more map ->list moves: WorkflowDeployment / WorkflowArtifact

docs/superpowers/plans/2026-05-20-typed-source-artifact-contracts.md

@@ -0,0 +1,272 @@
+# Typed Source Artifact Contracts Implementation Plan
+
+> **For agentic workers:** REQUIRED SUB-SKILL: Use superpowers:subagent-driven-development (recommended) or superpowers:executing-plans to implement this plan task-by-task. Steps use checkbox (`- [ ]`) syntax for tracking.
+
+**Goal:** Replace persisted dict-key source/capability contract shapes with explicit list-of-struct models backed by `SourceRef` and `CapabilityRef`.
+
+**Architecture:** Keep dot-joined strings as the JSON wire format for refs, but make Python/Pydantic fields use first-class ref objects. Accept old dict shapes as parse-only compatibility and dump the new list shapes. Runtime code should use helper indexes such as `binding_map()` and `required_capability_map()` instead of depending on serialized dict keys.
+
+**Tech Stack:** Python, Pydantic v2, `wf_platform` refs, `wf_artifacts` models, pytest, ruff, basedpyright.
+
+---
+
+### Task 1: Make Platform Refs Pydantic Boundary Types
+
+**Files:**
+- Modify: `src/wf_platform/refs.py`
+- Modify: `tests/refs/test_platform_refs.py`
+
+- [ ] **Step 1: Add tests for Pydantic validation and serialization**
+
+Add tests that prove:
+
+```python
+class Payload(BaseModel):
+    source: SourceRef
+    capability: CapabilityRef
+
+payload = Payload.model_validate(
+    {"source": "demo.personal", "capability": "demo.personal.echo_tool"}
+)
+
+assert payload.source == SourceRef.parse("demo.personal")
+assert payload.capability == CapabilityRef.parse("demo.personal.echo_tool")
+assert payload.model_dump(mode="json") == {
+    "source": "demo.personal",
+    "capability": "demo.personal.echo_tool",
+}
+```
+
+- [ ] **Step 2: Implement Pydantic core-schema hooks**
+
+Use `__get_pydantic_core_schema__` on `SourceRef` and `CapabilityRef` so both accept existing instances or strings and serialize back to strings.
+
+- [ ] **Step 3: Tighten segment validation modestly**
+
+Reject whitespace-only refs and empty segments. Do not over-restrict valid MCP/source names yet; external systems can use dashes, underscores, and other non-empty string segments.
+
+### Task 2: Convert Deployment Bindings to List-of-Struct
+
+**Files:**
+- Modify: `src/wf_artifacts/models.py`
+- Modify: `tests/artifacts/test_models.py`
+- Modify: `tests/artifacts/test_store.py`
+
+- [ ] **Step 1: Add `SourceBinding`**
+
+```python
+class SourceBinding(BaseModel):
+    logical_source: SourceRef
+    concrete_source: SourceRef
+```
+
+- [ ] **Step 2: Change `WorkflowDeployment.bindings`**
+
+Canonical model field:
+
+```python
+bindings: list[SourceBinding] = Field(default_factory=list)
+```
+
+Parse-only compatibility:
+
+```json
+{"bindings": {"demo": "demo.personal"}}
+```
+
+should normalize to:
+
+```json
+{"bindings": [{"logical_source": "demo", "concrete_source": "demo.personal"}]}
+```
+
+- [ ] **Step 3: Add `binding_map()`**
+
+```python
+def binding_map(self) -> dict[str, str]:
+    return {
+        str(binding.logical_source): str(binding.concrete_source)
+        for binding in self.bindings
+    }
+```
+
+Reject duplicate `logical_source` values during validation.
+
+### Task 3: Convert Required Capabilities to List-of-Struct
+
+**Files:**
+- Modify: `src/wf_artifacts/models.py`
+- Modify: `src/wf_artifacts/factory.py`
+- Modify: `src/wf_artifacts/references.py`
+- Modify: `tests/artifacts/test_models.py`
+- Modify: `tests/artifacts/test_factory.py`
+
+- [ ] **Step 1: Update `RequiredCapability`**
+
+Canonical field:
+
+```python
+ref: CapabilityRef
+```
+
+Keep compatibility properties:
+
+```python
+@property
+def logical_source(self) -> str: ...
+
+@property
+def capability_name(self) -> str: ...
+```
+
+Parse-only compatibility should accept old payloads with `logical_source` and `capability_name` and build `ref`.
+
+- [ ] **Step 2: Change `WorkflowArtifact.required_capabilities`**
+
+Canonical model field:
+
+```python
+required_capabilities: list[RequiredCapability] = Field(default_factory=list)
+```
+
+Parse-only compatibility:
+
+```json
+{
+  "required_capabilities": {
+    "demo.echo_tool": {
+      "logical_source": "demo",
+      "capability_name": "echo_tool",
+      "kind": "node_spec"
+    }
+  }
+}
+```
+
+should normalize to a list and dump as:
+
+```json
+{
+  "required_capabilities": [
+    {
+      "ref": "demo.echo_tool",
+      "kind": "node_spec"
+    }
+  ]
+}
+```
+
+- [ ] **Step 3: Add `required_capability_map()`**
+
+```python
+def required_capability_map(self) -> dict[str, RequiredCapability]:
+    return {str(capability.ref): capability for capability in self.required_capabilities}
+```
+
+Reject duplicate `ref` values during validation.
+
+### Task 4: Update Call Sites to Use Helper Maps
+
+**Files:**
+- Modify: `src/wf_artifacts/validation.py`
+- Modify: `src/wf_artifacts/catalog.py`
+- Modify: `src/wf_mcp/workflow_surface/runtime_dependencies.py`
+- Modify: `src/wf_mcp/workflow_surface/handlers.py`
+- Modify: `src/wf_mcp/workflow_surface/tools.py`
+- Modify: `src/wf_mcp/broker/artifact_tools.py`
+- Modify tests that directly index `.bindings` or `.required_capabilities`.
+
+- [ ] **Step 1: Replace `deployment.bindings.get(...)`**
+
+Use:
+
+```python
+bindings = deployment.binding_map()
+bound_source_id = bindings.get(required.logical_source)
+```
+
+- [ ] **Step 2: Replace `artifact.required_capabilities.items()`**
+
+Use:
+
+```python
+for logical_ref, required in artifact.required_capability_map().items():
+    ...
+```
+
+- [ ] **Step 3: Keep API input compatibility**
+
+MCP tools that accept `required_capabilities` from callers may still accept dict input, but constructed `WorkflowArtifact` should dump the canonical list shape.
+
+### Task 5: Update Docs
+
+**Files:**
+- Modify: `docs/workflow_artifacts.md`
+
+- [ ] **Step 1: Replace dict binding examples**
+
+Use:
+
+```json
+{
+  "bindings": [
+    {"logical_source": "context7", "concrete_source": "context7.default"}
+  ]
+}
+```
+
+- [ ] **Step 2: Replace dict required capability examples**
+
+Use:
+
+```json
+{
+  "required_capabilities": [
+    {
+      "ref": "context7.query-docs",
+      "kind": "node_spec",
+      "input_schema_hash": "sha256:..."
+    }
+  ]
+}
+```
+
+State that old dict shapes are accepted at parse boundaries but not emitted by model dumps.
+
+### Task 6: Verification
+
+**Files:**
+- All touched files
+
+- [ ] **Step 1: Run focused tests**
+
+Run:
+
+```bash
+uv run --with pytest pytest tests/refs/test_platform_refs.py tests/artifacts/test_models.py tests/artifacts/test_store.py tests/artifacts/test_factory.py tests/artifacts/test_validation.py -q
+```
+
+- [ ] **Step 2: Run full suite**
+
+Run:
+
+```bash
+uv run --with pytest pytest -q
+```
+
+- [ ] **Step 3: Run static checks**
+
+Run:
+
+```bash
+uvx ruff check
+uv run basedpyright --level error
+```
+
+---
+
+## Self-Review
+
+- Spec coverage: covers typed refs, deployment binding list shape, required capability list shape, compatibility parsing, helper indexes, docs, and tests.
+- Placeholder scan: no placeholders remain.
+- Type consistency: `SourceRef`, `CapabilityRef`, `SourceBinding`, `RequiredCapability`, `WorkflowArtifact`, and `WorkflowDeployment` names match current code or are introduced in this plan.

