Implemented durable workflow/job foundation

docs/asset-registry-implementation.md

@@ -65,6 +65,7 @@ and SQLite repositories are adapters behind those ports.
 - Idempotency-key reuse with a different payload raises a validation error.
 - Transformation run and derived lineage persistence for traceable derived
   artifact creation.
+- Workflow template and workflow run persistence for durable job execution.
 - In-memory repository for deterministic tests.
 - SQLite repository for local-first durable asset registry state.
 - SQLite foreign-key enforcement for representation and metadata asset
@@ -89,12 +90,14 @@ and SQLite repositories are adapters behind those ports.
 - `ingestion_jobs`
 - `transformation_runs`
 - `derived_lineage`
+- `workflow_templates`
+- `workflow_runs`
 
 Payloads are stored as compact JSON envelopes while indexed columns carry
 stable lookup fields such as asset ID, lifecycle, representation kind, digest,
 sequence, relationship source/target, actor ID, target, correlation ID,
-idempotency key, transformation status, operation ID, and derived output asset
+idempotency key, transformation status, operation ID, derived output asset ID,
-ID.
+workflow template ID, workflow version, and workflow run status.
 
 ## Not Yet Implemented
 

docs/workflow-jobs-implementation.md

@@ -0,0 +1,104 @@
+# Workflow Jobs Implementation Note
+
+Date: 2026-05-06
+
+Status: active implementation note for `KONT-WP-0008`.
+
+## Purpose
+
+This note records the first durable workflow and job-runner foundation. It
+extends traceable transformations into reusable workflow templates with
+dependency-ordered execution and recoverable run state.
+
+## Implemented Package Shape
+
+```text
+src/kontextual_engine/
+  core/
+    workflow_jobs.py
+  services/
+    workflow_service.py
+  ports/
+    repositories.py
+  adapters/
+    memory/asset_registry.py
+    sqlite/asset_registry.py
+```
+
+The workflow service uses the existing repository and policy ports. Executable
+steps delegate to `TransformationService`, so workflow runs do not bypass asset
+governance, transformation run records, derived lineage, or audit.
+
+## Implemented Capabilities
+
+- `WorkflowTemplate` for reusable template identity, version, metadata,
+  inputs, steps, policy checks, and failure behavior.
+- `WorkflowInputDefinition` supports asset, collection, query, source event,
+  and payload inputs.
+- `WorkflowStepDefinition` captures step kind, operation ID, dependencies,
+  input/output bindings, preconditions, failure behavior, and metadata.
+- Template registration validates duplicate step IDs, missing dependencies,
+  dependency cycles, and unsupported failure behavior.
+- `WorkflowRun` and `WorkflowStepRun` persist queued, running, waiting,
+  completed, partially completed, failed, retried, and canceled states.
+- `WorkflowService` can register templates, queue runs, invoke runs, resume
+  runs, retry runs, and cancel runs programmatically.
+- The MVP runner executes transformation-backed steps in dependency order.
+- Step inputs can resolve source asset IDs from invocation inputs or completed
+  upstream step outputs.
+- Retry and repeated invocation avoid silent overwrite by choosing a fresh
+  output asset ID when a fixed template output ID already exists.
+- Workflow template and run persistence are implemented for memory and SQLite.
+- Audit events are emitted for template registration, run queue/start/final
+  states, step start/final states, retry, and cancellation.
+
+## Current Boundaries
+
+The runner executes only `kind="transformation"` steps. Non-transformation
+steps are represented and fail with structured diagnostics until a later
+adapter or job worker handles them.
+
+Workflow inputs preserve collection, query, source event, and payload bindings,
+but the MVP runner only interprets asset bindings for transformation execution.
+Query expansion, source-event ingestion, human tasks, and exception queues stay
+in later WP-0008 tasks.
+
+Markdown-specific transformations remain adapter-backed through markitect-tool.
+Workflow orchestration may invoke those operations once the adapter boundary is
+wired, but the engine does not reimplement Markdown semantics.
+
+## Current SQLite Tables
+
+WP-0008 adds shared registry persistence for:
+
+- `workflow_templates`
+- `workflow_runs`
+
+The tables store compact JSON payloads with indexed lookup columns for template
+ID, template version, template name, run ID, run status, actor ID, correlation
+ID, queued timestamp, and updated timestamp.
+
+## Not Yet Implemented
+
+- Human review gates and approval tasks.
+- Exception queues for blocked, failed, low-confidence, or policy-conflicted
+  items.
+- Queue-worker adapters beyond embedded execution.
+- Rich retry policies by operation type.
+- Query/input expansion into dynamic asset sets.
+- Full workflow reconstruction views across all audit and lineage records.
+
+These remain in open tasks `KONT-WP-0008-T006` and `KONT-WP-0008-T007`.
+
+## Test Coverage
+
+`tests/test_workflow_service.py` covers:
+
+- template registration, input-kind persistence, missing dependency rejection,
+  and dependency-cycle rejection,
+- dependency-ordered execution of two transformation steps,
+- queue, cancel, resume denial, and retry without direct storage edits,
+- partial completion when a continue-on-failure step fails and another step
+  succeeds,
+- SQLite reload of workflow templates, workflow runs, step state, and derived
+  output representation state.

