):
+        apply_output_bindings(
+            workflow,
+            [
+                _binding("person.name", "state.person.name"),
+                _binding("missing", "state.person.extra"),
+            ],
+            {"person": {"name": "new"}},
+            state,
+        )
+
+    assert state["person"]["name"] == "old"
+    assert "extra" not in state["person"]
+
+
+def test_output_bindings_reject_overlapping_write_targets_before_mutation() -> None:
+    workflow = _workflow()
+    state = {"person": {"name": "old"}}
+
+    with pytest.raises(WorkflowExecutionError, match="overlapping"):
+        apply_output_bindings(
+            workflow,
+            [
+                _binding("person", "state.person"),
+                _binding("person.name", "state.person.name"),
+            ],
+            {"person": {"name": "Ada"}},
+            state,
+        )
+
+    assert state["person"]["name"] == "old"
+
+
+def test_output_bindings_prepare_reducer_results_before_mutation() -> None:
+    workflow = _workflow(
+        fields={
+            "person.name": StateField(type="string"),
+            "person.tags": StateField(
+                type="array",
+                reducer=ReducerRef(name="wf.std.set_union", config={"bad": True}),
+            ),
+        }
+    )
+    state = {"person": {"name": "old", "tags": ["seed"]}}
+
+    with pytest.raises(WorkflowExecutionError, match="reducer config"):
+        apply_output_bindings(
+            workflow,
+            [
+                _binding("person.name", "state.person.name"),
+                _binding("person.tags", "state.person.tags"),
+            ],
+            {"person": {"name": "new", "tags": ["next"]}},
+            state,
+        )
+
+    assert state["person"]["name"] == "old"
+    assert state["person"]["tags"][0] == "seed"
+    assert len(state["person"]["tags"]) == 1
+
+
+def test_output_bindings_commit_to_staged_state_before_mutating_original() -> None:
+    workflow = _workflow(
+        fields={
+            "person.name": StateField(type="string"),
+            "blocked.child": StateField(type="string"),
+        }
+    )
+    state = {"person": {"name": "old"}, "blocked": "not-an-object"}
+
+    with pytest.raises(WorkflowExecutionError, match="cannot descend"):
+        apply_output_bindings(
+            workflow,
+            [
+                _binding("person.name", "state.person.name"),
+                _binding("blocked.child", "state.blocked.child"),
+            ],
+            {"person": {"name": "new"}, "blocked": {"child": "value"}},
+            state,
+        )
+
+    assert state["person"]["name"] == "old"
+    assert state["blocked"] == "not-an-object"
+
+
+def test_full_workflow_execution_writes_canonical_output_bindings() -> None:
+    workflow = _workflow_with_node()
+    run = create_run_state(workflow, {})
+
+    run = resume_workflow(
+        workflow,
+        run,
+        {
+            "rename": lambda _payload, _ctx: {
+                "outcome": "ok",
+                "output": {"person": {"name": "Ada"}},
+            }
+        },
+    )
+
+    assert run.state["person"]["name"] == "Ada"
+    assert run.trace[0].state_changes["state.person.name"] == "Ada"
+
+
+def _binding(source: str, target: str) -> OutputBinding:
+    return OutputBinding.model_validate({"source": source, "target": target})
+
+
+def _workflow(
+    fields: dict[str, StateField] | None = None,
+) -> Workflow:
+    return Workflow(
+        name="patch",
+        input_schema=SchemaRef(type="object", properties={}),
+        state_schema=StateSchema(
+            fields=fields
+            or {
+                "person": StateField(type="object"),
+                "person.name": StateField(type="string"),
+                "person.extra": StateField(type="string"),
+            }
+        ),
+        output_schema=SchemaRef(type="object", properties={}),
+        start="n",
+        nodes=[],
+        edges=[],
+    )
+
+
+def _workflow_with_node() -> Workflow:
+    return Workflow(
+        name="canonical_output",
+        input_schema=SchemaRef(type="object", properties={}),
+        state_schema=StateSchema(fields={"person.name": StateField(type="string")}),
+        output_schema=SchemaRef(type="object", properties={"person": {"type": "object"}}),
+        node_defs=[
+            NodeDef(
+                name="rename",
+                input_schema=SchemaRef(type="object", properties={}),
+                output_schema=SchemaRef(
+                    type="object",
+                    properties={"person": {"type": "object"}},
+                ),
+                outcomes=["ok"],
+            )
+        ],
+        start="rename",
+        nodes=[
+            NodeUse.model_validate(
+                {
+                    "id": "rename",
+                    "type": "node",
+                    "node": "rename",
+                    "output": [{"source": "person.name", "target": "state.person.name"}],
+                }
+            )
+        ],
+        edges=[Edge.model_validate({"from": "rename", "outcome": "ok", "to": END})],
+    )