docs/workflow_artifacts.md

@@ -221,12 +221,12 @@ step:
   node_ref: context7.query-docs
 
 required_capabilities:
-  context7.query-docs:
+  - ref: context7.query-docs
     kind: tool
-    input_schema_hash
-    input_schema_snapshot
-    output_schema_hash
-    output_schema_snapshot
+    input_schema_hash: ...
+    input_schema_snapshot: ...
+    output_schema_hash: ...
+    output_schema_snapshot: ...
 ```
 
 This is similar to import resolution. The artifact stores stable logical
@@ -286,8 +286,7 @@ actually use:
 
 ```text
 RequiredCapability
-  logical_source: context7
-  capability_name: query-docs
+  ref: context7.query-docs
   kind: tool
   input_schema
   output_schema
@@ -446,8 +445,7 @@ and the artifact should declare a matching required capability:
 
 ```text
 RequiredCapability(
-  logical_source="demo",
-  capability_name="echo_tool",
+  ref="demo.echo_tool",
   kind="node_spec"
 )
 ```
@@ -456,7 +454,8 @@ The deployment then chooses the concrete account or connection profile:
 
 ```text
 bindings:
-  demo: demo.personal
+  - logical_source: demo
+    concrete_source: demo.personal
 ```
 
 This keeps reusable artifacts portable across accounts while still letting
@@ -494,7 +493,8 @@ WorkflowDeployment
   artifact_version: 1
   deployment_id: summarize_docs.context7_default
   bindings:
-    context7: context7.default
+    - logical_source: context7
+      concrete_source: context7.default
   drift_policy: block
 ```
 
@@ -510,12 +510,14 @@ logical sources to different MCP accounts or connection profiles:
 deployment: summarize_docs.personal
 artifact:   summarize_docs@1
 bindings:
-  context7: context7.personal
+  - logical_source: context7
+    concrete_source: context7.personal
 
 deployment: summarize_docs.work
 artifact:   summarize_docs@1
 bindings:
-  context7: context7.work
+  - logical_source: context7
+    concrete_source: context7.work
 ```
 
 This gives the stable MCP run tool a concrete target without requiring one MCP
@@ -599,6 +601,15 @@ Dot-joined names remain the wire/presentation format, but new runtime code
 should parse or format through those refs instead of rediscovering source/name
 boundaries with ad hoc string splits.
 
+Persisted artifact/deployment models use explicit list-of-struct shapes for
+source and capability contracts. Dict-key shapes such as
+`bindings: {"demo": "demo.personal"}` and
+`required_capabilities: {"demo.echo_tool": {...}}` are accepted at parse
+boundaries for compatibility, but model dumps emit the explicit list shapes.
+Runtime code that needs lookup tables should use helper indexes such as
+`WorkflowDeployment.binding_map()` and
+`WorkflowArtifact.required_capability_map()`.
+
 Saved workflow artifact names use a separate grammar and ref type:
 `WorkflowCapabilityRef(artifact_id, version)` serializes as
 `workflow.<artifact_id>.v<version>`. Artifact ids may contain dots, so this

examples/mcp_workflow_surface.py

@@ -207,10 +207,10 @@ async def create_and_run_echo_deployment(root: Path, *, text: str) -> dict[str,
             id="mcp_echo.personal",
             artifact_id="mcp_echo",
             artifact_version=1,
-            bindings={
-                "demo": "demo.personal",
-                "wf.std": "wf.std",
-            },
+            bindings=[
+                {"logical_source": "demo", "concrete_source": "demo.personal"},
+                {"logical_source": "wf.std", "concrete_source": "wf.std"},
+            ],
         )
     )
     return await handlers.run_deployment(

src/wf_artifacts/__init__.py

@@ -28,6 +28,7 @@ from .models import (
     DiagnosticSeverity,
     DriftPolicy,
     RequiredCapability,
+    SourceBinding,
     WorkflowArtifact,
     WorkflowDeployment,
 )
@@ -48,6 +49,7 @@ __all__ = [
     "FileDraftWorkspaceStore",
     "FileWorkflowArtifactStore",
     "RequiredCapability",
+    "SourceBinding",
     "WorkflowArtifact",
     "WorkflowArtifactCatalogEntry",
     "WorkflowCapabilityRef",