src/kontextual_engine/__init__.py

@@ -58,6 +58,14 @@ from .core import (
     TransformationRun,
     TransformationRunStatus,
     VersionChangeType,
+    WorkflowInputDefinition,
+    WorkflowInputKind,
+    WorkflowRun,
+    WorkflowRunStatus,
+    WorkflowStepDefinition,
+    WorkflowStepRun,
+    WorkflowStepRunStatus,
+    WorkflowTemplate,
 )
 from .errors import (
     AdapterUnavailableError,
@@ -109,6 +117,9 @@ from .services import (
     TransformationRequest,
     TransformationRunResult,
     TransformationService,
+    WorkflowInvocation,
+    WorkflowRunResult,
+    WorkflowService,
     default_transformation_registry,
 )
 from .storage import InMemoryKnowledgeRepository
@@ -232,7 +243,18 @@ __all__ = [
     "TransformationService",
     "ValidationError",
     "VersionChangeType",
+    "WorkflowInputDefinition",
+    "WorkflowInputKind",
+    "WorkflowInvocation",
+    "WorkflowRun",
+    "WorkflowRunResult",
+    "WorkflowRunStatus",
+    "WorkflowService",
     "WorkflowStep",
+    "WorkflowStepDefinition",
+    "WorkflowStepRun",
+    "WorkflowStepRunStatus",
+    "WorkflowTemplate",
     "bundle_digest",
     "content_digest",
     "default_transformation_registry",

src/kontextual_engine/adapters/memory/asset_registry.py

@@ -26,6 +26,9 @@ from kontextual_engine.core import (
     Sensitivity,
     TransformationRun,
     TransformationRunStatus,
+    WorkflowRun,
+    WorkflowRunStatus,
+    WorkflowTemplate,
 )
 from kontextual_engine.errors import NotFoundError, ValidationError
 
@@ -47,6 +50,8 @@ class InMemoryAssetRegistryRepository:
     ingestion_jobs: dict[str, IngestionJob] = field(default_factory=dict)
     transformation_runs: dict[str, TransformationRun] = field(default_factory=dict)
     derived_lineage: dict[str, DerivedArtifactLineage] = field(default_factory=dict)
+    workflow_templates: dict[tuple[str, str], WorkflowTemplate] = field(default_factory=dict)
+    workflow_runs: dict[str, WorkflowRun] = field(default_factory=dict)
 
     def save_actor(self, actor: Actor) -> Actor:
         self.actors[actor.id] = actor
@@ -357,6 +362,64 @@ class InMemoryAssetRegistryRepository:
             records = [record for record in records if record.transformation_run_id == transformation_run_id]
         return sorted(records, key=lambda record: (record.transformation_run_id, record.lineage_id))
 
+    def save_workflow_template(self, template: WorkflowTemplate) -> WorkflowTemplate:
+        self.workflow_templates[(template.template_id, template.version)] = template
+        return template
+
+    def get_workflow_template(
+        self,
+        template_id: str,
+        *,
+        version: str | None = None,
+    ) -> WorkflowTemplate:
+        if version is not None:
+            try:
+                return self.workflow_templates[(template_id, version)]
+            except KeyError as exc:
+                raise NotFoundError(
+                    "Workflow template not found",
+                    details={"template_id": template_id, "version": version},
+                ) from exc
+        candidates = [item for key, item in self.workflow_templates.items() if key[0] == template_id]
+        if not candidates:
+            raise NotFoundError("Workflow template not found", details={"template_id": template_id})
+        return sorted(candidates, key=lambda item: (item.updated_at, item.version))[-1]
+
+    def list_workflow_templates(
+        self,
+        *,
+        template_id: str | None = None,
+    ) -> list[WorkflowTemplate]:
+        templates: Iterable[WorkflowTemplate] = self.workflow_templates.values()
+        if template_id is not None:
+            templates = [item for item in templates if item.template_id == template_id]
+        return sorted(templates, key=lambda item: (item.name, item.version, item.template_id))
+
+    def save_workflow_run(self, run: WorkflowRun) -> WorkflowRun:
+        self.get_actor(run.actor_id)
+        self.get_workflow_template(run.template_id, version=run.template_version)
+        self.workflow_runs[run.run_id] = run
+        return run
+
+    def get_workflow_run(self, run_id: str) -> WorkflowRun:
+        try:
+            return self.workflow_runs[run_id]
+        except KeyError as exc:
+            raise NotFoundError("Workflow run not found", details={"run_id": run_id}) from exc
+
+    def list_workflow_runs(
+        self,
+        *,
+        status: WorkflowRunStatus | None = None,
+        template_id: str | None = None,
+    ) -> list[WorkflowRun]:
+        runs: Iterable[WorkflowRun] = self.workflow_runs.values()
+        if status is not None:
+            runs = [run for run in runs if run.status == status]
+        if template_id is not None:
+            runs = [run for run in runs if run.template_id == template_id]
+        return sorted(runs, key=lambda run: (run.queued_at, run.run_id))
+
 