tests/core/test_canonical_node_bindings.py

@@ -0,0 +1,166 @@
+import pytest
+from pydantic import ValidationError
+
+from wf_core.models.steps import InputPathBinding, InputValueBinding, NodeUse
+from wf_core.paths import GraphSourcePath, LocalPath, StatePath
+
+
+def test_node_use_accepts_canonical_input_and_output_bindings():
+    node = NodeUse.model_validate(
+        {
+            "id": "echo",
+            "type": "node",
+            "node": "echo",
+            "input": [
+                {"target": "message", "path": "input.message"},
+                {"target": "mode", "value": None},
+            ],
+            "output": [{"source": "echoed", "target": "state.echoed"}],
+        }
+    )
+
+    path_binding = node.input[0]
+    assert isinstance(path_binding, InputPathBinding)
+    assert path_binding.target == LocalPath.of("message")
+    assert path_binding.path == GraphSourcePath.input("message")
+
+    value_binding = node.input[1]
+    assert isinstance(value_binding, InputValueBinding)
+    assert value_binding.target == LocalPath.of("mode")
+    assert value_binding.value is None
+
+    assert node.output[0].source == LocalPath.of("echoed")
+    assert node.output[0].target == StatePath.of("echoed")
+
+
+def test_node_use_converts_old_maps_to_canonical_bindings():
+    node = NodeUse.model_validate(
+        {
+            "id": "echo",
+            "type": "node",
+            "node": "echo",
+            "in_map": {"input.message": "message"},
+            "input_values": {"mode": "fast"},
+            "out_map": {"echoed": "state.echoed"},
+        }
+    )
+
+    dumped = node.model_dump(mode="json")
+    assert "in_map" not in dumped
+    assert "input_values" not in dumped
+    assert "out_map" not in dumped
+    assert dumped["input"][0]["value"] == "fast"
+    assert dumped["input"][0]["target"] == "mode"
+    assert dumped["input"][1]["path"] == "input.message"
+    assert dumped["input"][1]["target"] == "message"
+    assert dumped["output"][0]["source"] == "echoed"
+    assert dumped["output"][0]["target"] == "state.echoed"
+
+
+def test_node_use_rejects_mixed_old_and_new_binding_styles():
+    with pytest.raises(ValidationError):
+        NodeUse.model_validate(
+            {
+                "id": "echo",
+                "type": "node",
+                "node": "echo",
+                "input": [{"target": "message", "path": "input.message"}],
+                "in_map": {"input.other": "other"},
+            }
+        )
+
+
+def test_input_binding_rejects_path_and_value_together():
+    with pytest.raises(ValidationError):
+        NodeUse.model_validate(
+            {
+                "id": "bad",
+                "type": "node",
+                "node": "bad",
+                "input": [
+                    {"target": "message", "path": "input.message", "value": "x"}
+                ],
+            }
+        )
+
+
+def test_input_binding_rejects_neither_path_nor_value():
+    with pytest.raises(ValidationError):
+        NodeUse.model_validate(
+            {
+                "id": "bad",
+                "type": "node",
+                "node": "bad",
+                "input": [{"target": "message"}],
+            }
+        )
+
+
+@pytest.mark.parametrize(
+    "field,binding",
+    [
+        ("input", {"target": "message", "path": "input.message", "extra": True}),
+        ("output", {"source": "echoed", "target": "state.echoed", "extra": True}),
+    ],
+)
+def test_bindings_reject_extra_fields(field: str, binding: dict[str, object]):
+    with pytest.raises(ValidationError):
+        NodeUse.model_validate(
+            {"id": "bad", "type": "node", "node": "bad", field: [binding]}
+        )
+
+
+@pytest.mark.parametrize(
+    "field,value",
+    [
+        ("in_map", None),
+        ("input_values", []),
+        ("out_map", "bad"),
+    ],
+)
+def test_deprecated_maps_reject_non_mapping_values(field: str, value: object):
+    with pytest.raises(ValidationError):
+        NodeUse.model_validate(
+            {"id": "bad", "type": "node", "node": "bad", field: value}
+        )
+
+
+def test_deprecated_conversion_preserves_input_value_then_in_map_order():
+    node = NodeUse.model_validate(
+        {
+            "id": "ordered",
+            "type": "node",
+            "node": "ordered",
+            "input_values": {"first": 1, "second": 2},
+            "in_map": {"input.third": "third", "state.fourth": "fourth"},
+        }
+    )
+
+    dumped_input = node.model_dump(mode="json")["input"]
+    assert dumped_input[0]["target"] == "first"
+    assert dumped_input[0]["value"] == 1
+    assert dumped_input[1]["target"] == "second"
+    assert dumped_input[1]["value"] == 2
+    assert dumped_input[2]["target"] == "third"
+    assert dumped_input[2]["path"] == "input.third"
+    assert dumped_input[3]["target"] == "fourth"
+    assert dumped_input[3]["path"] == "state.fourth"
+
+
+def test_deprecated_input_value_preserves_explicit_null():
+    node = NodeUse.model_validate(
+        {
+            "id": "null",
+            "type": "node",
+            "node": "null",
+            "input_values": {"maybe": None},
+        }
+    )
+
+    value_binding = node.input[0]
+    assert isinstance(value_binding, InputValueBinding)
+    assert value_binding.value is None
+
+    dumped_input = node.model_dump(mode="json")["input"]
+    assert dumped_input[0]["target"] == "maybe"
+    assert dumped_input[0]["value"] is None