src/wf_artifacts/catalog.py

@@ -41,7 +41,7 @@ def artifact_catalog_entry(
     required_sources = sorted(
         {
             capability.logical_source
-            for capability in artifact.required_capabilities.values()
+            for capability in artifact.required_capability_map().values()
         }
     )
     return WorkflowArtifactCatalogEntry(

src/wf_artifacts/factory.py

@@ -30,6 +30,11 @@ def create_workflow_artifact_from_plan(
         observed_node_specs,
     )
     _validate_workflow_plan(normalized_plan)
+    required = {
+        **_required_reducers_from_plan(normalized_plan),
+        **node_requirements,
+        **dict(required_capabilities or {}),
+    }
     return WorkflowArtifact(
         id=artifact_id,
         version=version,
@@ -40,11 +45,7 @@ def create_workflow_artifact_from_plan(
         output_schema=_required_object_field(normalized_plan, "output_schema"),
         outcomes=outcomes,
         plan=normalized_plan,
-        required_capabilities={
-            **_required_reducers_from_plan(normalized_plan),
-            **node_requirements,
-            **dict(required_capabilities or {}),
-        },
+        required_capabilities=list(required.values()),
         created_from_catalog_version=created_from_catalog_version,
     )
 
@@ -103,8 +104,7 @@ def _required_reducers_from_plan(plan: JsonObject) -> dict[str, RequiredCapabili
         except ValueError:
             continue
         requirements[reducer.name] = RequiredCapability(
-            logical_source=str(reducer_ref.source),
-            capability_name=reducer_ref.name,
+            ref=reducer_ref,
             kind="reducer",
         )
     return requirements

src/wf_artifacts/models.py

@@ -3,10 +3,14 @@ from __future__ import annotations
 from enum import StrEnum
 from typing import Any, Literal
 
-from pydantic import BaseModel, Field
+from pydantic import BaseModel, Field, model_validator
+
+from wf_platform import CapabilityRef, SourceRef
 
 JsonObject = dict[str, Any]
 ArtifactKind = Literal["workflow", "wrapper"]
+SourceRefInput = SourceRef | str
+CapabilityRefInput = CapabilityRef | str
 
 
 class DriftPolicy(StrEnum):
@@ -27,16 +31,46 @@ class DiagnosticSeverity(StrEnum):
 class RequiredCapability(BaseModel):
     """Saved contract for one capability an artifact references."""
 
-    logical_source: str
-    capability_name: str
+    ref: CapabilityRefInput
     kind: Literal["tool", "resource", "prompt", "node_spec", "reducer", "workflow"]
     input_schema_hash: str | None = None
     input_schema_snapshot: JsonObject | None = None
     output_schema_hash: str | None = None
     output_schema_snapshot: JsonObject | None = None
-    observed_concrete_source: str | None = None
+    observed_concrete_source: SourceRefInput | None = None
     observed_at_epoch_ms: int | None = Field(default=None, ge=0)
 
+    @property
+    def logical_source(self) -> str:
+        """Compatibility accessor for callers migrating to `ref`."""
+        return str(self.capability_ref().source)
+
+    @property
+    def capability_name(self) -> str:
+        """Compatibility accessor for callers migrating to `ref`."""
+        return self.capability_ref().name
+
+    def capability_ref(self) -> CapabilityRef:
+        """Return the typed ref even when constructed from JSON-compatible input."""
+        return (
+            self.ref
+            if isinstance(self.ref, CapabilityRef)
+            else CapabilityRef.parse(self.ref)
+        )
+
+    @model_validator(mode="before")
+    @classmethod
+    def _coerce_legacy_ref_fields(cls, value: object) -> object:
+        if not isinstance(value, dict):
+            return value
+        data = dict(value)
+        if "ref" not in data:
+            logical_source = data.pop("logical_source", None)
+            capability_name = data.pop("capability_name", None)
+            if isinstance(logical_source, str) and isinstance(capability_name, str):
+                data["ref"] = f"{logical_source}.{capability_name}"
+        return data
+
 
 class AvailableCapability(BaseModel):
     """Current contract for one capability exposed by a bound source."""
@@ -55,6 +89,29 @@ class AvailableSource(BaseModel):
     capabilities: dict[str, AvailableCapability] = Field(default_factory=dict)
 
 
+class SourceBinding(BaseModel):
+    """Deployment-time mapping from artifact logical source to concrete source."""
+
+    logical_source: SourceRefInput
+    concrete_source: SourceRefInput
+
+    def logical_ref(self) -> SourceRef:
+        """Return the typed logical source ref for runtime lookup code."""
+        return (
+            self.logical_source
+            if isinstance(self.logical_source, SourceRef)
+            else SourceRef.parse(self.logical_source)
+        )
+
+    def concrete_ref(self) -> SourceRef:
+        """Return the typed concrete source ref for runtime lookup code."""
+        return (
+            self.concrete_source
+            if isinstance(self.concrete_source, SourceRef)
+            else SourceRef.parse(self.concrete_source)
+        )
+
+
 class DependencyDiagnostic(BaseModel):
     """Machine-readable reason a deployment is degraded or unrunnable."""
 
@@ -78,10 +135,66 @@ class WorkflowArtifact(BaseModel):
     output_schema: JsonObject
     outcomes: tuple[str, ...]
     plan: JsonObject
-    required_capabilities: dict[str, RequiredCapability] = Field(default_factory=dict)
+    required_capabilities: (
+        list[RequiredCapability] | dict[str, RequiredCapability | JsonObject]
+    ) = Field(default_factory=list)
     workflow_dependencies: dict[str, int] = Field(default_factory=dict)
     created_from_catalog_version: str | None = None
 