 def _metadata_matches(
     records: list[MetadataRecord],

src/kontextual_engine/adapters/sqlite/asset_registry.py

@@ -29,6 +29,9 @@ from kontextual_engine.core import (
     Sensitivity,
     TransformationRun,
     TransformationRunStatus,
+    WorkflowRun,
+    WorkflowRunStatus,
+    WorkflowTemplate,
 )
 from kontextual_engine.errors import NotFoundError, ValidationError
 
@@ -723,6 +726,139 @@ class SQLiteAssetRegistryRepository:
             records = [record for record in records if source_asset_id in record.source_asset_ids]
         return records
 
+    def save_workflow_template(self, template: WorkflowTemplate) -> WorkflowTemplate:
+        with self._connect() as conn:
+            conn.execute(
+                """
+                insert into workflow_templates
+                  (id, version, name, updated_at, payload)
+                values (?, ?, ?, ?, ?)
+                on conflict(id, version) do update set
+                  name=excluded.name,
+                  updated_at=excluded.updated_at,
+                  payload=excluded.payload
+                """,
+                (
+                    template.template_id,
+                    template.version,
+                    template.name,
+                    template.updated_at,
+                    _json(template.to_dict()),
+                ),
+            )
+        return template
+
+    def get_workflow_template(
+        self,
+        template_id: str,
+        *,
+        version: str | None = None,
+    ) -> WorkflowTemplate:
+        if version is None:
+            row = self._one(
+                """
+                select payload from workflow_templates
+                where id = ?
+                order by updated_at desc, version desc
+                limit 1
+                """,
+                (template_id,),
+            )
+        else:
+            row = self._one(
+                "select payload from workflow_templates where id = ? and version = ?",
+                (template_id, version),
+            )
+        if row is None:
+            details = {"template_id": template_id}
+            if version is not None:
+                details["version"] = version
+            raise NotFoundError("Workflow template not found", details=details)
+        return WorkflowTemplate.from_dict(_loads(row["payload"]))
+
+    def list_workflow_templates(
+        self,
+        *,
+        template_id: str | None = None,
+    ) -> list[WorkflowTemplate]:
+        if template_id is None:
+            rows = self._all("select payload from workflow_templates order by name, version, id", ())
+        else:
+            rows = self._all(
+                "select payload from workflow_templates where id = ? order by name, version, id",
+                (template_id,),
+            )
+        return [WorkflowTemplate.from_dict(_loads(row["payload"])) for row in rows]
+
+    def save_workflow_run(self, run: WorkflowRun) -> WorkflowRun:
+        try:
+            with self._connect() as conn:
+                conn.execute(
+                    """
+                    insert into workflow_runs
+                      (id, template_id, template_version, status, actor_id, correlation_id,
+                       queued_at, updated_at, payload)
+                    values (?, ?, ?, ?, ?, ?, ?, ?, ?)
+                    on conflict(id) do update set
+                      template_id=excluded.template_id,
+                      template_version=excluded.template_version,
+                      status=excluded.status,
+                      actor_id=excluded.actor_id,
+                      correlation_id=excluded.correlation_id,
+                      queued_at=excluded.queued_at,
+                      updated_at=excluded.updated_at,
+                      payload=excluded.payload
+                    """,
+                    (
+                        run.run_id,
+                        run.template_id,
+                        run.template_version,
+                        run.status.value,
+                        run.actor_id,
+                        run.correlation_id,
+                        run.queued_at,
+                        run.updated_at,
+                        _json(run.to_dict()),
+                    ),
+                )
+        except sqlite3.IntegrityError as exc:
+            if _is_foreign_key_error(exc):
+                raise ValidationError(
+                    "Workflow run references an unknown actor or template",
+                    details={
+                        "actor_id": run.actor_id,
+                        "template_id": run.template_id,
+                        "template_version": run.template_version,
+                        "run_id": run.run_id,
+                    },
+                ) from exc
+            raise
+        return run
+
+    def get_workflow_run(self, run_id: str) -> WorkflowRun:
+        row = self._one("select payload from workflow_runs where id = ?", (run_id,))
+        if row is None:
+            raise NotFoundError("Workflow run not found", details={"run_id": run_id})
+        return WorkflowRun.from_dict(_loads(row["payload"]))
+
+    def list_workflow_runs(
+        self,
+        *,
+        status: WorkflowRunStatus | None = None,
+        template_id: str | None = None,
+    ) -> list[WorkflowRun]:
+        clauses = []
+        params: list[Any] = []
+        if status is not None:
+            clauses.append("status = ?")
+            params.append(status.value)
+        if template_id is not None:
+            clauses.append("template_id = ?")
+            params.append(template_id)
+        where = f" where {' and '.join(clauses)}" if clauses else ""
+        rows = self._all(f"select payload from workflow_runs{where} order by queued_at, id", tuple(params))
+        return [WorkflowRun.from_dict(_loads(row["payload"])) for row in rows]
+
     def _initialize(self) -> None:
         with self._connect() as conn:
             conn.executescript(
@@ -839,6 +975,28 @@ class SQLiteAssetRegistryRepository:
                   transformation_run_id text not null references transformation_runs(id) on delete cascade,
                   payload text not null
                 );
+                create table if not exists workflow_templates (
+                  id text not null,
+                  version text not null,
+                  name text not null,
+                  updated_at text not null,
+                  payload text not null,
+                  primary key(id, version)
+                );
+                create table if not exists workflow_runs (
+                  id text primary key,
+                  template_id text not null,
+                  template_version text not null,
+                  status text not null,
+                  actor_id text not null,
+                  correlation_id text not null,
+                  queued_at text not null,
+                  updated_at text not null,
+                  payload text not null,
+                  foreign key(actor_id) references actors(id),
+                  foreign key(template_id, template_version)
+                    references workflow_templates(id, version)
+                );
                 create index if not exists idx_assets_lifecycle on assets(lifecycle);
                 create index if not exists idx_representations_asset on representations(asset_id);
                 create index if not exists idx_metadata_asset on metadata_records(asset_id);
@@ -858,6 +1016,10 @@ class SQLiteAssetRegistryRepository:
                 create index if not exists idx_transformation_runs_correlation on transformation_runs(correlation_id);
                 create index if not exists idx_derived_lineage_output on derived_lineage(output_asset_id);
                 create index if not exists idx_derived_lineage_run on derived_lineage(transformation_run_id);
+                create index if not exists idx_workflow_templates_name on workflow_templates(name);
+                create index if not exists idx_workflow_runs_status on workflow_runs(status);
+                create index if not exists idx_workflow_runs_template on workflow_runs(template_id);
+                create index if not exists idx_workflow_runs_correlation on workflow_runs(correlation_id);
                 """
             )
 