tests/core/test_mapping_validation.py

@@ -1,6 +1,9 @@
 from __future__ import annotations
 
+from typing import Any, cast
+
 from wf_core import Edge, NodeDef, NodeUse, SchemaRef, StateField, StateSchema, Workflow
+from wf_core.validation.issues import ValidationIssueCode
 
 
 def test_validation_allows_nested_node_local_paths() -> None:
@@ -22,7 +25,10 @@ def test_validation_rejects_overlapping_node_input_destinations() -> None:
     ).validate_structure()
 
     assert any(
-        "overlapping node-local input paths" in issue.message for issue in report.errors
+        issue.code == ValidationIssueCode.INVALID_NODE_INPUT_FIELD
+        and issue.path == "nodes[0].input"
+        and "overlapping node-local input paths" in issue.message
+        for issue in report.errors
     )
 
 
@@ -36,18 +42,146 @@ def test_validation_rejects_overlapping_state_write_destinations() -> None:
     ).validate_structure()
 
     assert any(
-        "overlapping state destination paths" in issue.message
+        issue.code == ValidationIssueCode.INVALID_DESTINATION_PATH
+        and issue.path == "nodes[0].output"
+        and "overlapping state destination paths" in issue.message
         for issue in report.errors
     )
 
 
-def _workflow(*, in_map: dict[str, str], out_map: dict[str, str]) -> Workflow:
+def test_validation_rejects_invalid_canonical_input_source_path() -> None:
+    report = _workflow(
+        input=[{"target": "user.name", "path": "state.unknown.name"}],
+        output=[],
+    ).validate_structure()
+
+    assert any(
+        issue.code == ValidationIssueCode.INVALID_SOURCE_PATH
+        and issue.path == "nodes[0].input[0].path"
+        for issue in report.errors
+    )
+
+
+def test_validation_allows_canonical_input_source_under_declared_state_field_root() -> None:
+    report = _workflow(
+        input=[{"target": "user.name", "path": "state.person.name"}],
+        output=[],
+        state_fields={"person.name": StateField(type="string")},
+    ).validate_structure()
+
+    assert not any(
+        issue.code == ValidationIssueCode.INVALID_SOURCE_PATH for issue in report.errors
+    )
+
+
+def test_validation_rejects_invalid_canonical_output_destination() -> None:
+    workflow = _workflow(
+        input=[],
+        output=[{"source": "user.name", "target": "state.person.name"}],
+    )
+    # StatePath parsing rejects bad roots before workflow validation; mutate here so
+    # validate_node_use still guards malformed canonical destinations.
+    cast(Any, workflow.nodes[0]).output[0].target = "output.person.name"
+    report = workflow.validate_structure()
+
+    assert any(
+        issue.code == ValidationIssueCode.INVALID_DESTINATION_PATH
+        and issue.path == "nodes[0].output[0].target"
+        for issue in report.errors
+    )
+
+
+def test_validation_rejects_undeclared_canonical_output_destination_root() -> None:
+    report = _workflow(
+        input=[],
+        output=[{"source": "user.name", "target": "state.unknown.foo"}],
+    ).validate_structure()
+
+    assert any(
+        issue.code == ValidationIssueCode.INVALID_DESTINATION_PATH
+        and issue.path == "nodes[0].output[0].target"
+        for issue in report.errors
+    )
+
+
+def test_validation_rejects_overlapping_canonical_input_targets() -> None:
+    report = _workflow(
+        input=[