+    def required_capability_map(self) -> dict[str, RequiredCapability]:
+        """Return required capabilities keyed by dot-joined capability ref."""
+        return {
+            str(capability.capability_ref()): capability
+            for capability in self._required_capability_list()
+        }
+
+    @model_validator(mode="before")
+    @classmethod
+    def _coerce_legacy_required_capabilities(cls, value: object) -> object:
+        if not isinstance(value, dict):
+            return value
+        data = dict(value)
+        required = data.get("required_capabilities")
+        if not isinstance(required, dict):
+            return data
+
+        normalized: list[object] = []
+        for raw_ref, raw_capability in required.items():
+            if not isinstance(raw_capability, dict):
+                normalized.append(raw_capability)
+                continue
+            capability_data = dict(raw_capability)
+            capability_data.setdefault("ref", str(raw_ref))
+            normalized.append(capability_data)
+        data["required_capabilities"] = normalized
+        return data
+
+    @model_validator(mode="after")
+    def _reject_duplicate_required_capabilities(self) -> WorkflowArtifact:
+        self.required_capabilities = self._required_capability_list()
+        refs = [
+            str(capability.capability_ref())
+            for capability in self.required_capabilities
+        ]
+        duplicates = {ref for ref in refs if refs.count(ref) > 1}
+        if duplicates:
+            raise ValueError(
+                "duplicate required capability refs: " + ", ".join(sorted(duplicates))
+            )
+        return self
+
+    def _required_capability_list(self) -> list[RequiredCapability]:
+        if isinstance(self.required_capabilities, dict):
+            return [
+                RequiredCapability.model_validate(
+                    {"ref": ref, **capability}
+                    if isinstance(capability, dict)
+                    else capability
+                )
+                for ref, capability in self.required_capabilities.items()
+            ]
+        return self.required_capabilities
+
 
 class WorkflowDeployment(BaseModel):
     """One configured way to run an artifact version in an environment."""
@@ -89,5 +202,57 @@ class WorkflowDeployment(BaseModel):
     id: str
     artifact_id: str
     artifact_version: int = Field(ge=1)
-    bindings: dict[str, str] = Field(default_factory=dict)
+    bindings: list[SourceBinding | dict[str, str]] | dict[str, str] = Field(
+        default_factory=list
+    )
     drift_policy: DriftPolicy = DriftPolicy.BLOCK
+
+    def binding_map(self) -> dict[str, str]:
+        """Return bindings keyed by dot-joined logical source ref."""
+        return {
+            str(binding.logical_ref()): str(binding.concrete_ref())
+            for binding in self._binding_list()
+        }
+
+    @model_validator(mode="before")
+    @classmethod
+    def _coerce_legacy_binding_map(cls, value: object) -> object:
+        if not isinstance(value, dict):
+            return value
+        data = dict(value)
+        bindings = data.get("bindings")
+        if not isinstance(bindings, dict):
+            return data
+
+        data["bindings"] = [
+            {"logical_source": logical, "concrete_source": concrete}
+            for logical, concrete in bindings.items()
+        ]
+        return data
+
+    @model_validator(mode="after")
+    def _reject_duplicate_bindings(self) -> WorkflowDeployment:
+        normalized = self._binding_list()
+        self.bindings = [*normalized]
+        logical_sources = [str(binding.logical_ref()) for binding in normalized]
+        duplicates = {
+            source for source in logical_sources if logical_sources.count(source) > 1
+        }
+        if duplicates:
+            raise ValueError(
+                "duplicate deployment source bindings: " + ", ".join(sorted(duplicates))
+            )
+        return self
+
+    def _binding_list(self) -> list[SourceBinding]:
+        if isinstance(self.bindings, dict):
+            return [
+                SourceBinding(logical_source=logical, concrete_source=concrete)
+                for logical, concrete in self.bindings.items()
+            ]
+        return [
+            binding
+            if isinstance(binding, SourceBinding)
+            else SourceBinding.model_validate(binding)
+            for binding in self.bindings
+        ]

src/wf_artifacts/references.py

@@ -46,8 +46,7 @@ def normalize_plan_node_refs(
             else None
         )
         requirements[logical_ref] = RequiredCapability(
-            logical_source=logical_source,
-            capability_name=capability_name,
+            ref=CapabilityRef.parse(logical_ref),
             kind="node_spec",
             input_schema_hash=(
                 hash_json_schema(observed.input_schema)
@@ -65,7 +64,7 @@ def normalize_plan_node_refs(
             output_schema_snapshot=(
                 observed.output_schema if observed is not None else None
             ),
-            observed_concrete_source=concrete_source,
+            observed_concrete_source=SourceRef.parse(concrete_source),
         )
 
     return normalized, requirements

src/wf_artifacts/validation.py

@@ -19,10 +19,11 @@ def validate_deployment_dependencies(
 ) -> list[DependencyDiagnostic]:
     """Validate that a deployment can satisfy an artifact's required contracts."""
     sources_by_id = {source.id: source for source in sources}
+    bindings = deployment.binding_map()
     diagnostics: list[DependencyDiagnostic] = []
 
-    for logical_ref, required in artifact.required_capabilities.items():
-        bound_source_id = deployment.bindings.get(required.logical_source)
+    for logical_ref, required in artifact.required_capability_map().items():
+        bound_source_id = bindings.get(required.logical_source)
         if bound_source_id is None:
             diagnostics.append(
                 _diagnostic(

src/wf_mcp/workflow_surface/handlers.py

@@ -234,10 +234,9 @@ class WorkflowSurfaceHandlers:
             "is_async": True,
             "input_schema": artifact.input_schema,
             "output_schema": artifact.output_schema,
-            "required_capabilities": {
-                name: capability.model_dump(mode="json")
-                for name, capability in sorted(artifact.required_capabilities.items())
-            },
+            "required_capabilities": _required_capability_payloads(
+                artifact.required_capability_map()
+            ),
         }
 
     async def _call_wrapper_artifact(
@@ -428,7 +427,7 @@ class WorkflowSurfaceHandlers:
         required_sources = sorted(
             {
                 capability.logical_source
-                for capability in dict(workflow_artifact.required_capabilities).values()
+                for capability in workflow_artifact.required_capability_map().values()
             }
         )
         return {
@@ -892,7 +891,7 @@ def _required_capabilities_for_plan(
         source_bindings=source_bindings,
         observed_node_specs=_observed_node_specs(service),
     )
-    requirements = dict(artifact.required_capabilities)
+    requirements = artifact.required_capability_map()
     for node in _plan_nodes(artifact):
         raw_ref = node.get("node")
         if not isinstance(raw_ref, str) or raw_ref in requirements:
@@ -902,8 +901,7 @@ def _required_capabilities_for_plan(
         except ValueError:
             continue
         requirements[raw_ref] = RequiredCapability(
-            logical_source=str(parsed.source),
-            capability_name=parsed.name,
+            ref=parsed,
             kind="node_spec",
         )
     return requirements