src/kontextual_engine/core/__init__.py

@@ -42,6 +42,16 @@ from .relationships import (
 )
 from .retrieval_feedback import RetrievalFeedbackLabel, RetrievalFeedbackRecord
 from .transformations import TransformationOperation, TransformationRun, TransformationRunStatus
+from .workflow_jobs import (
+    WorkflowInputDefinition,
+    WorkflowInputKind,
+    WorkflowRun,
+    WorkflowRunStatus,
+    WorkflowStepDefinition,
+    WorkflowStepRun,
+    WorkflowStepRunStatus,
+    WorkflowTemplate,
+)
 
 __all__ = [
     "Actor",
@@ -87,6 +97,14 @@ __all__ = [
     "TransformationRun",
     "TransformationRunStatus",
     "VersionChangeType",
+    "WorkflowInputDefinition",
+    "WorkflowInputKind",
+    "WorkflowRun",
+    "WorkflowRunStatus",
+    "WorkflowStepDefinition",
+    "WorkflowStepRun",
+    "WorkflowStepRunStatus",
+    "WorkflowTemplate",
     "content_digest",
     "mapping_digest",
     "new_id",

src/kontextual_engine/core/workflow_jobs.py

@@ -0,0 +1,462 @@
+"""Workflow template and durable run primitives."""
+
+from __future__ import annotations
+
+from dataclasses import dataclass, field, replace
+from enum import Enum
+from typing import Any
+
+from .primitives import compact_dict, mapping_digest, new_id, utc_now
+
+
+class WorkflowInputKind(str, Enum):
+    ASSET = "asset"
+    COLLECTION = "collection"
+    QUERY = "query"
+    SOURCE_EVENT = "source_event"
+    PAYLOAD = "payload"
+
+
+class WorkflowRunStatus(str, Enum):
+    QUEUED = "queued"
+    RUNNING = "running"
+    WAITING = "waiting"
+    COMPLETED = "completed"
+    PARTIALLY_COMPLETED = "partially_completed"
+    FAILED = "failed"
+    RETRIED = "retried"
+    CANCELED = "canceled"
+
+
+class WorkflowStepRunStatus(str, Enum):
+    QUEUED = "queued"
+    RUNNING = "running"
+    WAITING = "waiting"
+    COMPLETED = "completed"
+    FAILED = "failed"
+    SKIPPED = "skipped"
+    CANCELED = "canceled"
+
+
+@dataclass(frozen=True)
+class WorkflowInputDefinition:
+    name: str
+    kind: WorkflowInputKind | str
+    required: bool = True
+    description: str = ""
+    metadata: dict[str, Any] = field(default_factory=dict)
+
+    def __post_init__(self) -> None:
+        object.__setattr__(self, "kind", WorkflowInputKind(self.kind))
+
+    def to_dict(self) -> dict[str, Any]:
+        return compact_dict(
+            {
+                "name": self.name,
+                "kind": self.kind.value,
+                "required": self.required,
+                "description": self.description,
+                "metadata": dict(self.metadata),
+            }
+        )
+
+    @classmethod
+    def from_dict(cls, data: dict[str, Any]) -> "WorkflowInputDefinition":
+        return cls(
+            name=data["name"],
+            kind=WorkflowInputKind(data["kind"]),
+            required=bool(data.get("required", True)),
+            description=data.get("description", ""),
+            metadata=dict(data.get("metadata", {})),
+        )
+
+
+@dataclass(frozen=True)
+class WorkflowStepDefinition:
+    step_id: str
+    kind: str = "transformation"
+    operation_id: str | None = None
+    depends_on: tuple[str, ...] = ()
+    inputs: dict[str, Any] = field(default_factory=dict)
+    parameters: dict[str, Any] = field(default_factory=dict)
+    outputs: dict[str, Any] = field(default_factory=dict)
+    preconditions: tuple[dict[str, Any], ...] = ()
+    failure_behavior: str = "fail_workflow"
+    required_permissions: tuple[str, ...] = ()
+    metadata: dict[str, Any] = field(default_factory=dict)
+
+    def __post_init__(self) -> None:
+        object.__setattr__(self, "depends_on", tuple(self.depends_on))
+        object.__setattr__(self, "preconditions", tuple(dict(item) for item in self.preconditions))
+        object.__setattr__(self, "required_permissions", tuple(self.required_permissions))
+
+    def to_dict(self) -> dict[str, Any]:
+        return compact_dict(
+            {
+                "step_id": self.step_id,
+                "kind": self.kind,
+                "operation_id": self.operation_id,
+                "depends_on": list(self.depends_on),
+                "inputs": dict(self.inputs),
+                "parameters": dict(self.parameters),
+                "outputs": dict(self.outputs),
+                "preconditions": list(self.preconditions),
+                "failure_behavior": self.failure_behavior,
+                "required_permissions": list(self.required_permissions),
+                "metadata": dict(self.metadata),
+            }
+        )
+
+    @classmethod
+    def from_dict(cls, data: dict[str, Any]) -> "WorkflowStepDefinition":
+        return cls(
+            step_id=data["step_id"],
+            kind=data.get("kind", "transformation"),