+            {"target": "user", "value": {"name": "Ada"}},
+            {"target": "user.name", "path": "input.person.name"},
+        ],
+        output=[],
+    ).validate_structure()
+
+    assert any(
+        issue.code == ValidationIssueCode.INVALID_NODE_INPUT_FIELD
+        and issue.path == "nodes[0].input"
+        for issue in report.errors
+    )
+
+
+def test_validation_rejects_overlapping_canonical_output_targets() -> None:
+    report = _workflow(
+        input=[],
+        output=[
+            {"source": "user", "target": "state.person"},
+            {"source": "user.name", "target": "state.person.name"},
+        ],
+    ).validate_structure()
+
+    assert any(
+        issue.code == ValidationIssueCode.INVALID_DESTINATION_PATH
+        and issue.path == "nodes[0].output"
+        for issue in report.errors
+    )
+
+
+def test_validation_allows_valid_canonical_mapping() -> None:
+    report = _workflow(
+        input=[
+            {"target": "user.name", "path": "input.person.name"},
+            {"target": "user.nickname", "value": "Ada"},
+        ],
+        output=[{"source": "user.age", "target": "state.person.age"}],
+    ).validate_structure()
+
+    mapping_issue_codes = {
+        ValidationIssueCode.INVALID_NODE_INPUT_FIELD,
+        ValidationIssueCode.INVALID_NODE_OUTPUT_FIELD,
+        ValidationIssueCode.INVALID_SOURCE_PATH,
+        ValidationIssueCode.INVALID_DESTINATION_PATH,
+    }
+    assert not any(issue.code in mapping_issue_codes for issue in report.errors)
+
+
+def _workflow(
+    *,
+    in_map: dict[str, str] | None = None,
+    out_map: dict[str, str] | None = None,
+    input: list[dict[str, object]] | None = None,
+    output: list[dict[str, str]] | None = None,
+    state_fields: dict[str, StateField] | None = None,
+) -> Workflow:
+    node_data: dict[str, object] = {
+        "id": "tool",
+        "type": "node",
+        "node": "tool",
+    }
+    if input is not None or output is not None:
+        node_data["input"] = input or []
+        node_data["output"] = output or []
+    else:
+        node_data["in_map"] = in_map or {}
+        node_data["out_map"] = out_map or {}
+
     return Workflow(
         name="mapping_validation",
         input_schema=SchemaRef.model_validate(
             {"type": "object", "properties": {"person": {"type": "object"}}}
         ),
-        state_schema=StateSchema(fields={"person": StateField(type="object")}),
+        state_schema=StateSchema(
+            fields=state_fields or {"person": StateField(type="object")}
+        ),
         output_schema=SchemaRef(type="object", properties={}),
         node_defs=[
             NodeDef(
@@ -62,14 +196,6 @@ def _workflow(*, in_map: dict[str, str], out_map: dict[str, str]) -> Workflow:
             )
         ],
         start="tool",
-        nodes=[
-            NodeUse(
-                id="tool",
-                type="node",
-                node="tool",
-                in_map=in_map,
-                out_map=out_map,
-            )
-        ],
+        nodes=[NodeUse.model_validate(node_data)],
         edges=[Edge.model_validate({"from": "tool", "outcome": "ok", "to": "__end__"})],
     )

tests/core/test_nested_mappings.py

@@ -1,5 +1,7 @@
 from __future__ import annotations
 
+from typing import Any, cast
+
 import pytest
 
 from wf_core import (
@@ -16,6 +18,72 @@ from wf_core import (
 )
 