src/wf_mcp/workflow_surface/runtime_dependencies.py

@@ -44,7 +44,7 @@ def resolve_runtime_dependencies(
         node_name_bindings[node_name] = concrete_name
         node_specs[concrete_name] = spec
     reducers = _resolve_reducers(
-        required_capabilities=artifact.required_capabilities,
+        required_capabilities=artifact.required_capability_map(),
         deployment=deployment,
         sources=sources,
     )
@@ -68,7 +68,7 @@ def _resolve_node_spec(
 
     if deployment is not None:
         try:
-            bound_ref = CapabilityRef.parse(node_name).bind(deployment.bindings)
+            bound_ref = CapabilityRef.parse(node_name).bind(deployment.binding_map())
         except ValueError:
             bound_ref = None
         if bound_ref is not None:
@@ -104,7 +104,7 @@ def _resolve_reducers(
     for logical_ref, required in required_capabilities.items():
         if required.kind != "reducer":
             continue
-        bound_source_id = deployment.bindings.get(required.logical_source)
+        bound_source_id = deployment.binding_map().get(required.logical_source)
         if bound_source_id is None:
             continue
         source = sources.get(bound_source_id)

src/wf_platform/refs.py

@@ -2,6 +2,9 @@ from __future__ import annotations
 
 from collections.abc import Mapping
 from dataclasses import dataclass
+from typing import Any
+
+from pydantic_core import core_schema
 
 
 @dataclass(frozen=True, slots=True)
@@ -11,7 +14,7 @@ class SourceRef:
     parts: tuple[str, ...]
 
     def __post_init__(self) -> None:
-        if not self.parts or any(not part for part in self.parts):
+        if not self.parts or any(not part or not part.strip() for part in self.parts):
             raise ValueError("source ref requires non-empty path segments")
 
     @classmethod
@@ -22,6 +25,29 @@ class SourceRef:
     def __str__(self) -> str:
         return ".".join(self.parts)
 
+    @classmethod
+    def __get_pydantic_core_schema__(
+        cls,
+        _source_type: object,
+        _handler: object,
+    ) -> core_schema.CoreSchema:
+        """Validate refs from strings while serializing back to wire strings."""
+        return core_schema.no_info_plain_validator_function(
+            cls._validate,
+            serialization=core_schema.plain_serializer_function_ser_schema(
+                str,
+                when_used="json",
+            ),
+        )
+
+    @classmethod
+    def _validate(cls, value: Any) -> SourceRef:
+        if isinstance(value, SourceRef):
+            return value
+        if isinstance(value, str):
+            return cls.parse(value)
+        raise TypeError("source ref must be a string")
+
 
 @dataclass(frozen=True, slots=True)
 class CapabilityRef:
@@ -51,3 +77,26 @@ class CapabilityRef:
 
     def __str__(self) -> str:
         return f"{self.source}.{self.name}"
+
+    @classmethod
+    def __get_pydantic_core_schema__(
+        cls,
+        _source_type: object,
+        _handler: object,
+    ) -> core_schema.CoreSchema:
+        """Validate refs from strings while serializing back to wire strings."""
+        return core_schema.no_info_plain_validator_function(
+            cls._validate,
+            serialization=core_schema.plain_serializer_function_ser_schema(
+                str,
+                when_used="json",
+            ),
+        )
+
+    @classmethod
+    def _validate(cls, value: Any) -> CapabilityRef:
+        if isinstance(value, CapabilityRef):
+            return value
+        if isinstance(value, str):
+            return cls.parse(value)
+        raise TypeError("capability ref must be a string")

tests/artifacts/test_catalog.py

@@ -23,13 +23,12 @@ def artifact() -> WorkflowArtifact:
         },
         outcomes=("done", "failed"),
         plan={"name": "summarize_docs", "nodes": [{"id": "hidden"}]},
-        required_capabilities={
-            "context7.query-docs": RequiredCapability(
-                logical_source="context7",
-                capability_name="query-docs",
+        required_capabilities=[
+            RequiredCapability(
+                ref="context7.query-docs",
                 kind="tool",
             )
-        },
+        ],
     )
 
 

tests/artifacts/test_factory.py

@@ -16,8 +16,7 @@ def test_create_workflow_artifact_from_plan_derives_boundary_schemas() -> None:
         outcomes=("done",),
         required_capabilities={
             "demo.echo_tool": RequiredCapability(
-                logical_source="demo",
-                capability_name="echo_tool",
+                ref="demo.echo_tool",
                 kind="node_spec",
             )
         },
@@ -32,7 +31,7 @@ def test_create_workflow_artifact_from_plan_derives_boundary_schemas() -> None:
     assert artifact.output_schema["properties"]["echoed"]["type"] == "string"
     assert artifact.outcomes == ("done",)
     assert artifact.plan["name"] == "echo"
-    assert "demo.echo_tool" in artifact.required_capabilities
+    assert "demo.echo_tool" in artifact.required_capability_map()
     assert artifact.created_from_catalog_version == "catalog-1"
 
 
@@ -51,7 +50,7 @@ def test_create_workflow_artifact_from_plan_adds_reducer_dependencies() -> None:
         outcomes=("done",),
     )
 
-    reducer = artifact.required_capabilities["wf.std.max"]
+    reducer = artifact.required_capability_map()["wf.std.max"]
     assert reducer.logical_source == "wf.std"
     assert reducer.capability_name == "max"
     assert reducer.kind == "reducer"
@@ -71,12 +70,12 @@ def test_create_workflow_artifact_from_plan_rewrites_bound_node_specs() -> None:
     )
 
     node = artifact.plan["nodes"][0]
-    required = artifact.required_capabilities["demo.echo_tool"]
+    required = artifact.required_capability_map()["demo.echo_tool"]
     assert node["node"] == "demo.echo_tool"
     assert required.logical_source == "demo"
     assert required.capability_name == "echo_tool"
     assert required.kind == "node_spec"
-    assert required.observed_concrete_source == "demo.personal"
+    assert str(required.observed_concrete_source) == "demo.personal"
 
 
 def test_create_workflow_artifact_from_plan_snapshots_observed_node_spec() -> None:
@@ -108,7 +107,7 @@ def test_create_workflow_artifact_from_plan_snapshots_observed_node_spec() -> No
         },
     )
 