+            operation_id=data.get("operation_id"),
+            depends_on=tuple(data.get("depends_on", ())),
+            inputs=dict(data.get("inputs", {})),
+            parameters=dict(data.get("parameters", {})),
+            outputs=dict(data.get("outputs", {})),
+            preconditions=tuple(dict(item) for item in data.get("preconditions", ())),
+            failure_behavior=data.get("failure_behavior", "fail_workflow"),
+            required_permissions=tuple(data.get("required_permissions", ())),
+            metadata=dict(data.get("metadata", {})),
+        )
+
+
+@dataclass(frozen=True)
+class WorkflowTemplate:
+    name: str
+    inputs: tuple[WorkflowInputDefinition, ...] = ()
+    steps: tuple[WorkflowStepDefinition, ...] = ()
+    version: str = "1"
+    description: str = ""
+    policy_checks: tuple[dict[str, Any], ...] = ()
+    failure_behavior: str = "fail_workflow"
+    metadata: dict[str, Any] = field(default_factory=dict)
+    created_by: str | None = None
+    template_id: str = field(default_factory=lambda: new_id("wftpl"))
+    created_at: str = field(default_factory=lambda: utc_now().isoformat())
+    updated_at: str = field(default_factory=lambda: utc_now().isoformat())
+
+    def __post_init__(self) -> None:
+        object.__setattr__(self, "inputs", tuple(self.inputs))
+        object.__setattr__(self, "steps", tuple(self.steps))
+        object.__setattr__(self, "policy_checks", tuple(dict(item) for item in self.policy_checks))
+
+    @property
+    def template_hash(self) -> str:
+        return mapping_digest(self.to_dict(include_hash=False))
+
+    def with_actor(self, actor_id: str) -> "WorkflowTemplate":
+        return replace(self, created_by=self.created_by or actor_id, updated_at=utc_now().isoformat())
+
+    def to_dict(self, *, include_hash: bool = True) -> dict[str, Any]:
+        data = compact_dict(
+            {
+                "template_id": self.template_id,
+                "name": self.name,
+                "version": self.version,
+                "description": self.description,
+                "inputs": [item.to_dict() for item in self.inputs],
+                "steps": [item.to_dict() for item in self.steps],
+                "policy_checks": list(self.policy_checks),
+                "failure_behavior": self.failure_behavior,
+                "metadata": dict(self.metadata),
+                "created_by": self.created_by,
+                "created_at": self.created_at,
+                "updated_at": self.updated_at,
+            }
+        )
+        if include_hash:
+            data["template_hash"] = self.template_hash
+        return data
+
+    @classmethod
+    def from_dict(cls, data: dict[str, Any]) -> "WorkflowTemplate":
+        return cls(
+            template_id=data["template_id"],
+            name=data["name"],
+            version=data.get("version", "1"),
+            description=data.get("description", ""),
+            inputs=tuple(WorkflowInputDefinition.from_dict(item) for item in data.get("inputs", ())),
+            steps=tuple(WorkflowStepDefinition.from_dict(item) for item in data.get("steps", ())),
+            policy_checks=tuple(dict(item) for item in data.get("policy_checks", ())),
+            failure_behavior=data.get("failure_behavior", "fail_workflow"),
+            metadata=dict(data.get("metadata", {})),
+            created_by=data.get("created_by"),
+            created_at=data["created_at"],
+            updated_at=data["updated_at"],
+        )
+
+
+@dataclass(frozen=True)
+class WorkflowStepRun:
+    step_id: str
+    operation_id: str | None = None
+    status: WorkflowStepRunStatus = WorkflowStepRunStatus.QUEUED
+    transformation_run_id: str | None = None
+    output_asset_ids: tuple[str, ...] = ()
+    diagnostics: tuple[dict[str, Any], ...] = ()
+    started_at: str | None = None
+    completed_at: str | None = None
+    updated_at: str = field(default_factory=lambda: utc_now().isoformat())
+
+    def __post_init__(self) -> None:
+        object.__setattr__(self, "status", WorkflowStepRunStatus(self.status))
+        object.__setattr__(self, "output_asset_ids", tuple(self.output_asset_ids))
+        object.__setattr__(self, "diagnostics", tuple(self.diagnostics))
+
+    def running(self) -> "WorkflowStepRun":
+        now = utc_now().isoformat()
+        return replace(
+            self,
+            status=WorkflowStepRunStatus.RUNNING,
+            started_at=self.started_at or now,
+            updated_at=now,
+        )
+
+    def waiting(self, diagnostics: tuple[dict[str, Any], ...] = ()) -> "WorkflowStepRun":
+        return replace(
+            self,
+            status=WorkflowStepRunStatus.WAITING,
+            diagnostics=self.diagnostics + tuple(diagnostics),