 
+def test_canonical_bindings_resolve_input_values_paths_and_explicit_null() -> None:
+    workflow = Workflow.model_validate(
+        {
+            "name": "canonical",
+            "input_schema": {
+                "type": "object",
+                "properties": {"message": {"type": "string"}},
+            },
+            "state_schema": {"fields": {"echoed": {"type": "string"}}},
+            "output_schema": {
+                "type": "object",
+                "properties": {"echoed": {"type": "string"}},
+            },
+            "start": "echo",
+            "node_defs": [
+                {
+                    "name": "echo",
+                    "input_schema": {
+                        "type": "object",
+                        "properties": {
+                            "message": {"type": "string"},
+                            "mode": {"type": "string"},
+                            "maybe": {"type": "null"},
+                        },
+                        "required": ["message", "mode", "maybe"],
+                    },
+                    "output_schema": {
+                        "type": "object",
+                        "properties": {"echoed": {"type": "string"}},
+                    },
+                    "outcomes": ["ok"],
+                }
+            ],
+            "nodes": [
+                {
+                    "id": "echo",
+                    "type": "node",
+                    "node": "echo",
+                    "input": [
+                        {"target": "message", "path": "input.message"},
+                        {"target": "mode", "value": "fast"},
+                        {"target": "maybe", "value": None},
+                    ],
+                    "output": [{"source": "echoed", "target": "state.echoed"}],
+                }
+            ],
+            "edges": [{"from": "echo", "outcome": "ok", "to": END}],
+        }
+    )
+    run = execute_workflow(
+        workflow,
+        {"message": "hi"},
+        registry={
+            "echo": lambda payload, _ctx: {
+                "outcome": "ok",
+                "output": {"echoed": payload["message"]},
+            }
+        },
+    )
+
+    assert run.trace[0].resolved_input["message"] == "hi"
+    assert run.trace[0].resolved_input["mode"] == "fast"
+    assert run.trace[0].resolved_input["maybe"] is None
+    assert run.state["echoed"] == "hi"
+
+
 def test_nested_node_local_paths_build_input_and_read_output() -> None:
     workflow = _nested_mapping_workflow()
 
@@ -33,9 +101,8 @@ def test_nested_node_local_paths_build_input_and_read_output() -> None:
         },
     )
 
-    assert run.trace[0].resolved_input == {
-        "user": {"name": "Ada", "email": "[email protected]"}
-    }
+    assert run.trace[0].resolved_input["user"]["name"] == "Ada"
+    assert run.trace[0].resolved_input["user"]["email"] == "[email protected]"
     assert run.state["person"]["age"] == 36
     assert run.state["person"]["gender"] == "x"
     assert run.state["experience"]["years"] == 12
@@ -46,7 +113,7 @@ def test_missing_nested_node_output_path_fails() -> None:
 
     with pytest.raises(
         WorkflowExecutionError,
-        match="did not return required mapped field 'user.gender'",
+        match="node output did not include required field 'user.gender'",
     ):
         execute_workflow(
             workflow,
@@ -87,12 +154,17 @@ def test_root_node_local_paths_map_whole_input_and_output_payloads() -> None:
         ],
         start="force",
         nodes=[
-            NodeUse(
-                id="force",
-                type="node",
-                node="force_rates",
-                in_map={"input.rates": "."},
-                out_map={".": "state.rates"},
+            cast(
+                Any,
+                NodeUse.model_validate(
+                    {
+                        "id": "force",
+                        "type": "node",
+                        "node": "force_rates",
+                        "in_map": {"input.rates": "."},
+                        "out_map": {".": "state.rates"},
+                    }
+                ),
             )
         ],
         edges=[Edge.model_validate({"from": "force", "outcome": "ok", "to": END})],
@@ -109,8 +181,10 @@ def test_root_node_local_paths_map_whole_input_and_output_payloads() -> None:
         },
     )
 