-    required = artifact.required_capabilities["demo.echo_tool"]
+    required = artifact.required_capability_map()["demo.echo_tool"]
     assert required.input_schema_snapshot == {
         "type": "object",
         "properties": {"text": {"type": "string"}},
@@ -136,15 +135,14 @@ def test_create_workflow_artifact_from_plan_keeps_explicit_capability_metadata()
         source_bindings={"demo": "demo.personal"},
         required_capabilities={
             "demo.echo_tool": RequiredCapability(
-                logical_source="demo",
-                capability_name="echo_tool",
+                ref="demo.echo_tool",
                 kind="node_spec",
                 input_schema_hash="sha256:explicit",
             )
         },
     )
 
-    required = artifact.required_capabilities["demo.echo_tool"]
+    required = artifact.required_capability_map()["demo.echo_tool"]
     assert required.input_schema_hash == "sha256:explicit"
 
 

tests/artifacts/test_models.py

@@ -12,8 +12,7 @@ from wf_artifacts import (
 
 def test_workflow_artifact_serializes_required_capability_contract() -> None:
     capability = RequiredCapability(
-        logical_source="context7",
-        capability_name="query-docs",
+        ref="context7.query-docs",
         kind="tool",
         input_schema_hash="sha256:input",
         input_schema_snapshot={"type": "object", "properties": {}},
@@ -31,7 +30,7 @@ def test_workflow_artifact_serializes_required_capability_contract() -> None:
         output_schema={"type": "object", "properties": {}},
         outcomes=("done", "failed"),
         plan={"name": "summarize_docs", "nodes": [], "edges": []},
-        required_capabilities={"context7.query-docs": capability},
+        required_capabilities=[capability],
         created_from_catalog_version="catalog-1",
     )
 
@@ -41,9 +40,10 @@ def test_workflow_artifact_serializes_required_capability_contract() -> None:
     assert dumped["kind"] == "workflow"
     assert dumped["version"] == 1
     assert dumped["outcomes"] == ["done", "failed"]
-    required = dumped["required_capabilities"]["context7.query-docs"]
-    assert required["logical_source"] == "context7"
+    required = dumped["required_capabilities"][0]
+    assert required["ref"] == "context7.query-docs"
     assert required["input_schema_hash"] == "sha256:input"
+    assert artifact.required_capability_map()["context7.query-docs"] == capability
 
 
 def test_workflow_artifact_can_be_marked_as_wrapper_intent() -> None:
@@ -68,7 +68,9 @@ def test_workflow_deployment_binds_logical_sources_to_concrete_sources() -> None
         id="summarize_docs.personal",
         artifact_id="summarize_docs",
         artifact_version=1,
-        bindings={"context7": "context7.personal"},
+        bindings=[
+            {"logical_source": "context7", "concrete_source": "context7.personal"}
+        ],
         drift_policy=DriftPolicy.BLOCK,
     )
 
@@ -77,8 +79,10 @@ def test_workflow_deployment_binds_logical_sources_to_concrete_sources() -> None
     assert dumped["id"] == "summarize_docs.personal"
     assert dumped["artifact_id"] == "summarize_docs"
     assert dumped["artifact_version"] == 1
-    assert dumped["bindings"]["context7"] == "context7.personal"
+    assert dumped["bindings"][0]["logical_source"] == "context7"
+    assert dumped["bindings"][0]["concrete_source"] == "context7.personal"
     assert dumped["drift_policy"] == "block"
+    assert deployment.binding_map()["context7"] == "context7.personal"
 
 
 def test_dependency_diagnostic_is_structured() -> None:

tests/artifacts/test_store.py

@@ -50,7 +50,9 @@ def test_file_store_round_trips_deployment(tmp_path) -> None:
         id="summarize_docs.personal",
         artifact_id="summarize_docs",
         artifact_version=1,
-        bindings={"context7": "context7.personal"},
+        bindings=[
+            {"logical_source": "context7", "concrete_source": "context7.personal"}
+        ],
     )
 
     store.save_deployment(deployment)
@@ -58,7 +60,7 @@ def test_file_store_round_trips_deployment(tmp_path) -> None:
 
     assert loaded.id == "summarize_docs.personal"
     assert loaded.artifact_id == "summarize_docs"
-    assert loaded.bindings["context7"] == "context7.personal"
+    assert loaded.binding_map()["context7"] == "context7.personal"
 
 
 def test_file_store_lists_deployments_in_id_order(tmp_path) -> None:
@@ -68,7 +70,9 @@ def test_file_store_lists_deployments_in_id_order(tmp_path) -> None:
             id="summarize_docs.work",
             artifact_id="summarize_docs",
             artifact_version=1,
-            bindings={"context7": "context7.work"},
+            bindings=[
+                {"logical_source": "context7", "concrete_source": "context7.work"}
+            ],
         )
     )
     store.save_deployment(
@@ -76,7 +80,9 @@ def test_file_store_lists_deployments_in_id_order(tmp_path) -> None:
             id="summarize_docs.personal",
             artifact_id="summarize_docs",
             artifact_version=1,
-            bindings={"context7": "context7.personal"},
+            bindings=[
+                {"logical_source": "context7", "concrete_source": "context7.personal"}
+            ],
         )
     )
 
@@ -86,4 +92,4 @@ def test_file_store_lists_deployments_in_id_order(tmp_path) -> None:
         "summarize_docs.personal",
         "summarize_docs.work",
     ]
-    assert deployments[0].bindings["context7"] == "context7.personal"
+    assert deployments[0].binding_map()["context7"] == "context7.personal"

tests/artifacts/test_validation.py

@@ -19,8 +19,7 @@ def required_capability(
     output_hash: str = "sha256:output",
 ) -> RequiredCapability:
     return RequiredCapability(
-        logical_source=logical_source,
-        capability_name=capability_name,
+        ref=f"{logical_source}.{capability_name}",
         kind="tool",
         input_schema_hash=input_hash,
         input_schema_snapshot={"type": "object", "properties": {}},
@@ -38,9 +37,7 @@ def artifact_with(capability: RequiredCapability) -> WorkflowArtifact:
         output_schema={"type": "object", "properties": {}},
         outcomes=("done",),
         plan={"name": "summarize_docs", "nodes": [], "edges": []},
-        required_capabilities={
-            f"{capability.logical_source}.{capability.capability_name}": capability
-        },
+        required_capabilities=[capability],
     )
 