+            updated_at=utc_now().isoformat(),
+        )
+
+    def completed(
+        self,
+        *,
+        transformation_run_id: str | None = None,
+        output_asset_ids: tuple[str, ...] = (),
+    ) -> "WorkflowStepRun":
+        now = utc_now().isoformat()
+        return replace(
+            self,
+            status=WorkflowStepRunStatus.COMPLETED,
+            transformation_run_id=transformation_run_id or self.transformation_run_id,
+            output_asset_ids=tuple(output_asset_ids),
+            completed_at=now,
+            updated_at=now,
+        )
+
+    def failed(self, diagnostics: tuple[dict[str, Any], ...]) -> "WorkflowStepRun":
+        now = utc_now().isoformat()
+        return replace(
+            self,
+            status=WorkflowStepRunStatus.FAILED,
+            diagnostics=self.diagnostics + tuple(diagnostics),
+            completed_at=now,
+            updated_at=now,
+        )
+
+    def skipped(self, diagnostics: tuple[dict[str, Any], ...]) -> "WorkflowStepRun":
+        now = utc_now().isoformat()
+        return replace(
+            self,
+            status=WorkflowStepRunStatus.SKIPPED,
+            diagnostics=self.diagnostics + tuple(diagnostics),
+            completed_at=now,
+            updated_at=now,
+        )
+
+    def canceled(self, diagnostics: tuple[dict[str, Any], ...] = ()) -> "WorkflowStepRun":
+        now = utc_now().isoformat()
+        return replace(
+            self,
+            status=WorkflowStepRunStatus.CANCELED,
+            diagnostics=self.diagnostics + tuple(diagnostics),
+            completed_at=now,
+            updated_at=now,
+        )
+
+    def to_dict(self) -> dict[str, Any]:
+        return compact_dict(
+            {
+                "step_id": self.step_id,
+                "operation_id": self.operation_id,
+                "status": self.status.value,
+                "transformation_run_id": self.transformation_run_id,
+                "output_asset_ids": list(self.output_asset_ids),
+                "diagnostics": list(self.diagnostics),
+                "started_at": self.started_at,
+                "completed_at": self.completed_at,
+                "updated_at": self.updated_at,
+            }
+        )
+
+    @classmethod
+    def from_dict(cls, data: dict[str, Any]) -> "WorkflowStepRun":
+        return cls(
+            step_id=data["step_id"],
+            operation_id=data.get("operation_id"),
+            status=WorkflowStepRunStatus(data.get("status", WorkflowStepRunStatus.QUEUED.value)),
+            transformation_run_id=data.get("transformation_run_id"),
+            output_asset_ids=tuple(data.get("output_asset_ids", ())),
+            diagnostics=tuple(data.get("diagnostics", ())),
+            started_at=data.get("started_at"),
+            completed_at=data.get("completed_at"),
+            updated_at=data["updated_at"],
+        )
+
+
+@dataclass(frozen=True)
+class WorkflowRun:
+    template_id: str
+    template_version: str
+    input_bindings: dict[str, Any]
+    actor_id: str
+    correlation_id: str
+    policy_context: dict[str, Any] = field(default_factory=dict)
+    status: WorkflowRunStatus = WorkflowRunStatus.QUEUED
+    step_runs: tuple[WorkflowStepRun, ...] = ()
+    output_asset_ids: tuple[str, ...] = ()
+    diagnostics: tuple[dict[str, Any], ...] = ()
+    retry_of_run_id: str | None = None
+    attempt: int = 1
+    run_id: str = field(default_factory=lambda: new_id("wfrun"))
+    queued_at: str = field(default_factory=lambda: utc_now().isoformat())
+    started_at: str | None = None
+    completed_at: str | None = None
+    updated_at: str = field(default_factory=lambda: utc_now().isoformat())
+
+    def __post_init__(self) -> None:
+        object.__setattr__(self, "status", WorkflowRunStatus(self.status))
+        object.__setattr__(self, "step_runs", tuple(self.step_runs))
+        object.__setattr__(self, "output_asset_ids", tuple(self.output_asset_ids))
+        object.__setattr__(self, "diagnostics", tuple(self.diagnostics))
+
+    def running(self) -> "WorkflowRun":
+        now = utc_now().isoformat()
+        return replace(
+            self,
+            status=WorkflowRunStatus.RUNNING,
+            started_at=self.started_at or now,
+            updated_at=now,
+        )
+
+    def waiting(self, diagnostics: tuple[dict[str, Any], ...] = ()) -> "WorkflowRun":
+        return replace(
+            self,
+            status=WorkflowRunStatus.WAITING,
+            diagnostics=self.diagnostics + tuple(diagnostics),
+            updated_at=utc_now().isoformat(),
+        )
+
+    def completed(self, *, output_asset_ids: tuple[str, ...] = ()) -> "WorkflowRun":
+        now = utc_now().isoformat()