-    assert run.trace[0].resolved_input == {"r_1": 0.9, "r_10": 0.1}
-    assert run.state["rates"] == {"r_1": 0.0, "r_10": 0.1}
+    assert run.trace[0].resolved_input["r_1"] == 0.9
+    assert run.trace[0].resolved_input["r_10"] == 0.1
+    assert run.state["rates"]["r_1"] == 0.0
+    assert run.state["rates"]["r_10"] == 0.1
 
 
 def test_static_input_values_are_merged_into_node_local_input() -> None:
@@ -141,12 +215,17 @@ def test_static_input_values_are_merged_into_node_local_input() -> None:
         ],
         start="constant",
         nodes=[
-            NodeUse(
-                id="constant",
-                type="node",
-                node="constant",
-                input_values={"value": "CLICKED"},
-                out_map={"value": "state.message"},
+            cast(
+                Any,
+                NodeUse.model_validate(
+                    {
+                        "id": "constant",
+                        "type": "node",
+                        "node": "constant",
+                        "input_values": {"value": "CLICKED"},
+                        "out_map": {"value": "state.message"},
+                    }
+                ),
             )
         ],
         edges=[Edge.model_validate({"from": "constant", "outcome": "ok", "to": END})],
@@ -204,19 +283,24 @@ def _nested_mapping_workflow() -> Workflow:
         ],
         start="big",
         nodes=[
-            NodeUse(
-                id="big",
-                type="node",
-                node="big_tool",
-                in_map={
+            cast(
+                Any,
+                NodeUse.model_validate(
+                    {
+                        "id": "big",
+                        "type": "node",
+                        "node": "big_tool",
+                        "in_map": {
                             "input.person.name": "user.name",
                             "input.digital.email": "user.email",
                         },
-                out_map={
+                        "out_map": {
                             "user.age": "state.person.age",
                             "user.gender": "state.person.gender",
                             "job.years": "state.experience.years",
                         },
+                    }
+                ),
             )
         ],
         edges=[Edge.model_validate({"from": "big", "outcome": "ok", "to": END})],