 
@@ -53,7 +50,14 @@ def deployment(
         id="summarize_docs.personal",
         artifact_id="summarize_docs",
         artifact_version=1,
-        bindings={"context7": "context7.personal"} if bindings is None else bindings,
+        bindings=(
+            [{"logical_source": "context7", "concrete_source": "context7.personal"}]
+            if bindings is None
+            else [
+                {"logical_source": logical, "concrete_source": concrete}
+                for logical, concrete in bindings.items()
+            ]
+        ),
         drift_policy=drift_policy,
     )
 
@@ -180,8 +184,7 @@ def test_validate_deployment_allows_changed_schema_when_policy_allows() -> None:
 
 def test_validate_deployment_accepts_reducer_capability() -> None:
     reducer = RequiredCapability(
-        logical_source="wf.std",
-        capability_name="set_union",
+        ref="wf.std.set_union",
         kind="reducer",
     )
     diagnostics = validate_deployment_dependencies(

tests/refs/test_platform_refs.py

@@ -1,5 +1,7 @@
 from __future__ import annotations
 
+from pydantic import BaseModel
+
 from wf_platform import CapabilityRef, SourceRef
 
 
@@ -26,6 +28,35 @@ def test_capability_ref_binds_logical_source_to_concrete_source() -> None:
     assert str(bound) == "demo.personal.echo_tool"
 
 
+def test_platform_refs_validate_and_serialize_through_pydantic() -> None:
+    class Payload(BaseModel):
+        source: SourceRef
+        capability: CapabilityRef
+
+    payload = Payload.model_validate(
+        {
+            "source": "demo.personal",
+            "capability": "demo.personal.echo_tool",
+        }
+    )
+
+    assert payload.source == SourceRef.parse("demo.personal")
+    assert payload.capability == CapabilityRef.parse("demo.personal.echo_tool")
+    assert payload.model_dump(mode="json") == {
+        "source": "demo.personal",
+        "capability": "demo.personal.echo_tool",
+    }
+
+
+def test_source_ref_rejects_whitespace_segments() -> None:
+    try:
+        SourceRef.parse("demo. .personal")
+    except ValueError as exc:
+        assert "non-empty path segments" in str(exc)
+    else:
+        raise AssertionError("expected whitespace source segment to fail")
+
+
 def test_capability_ref_rejects_missing_capability_name() -> None:
     try:
         CapabilityRef.parse("demo")

tests/wf_mcp/test_broker_server.py

@@ -204,7 +204,9 @@ def test_broker_validates_workflow_deployment_from_artifact_store() -> None:
             id="summarize_docs.personal",
             artifact_id="summarize_docs",
             artifact_version=1,
-            bindings={"context7": "context7.personal"},
+            bindings=[
+                {"logical_source": "context7", "concrete_source": "context7.personal"}
+            ],
         )
     )
     service = WfMcpService(
@@ -290,7 +292,7 @@ def test_broker_creates_workflow_artifact_from_plan() -> None:
     assert payload["saved"] is True
     assert loaded.input_schema["properties"]["text"]["type"] == "string"
     assert loaded.output_schema["properties"]["echoed"]["type"] == "string"
-    assert loaded.required_capabilities["demo.echo_tool"].logical_source == "demo"
+    assert loaded.required_capability_map()["demo.echo_tool"].logical_source == "demo"
 
 
 def test_broker_saves_and_lists_workflow_deployments() -> None:
@@ -311,7 +313,12 @@ def test_broker_saves_and_lists_workflow_deployments() -> None:
                     id="summarize_docs.personal",
                     artifact_id="summarize_docs",
                     artifact_version=1,
-                    bindings={"context7": "context7.personal"},
+                    bindings=[
+                        {
+                            "logical_source": "context7",
+                            "concrete_source": "context7.personal",
+                        }
+                    ],
                 ).model_dump(mode="json")
             },
         )
@@ -324,7 +331,9 @@ def test_broker_saves_and_lists_workflow_deployments() -> None:
 
     assert save_payload["deployment_id"] == "summarize_docs.personal"
     assert list_payload["deployments"][0]["id"] == "summarize_docs.personal"
-    assert list_payload["deployments"][0]["bindings"]["context7"] == "context7.personal"
+    binding = list_payload["deployments"][0]["bindings"][0]
+    assert binding["logical_source"] == "context7"
+    assert binding["concrete_source"] == "context7.personal"
 
 
 def test_broker_runs_non_interrupting_workflow_deployment() -> None:
@@ -337,7 +346,7 @@ def test_broker_runs_non_interrupting_workflow_deployment() -> None:
             id="echo.personal",
             artifact_id="echo",
             artifact_version=1,
-            bindings={"demo": "demo.personal"},
+            bindings=[{"logical_source": "demo", "concrete_source": "demo.personal"}],
         )
     )
     service = WfMcpService(
@@ -379,7 +388,9 @@ def test_broker_run_deployment_returns_unrunnable_for_dependency_errors() -> Non
             id="summarize_docs.personal",
             artifact_id="summarize_docs",
             artifact_version=1,
-            bindings={"context7": "context7.personal"},
+            bindings=[
+                {"logical_source": "context7", "concrete_source": "context7.personal"}
+            ],
         )
     )
     service = WfMcpService(
@@ -414,7 +425,7 @@ def test_broker_run_deployment_rejects_interrupting_artifacts() -> None:
             id="approval.personal",
             artifact_id="approval",
             artifact_version=1,
-            bindings={},
+            bindings=[],
         )
     )
     service = WfMcpService(
@@ -464,8 +475,7 @@ def _artifact() -> WorkflowArtifact:
         plan={"name": "summarize_docs", "nodes": [], "edges": []},
         required_capabilities={
             "context7.query-docs": RequiredCapability(
-                logical_source="context7",
-                capability_name="query-docs",
+                ref="context7.query-docs",
                 kind="tool",
                 input_schema_hash="sha256:input",
                 output_schema_hash="sha256:output",
@@ -518,8 +528,7 @@ def _echo_artifact() -> WorkflowArtifact:
         },
         required_capabilities={
             "demo.echo_tool": RequiredCapability(
-                logical_source="demo",
-                capability_name="echo_tool",
+                ref="demo.echo_tool",
                 kind="node_spec",
             )
         },