+        return replace(
+            self,
+            status=WorkflowRunStatus.COMPLETED,
+            output_asset_ids=tuple(output_asset_ids),
+            completed_at=now,
+            updated_at=now,
+        )
+
+    def partially_completed(
+        self,
+        *,
+        output_asset_ids: tuple[str, ...] = (),
+        diagnostics: tuple[dict[str, Any], ...] = (),
+    ) -> "WorkflowRun":
+        now = utc_now().isoformat()
+        return replace(
+            self,
+            status=WorkflowRunStatus.PARTIALLY_COMPLETED,
+            output_asset_ids=tuple(output_asset_ids),
+            diagnostics=self.diagnostics + tuple(diagnostics),
+            completed_at=now,
+            updated_at=now,
+        )
+
+    def failed(self, diagnostics: tuple[dict[str, Any], ...]) -> "WorkflowRun":
+        now = utc_now().isoformat()
+        return replace(
+            self,
+            status=WorkflowRunStatus.FAILED,
+            diagnostics=self.diagnostics + tuple(diagnostics),
+            completed_at=now,
+            updated_at=now,
+        )
+
+    def canceled(self, diagnostics: tuple[dict[str, Any], ...] = ()) -> "WorkflowRun":
+        now = utc_now().isoformat()
+        return replace(
+            self,
+            status=WorkflowRunStatus.CANCELED,
+            diagnostics=self.diagnostics + tuple(diagnostics),
+            completed_at=now,
+            updated_at=now,
+        )
+
+    def retried(self) -> "WorkflowRun":
+        return replace(
+            self,
+            status=WorkflowRunStatus.RETRIED,
+            updated_at=utc_now().isoformat(),
+        )
+
+    def retry(self, *, actor_id: str, correlation_id: str) -> "WorkflowRun":
+        return WorkflowRun(
+            template_id=self.template_id,
+            template_version=self.template_version,
+            input_bindings=dict(self.input_bindings),
+            actor_id=actor_id,
+            correlation_id=correlation_id,
+            policy_context=dict(self.policy_context),
+            retry_of_run_id=self.run_id,
+            attempt=self.attempt + 1,
+        )
+
+    def with_step_run(self, step_run: WorkflowStepRun) -> "WorkflowRun":
+        existing = {item.step_id: item for item in self.step_runs}
+        existing[step_run.step_id] = step_run
+        ordered = tuple(existing[item.step_id] for item in self.step_runs if item.step_id in existing)
+        if step_run.step_id not in {item.step_id for item in self.step_runs}:
+            ordered = ordered + (step_run,)
+        return replace(self, step_runs=ordered, updated_at=utc_now().isoformat())
+
+    def to_dict(self) -> dict[str, Any]:
+        return compact_dict(
+            {
+                "run_id": self.run_id,
+                "template_id": self.template_id,
+                "template_version": self.template_version,
+                "input_bindings": dict(self.input_bindings),
+                "actor_id": self.actor_id,
+                "correlation_id": self.correlation_id,
+                "policy_context": dict(self.policy_context),
+                "status": self.status.value,
+                "step_runs": [item.to_dict() for item in self.step_runs],
+                "output_asset_ids": list(self.output_asset_ids),
+                "diagnostics": list(self.diagnostics),
+                "retry_of_run_id": self.retry_of_run_id,
+                "attempt": self.attempt,
+                "queued_at": self.queued_at,
+                "started_at": self.started_at,
+                "completed_at": self.completed_at,
+                "updated_at": self.updated_at,
+            }
+        )
+
+    @classmethod
+    def from_dict(cls, data: dict[str, Any]) -> "WorkflowRun":
+        return cls(
+            run_id=data["run_id"],
+            template_id=data["template_id"],
+            template_version=data.get("template_version", "1"),
+            input_bindings=dict(data.get("input_bindings", {})),
+            actor_id=data["actor_id"],
+            correlation_id=data["correlation_id"],
+            policy_context=dict(data.get("policy_context", {})),
+            status=WorkflowRunStatus(data.get("status", WorkflowRunStatus.QUEUED.value)),
+            step_runs=tuple(WorkflowStepRun.from_dict(item) for item in data.get("step_runs", ())),
+            output_asset_ids=tuple(data.get("output_asset_ids", ())),
+            diagnostics=tuple(data.get("diagnostics", ())),
+            retry_of_run_id=data.get("retry_of_run_id"),
+            attempt=int(data.get("attempt", 1)),
+            queued_at=data["queued_at"],
+            started_at=data.get("started_at"),
+            completed_at=data.get("completed_at"),
+            updated_at=data["updated_at"],
+        )