tests/core/test_path_values.py

@@ -0,0 +1,185 @@
+from __future__ import annotations
+
+import pytest
+from pydantic import BaseModel, ValidationError
+
+from wf_core.paths import (
+    GraphSourcePath,
+    LocalPath,
+    PathResolutionError,
+    StatePath,
+    is_valid_destination_path,
+    is_valid_source_path,
+)
+
+
+def test_graph_source_path_accepts_root_and_nested_paths() -> None:
+    assert str(GraphSourcePath.parse("state")) == "state"
+    assert str(GraphSourcePath.parse("input")) == "input"
+    assert str(GraphSourcePath.parse("context")) == "context"
+    assert str(GraphSourcePath.parse("input.user")) == "input.user"
+    assert str(GraphSourcePath.parse("state.person.name")) == "state.person.name"
+    assert str(GraphSourcePath.context("loop_item")) == "context.loop_item"
+
+
+def test_state_path_serializes_with_state_prefix() -> None:
+    assert str(StatePath.of("person.name")) == "state.person.name"
+    assert str(StatePath.parse("state.person.name")) == "state.person.name"
+
+
+def test_state_path_rejects_bare_state_write_target() -> None:
+    with pytest.raises(PathResolutionError, match="state path"):
+        StatePath.parse("state")
+
+
+def test_local_path_supports_root_marker_and_fragments() -> None:
+    assert str(LocalPath.root()) == "."
+    assert str(LocalPath.of("user.name")) == "user.name"
+    assert str(LocalPath.of("user", "name")) == "user.name"
+    assert LocalPath.parse(".") == LocalPath.root()
+
+
+@pytest.mark.parametrize(
+    "raw",
+    [
+        "",
+        ".",
+        "state.",
+        "state..name",
+        "state.items.0",
+        "state.user-name",
+        "state.items[0]",
+        "output.foo",
+    ],
+)
+def test_graph_source_paths_reject_invalid_segments(raw: str) -> None:
+    with pytest.raises(PathResolutionError):
+        GraphSourcePath.parse(raw)
+
+
+@pytest.mark.parametrize(
+    "factory",
+    [
+        LocalPath.parse,
+        StatePath.parse,
+        GraphSourcePath.parse,
+    ],
+)
+@pytest.mark.parametrize(
+    "raw",
+    [
+        "state.",
+        "state..name",
+        "state.items.0",
+        "state.user-name",
+        "state.items[0]",
+    ],
+)
+def test_all_path_types_reject_invalid_segments(factory, raw: str) -> None:
+    with pytest.raises(PathResolutionError):
+        factory(raw)
+
+
+def test_path_objects_are_immutable_and_hashable() -> None:
+    paths = {StatePath.of("person.name"), StatePath.of("person.name")}
+    assert len(paths) == 1
+
+    with pytest.raises(Exception):
+        StatePath.of("person.name").parts = ("other",)  # type: ignore[misc]
+
+
+@pytest.mark.parametrize(
+    ("factory", "args"),
+    [
+        (GraphSourcePath, ("output", ("user-name",))),
+        (StatePath, (("0",),)),
+        (LocalPath, (("items[0]",),)),
+    ],
+)
+def test_direct_constructors_enforce_path_invariants(factory, args: tuple[object, ...]) -> None:
+    with pytest.raises(PathResolutionError):
+        factory(*args)
+
+
+def test_pydantic_revalidates_existing_path_objects() -> None:
+    class Payload(BaseModel):
+        source: GraphSourcePath
+        target: StatePath
+        local: LocalPath
+
+    # Bypass constructors to simulate stale or malicious objects that predate
+    # constructor validation. Pydantic must not blindly trust existing instances.
+    source = object.__new__(GraphSourcePath)
+    object.__setattr__(source, "root", "output")
+    object.__setattr__(source, "parts", ("user-name",))
+    target = object.__new__(StatePath)
+    object.__setattr__(target, "parts", ("0",))
+    local = object.__new__(LocalPath)
+    object.__setattr__(local, "parts", ("items[0]",))
+
+    with pytest.raises(ValidationError):
+        Payload.model_validate({"source": source, "target": StatePath.of("person"), "local": LocalPath.root()})
+    with pytest.raises(ValidationError):
+        Payload.model_validate({"source": GraphSourcePath.input("user"), "target": target, "local": LocalPath.root()})
+    with pytest.raises(ValidationError):
+        Payload.model_validate({"source": GraphSourcePath.input("user"), "target": StatePath.of("person"), "local": local})
+
+
+def test_pydantic_accepts_path_strings_and_serializes_strings() -> None:
+    class Payload(BaseModel):
+        source: GraphSourcePath
+        target: StatePath
+        local: LocalPath
+
+    payload = Payload.model_validate(
+        {"source": "input.user", "target": "state.person", "local": "user"}
+    )
+
+    assert payload.source == GraphSourcePath.input("user")
+    assert payload.target == StatePath.of("person")
+    assert payload.local == LocalPath.of("user")
+
+    dumped = payload.model_dump(mode="json")
+    assert dumped["source"] == "input.user"
+    assert dumped["target"] == "state.person"
+    assert dumped["local"] == "user"
+
+
+def test_pydantic_accepts_existing_path_objects() -> None:
+    class Payload(BaseModel):
+        source: GraphSourcePath
+        target: StatePath
+        local: LocalPath
+
+    payload = Payload.model_validate(
+        {
+            "source": GraphSourcePath.state("person"),
+            "target": StatePath.of("person.name"),
+            "local": LocalPath.root(),
+        }
+    )
+
+    assert str(payload.source) == "state.person"
+    assert str(payload.target) == "state.person.name"
+    assert str(payload.local) == "."
+
+
+def test_pydantic_rejects_bad_path_string() -> None:
+    class Payload(BaseModel):
+        source: GraphSourcePath
+
+    with pytest.raises(ValidationError):
+        Payload.model_validate({"source": "output.foo"})
+
+
+def test_existing_source_and_destination_validation_helpers_use_new_parsers() -> None:
+    assert is_valid_source_path("state", set(), set()) is True
+    assert is_valid_source_path("input", set(), set()) is True
+    assert is_valid_source_path("context", set(), set(), allow_context=True) is True
+    assert is_valid_source_path("state.person", {"person"}, set()) is True
+    assert is_valid_source_path("input.person", set(), {"person"}) is True
+    assert is_valid_source_path("state.person-name", {"person-name"}, set()) is False
+
+    assert is_valid_destination_path("state") is False
+    assert is_valid_destination_path("state.person") is True
+    assert is_valid_destination_path("input.person") is False