tests/wf_mcp/test_workflow_surface.py

@@ -189,7 +189,9 @@ def test_workflow_surface_validates_deployment_dependencies() -> None:
             id="summarize_docs.personal",
             artifact_id="summarize_docs",
             artifact_version=1,
-            bindings={"context7": "context7.personal"},
+            bindings=[
+                {"logical_source": "context7", "concrete_source": "context7.personal"}
+            ],
         )
     )
     handlers = _handlers(artifact_store)
@@ -215,7 +217,9 @@ def test_workflow_surface_records_artifact_and_deployment_save_events() -> None:
                 id="echo.personal",
                 artifact_id="echo",
                 artifact_version=1,
-                bindings={"demo": "demo.personal"},
+                bindings=[
+                    {"logical_source": "demo", "concrete_source": "demo.personal"}
+                ],
             ).model_dump(mode="json")
         )
     )
@@ -274,7 +278,7 @@ def test_workflow_surface_creates_artifact_with_logical_node_refs() -> None:
     artifact = artifact_store.get_artifact("echo_logical", 1)
 
     assert artifact.plan["nodes"][0]["node"] == "demo.echo_tool"
-    assert artifact.required_capabilities["demo.echo_tool"].logical_source == "demo"
+    assert artifact.required_capability_map()["demo.echo_tool"].logical_source == "demo"
 
 
 def test_workflow_surface_validates_draft_without_saving() -> None:
@@ -349,7 +353,7 @@ def test_workflow_surface_creates_artifact_from_draft_with_binding_suggestions()
     assert payload["required_logical_sources"] == ["demo", "wf.std"]
     assert payload["suggested_bindings"]["wf.std"] == "wf.std"
     assert artifact.plan["nodes"][0]["node"] == "demo.echo_tool"
-    assert artifact.required_capabilities["demo.echo_tool"].logical_source == "demo"
+    assert artifact.required_capability_map()["demo.echo_tool"].logical_source == "demo"
 
 
 def test_workflow_surface_draft_artifact_requires_std_self_binding() -> None:
@@ -373,7 +377,7 @@ def test_workflow_surface_draft_artifact_requires_std_self_binding() -> None:
             id="draft_echo_missing_std.personal",
             artifact_id="draft_echo_missing_std",
             artifact_version=1,
-            bindings={"demo": "demo.personal"},
+            bindings=[{"logical_source": "demo", "concrete_source": "demo.personal"}],
         )
     )
 
@@ -733,7 +737,7 @@ def test_workflow_surface_runs_non_interrupting_deployment() -> None:
             id="echo.personal",
             artifact_id="echo",
             artifact_version=1,
-            bindings={"demo": "demo.personal"},
+            bindings=[{"logical_source": "demo", "concrete_source": "demo.personal"}],
         )
     )
     service = WfMcpService(
@@ -766,7 +770,7 @@ def test_workflow_surface_runs_deployment_with_bound_node_spec_dependency() -> N
             id="echo.personal",
             artifact_id="logical_echo",
             artifact_version=1,
-            bindings={"demo": "demo.personal"},
+            bindings=[{"logical_source": "demo", "concrete_source": "demo.personal"}],
         )
     )
     service = WfMcpService(
@@ -836,7 +840,7 @@ def test_workflow_surface_runs_artifact_created_from_concrete_node_ref() -> None
     artifact = artifact_store.get_artifact("created_echo", 1)
 
     assert artifact.plan["nodes"][0]["node"] == "demo.echo_tool"
-    assert artifact.required_capabilities["demo.echo_tool"].logical_source == "demo"
+    assert artifact.required_capability_map()["demo.echo_tool"].logical_source == "demo"
     assert payload["status"] == "completed"
     assert payload["output"]["echoed"] == "hello"
     assert payload["diagnostics"] == []
@@ -881,7 +885,7 @@ def test_workflow_surface_detects_drift_from_saved_node_spec_snapshot() -> None:
     required = artifact_store.get_artifact(
         "created_echo_drift",
         1,
-    ).required_capabilities["demo.echo_tool"]
+    ).required_capability_map()["demo.echo_tool"]
     assert required.input_schema_hash is not None
 
     service.register_connection(
@@ -1034,7 +1038,7 @@ def test_workflow_surface_calls_saved_wrapper_artifact_with_deployment_bindings(
             id="logical_echo_wrapper.personal",
             artifact_id="logical_echo_wrapper",
             artifact_version=1,
-            bindings={"demo": "demo.personal"},
+            bindings=[{"logical_source": "demo", "concrete_source": "demo.personal"}],
         )
     )
     service = WfMcpService(
@@ -1084,8 +1088,7 @@ def _artifact() -> WorkflowArtifact:
         plan={"name": "summarize_docs", "nodes": [], "edges": []},
         required_capabilities={
             "context7.query-docs": RequiredCapability(
-                logical_source="context7",
-                capability_name="query-docs",
+                ref="context7.query-docs",
                 kind="tool",
                 input_schema_hash="sha256:input",
                 output_schema_hash="sha256:output",
@@ -1130,8 +1133,7 @@ def _echo_artifact() -> WorkflowArtifact:
         plan=plan,
         required_capabilities={
             "demo.echo_tool": RequiredCapability(
-                logical_source="demo",
-                capability_name="echo_tool",
+                ref="demo.echo_tool",
                 kind="node_spec",
             )
         },
@@ -1196,7 +1198,7 @@ def _custom_reducer_artifact() -> WorkflowArtifact:
                     "type": "integer",
                     "reducer": "custom.multiply",
                 }
-            }
+            },
         },
         "output_schema": {
             "type": "object",
@@ -1225,13 +1227,11 @@ def _custom_reducer_artifact() -> WorkflowArtifact:
         plan=plan,
         required_capabilities={
             "demo.amount_tool": RequiredCapability(
-                logical_source="demo",
-                capability_name="amount_tool",
+                ref="demo.amount_tool",
                 kind="node_spec",
             ),
             "custom.multiply": RequiredCapability(
-                logical_source="custom",
-                capability_name="multiply",
+                ref="custom.multiply",
                 kind="reducer",
             ),
         },