src/kontextual_engine/ports/repositories.py

@@ -25,6 +25,9 @@ from kontextual_engine.core import (
     Sensitivity,
     TransformationRun,
     TransformationRunStatus,
+    WorkflowRun,
+    WorkflowRunStatus,
+    WorkflowTemplate,
 )
 
 
@@ -131,3 +134,25 @@ class AssetRegistryRepository(Protocol):
         source_asset_id: str | None = None,
         transformation_run_id: str | None = None,
     ) -> list[DerivedArtifactLineage]: ...
+
+    def save_workflow_template(self, template: WorkflowTemplate) -> WorkflowTemplate: ...
+    def get_workflow_template(
+        self,
+        template_id: str,
+        *,
+        version: str | None = None,
+    ) -> WorkflowTemplate: ...
+    def list_workflow_templates(
+        self,
+        *,
+        template_id: str | None = None,
+    ) -> list[WorkflowTemplate]: ...
+
+    def save_workflow_run(self, run: WorkflowRun) -> WorkflowRun: ...
+    def get_workflow_run(self, run_id: str) -> WorkflowRun: ...
+    def list_workflow_runs(
+        self,
+        *,
+        status: WorkflowRunStatus | None = None,
+        template_id: str | None = None,
+    ) -> list[WorkflowRun]: ...

src/kontextual_engine/services/__init__.py

@@ -32,6 +32,7 @@ from .transformation_service import (
     TransformationService,
     default_transformation_registry,
 )
+from .workflow_service import WorkflowInvocation, WorkflowRunResult, WorkflowService
 
 __all__ = [
     "AssetChangeResult",
@@ -60,5 +61,8 @@ __all__ = [
     "TransformationRequest",
     "TransformationRunResult",
     "TransformationService",
+    "WorkflowInvocation",
+    "WorkflowRunResult",
+    "WorkflowService",
     "default_transformation_registry",
 ]

src/kontextual_engine/services/workflow_service.py

@@ -0,0 +1,853 @@
+"""Durable workflow templates and MVP job runner."""
+
+from __future__ import annotations
+
+from dataclasses import dataclass, field, replace
+from typing import Any
+
+from kontextual_engine.core import (
+    AuditEvent,
+    AuditOutcome,
+    OperationContext,
+    PolicyDecision,
+    WorkflowRun,
+    WorkflowRunStatus,
+    WorkflowStepDefinition,
+    WorkflowStepRun,
+    WorkflowStepRunStatus,
+    WorkflowTemplate,
+)
+from kontextual_engine.errors import AuthorizationError, Diagnostic, NotFoundError, ValidationError
+from kontextual_engine.ports import AllowAllPolicyGateway, AssetRegistryRepository, PolicyGateway
+
+from .transformation_service import TransformationRequest, TransformationRunResult, TransformationService
+
+
+@dataclass(frozen=True)
+class WorkflowInvocation:
+    template_id: str
+    inputs: dict[str, Any] = field(default_factory=dict)
+    template_version: str | None = None
+    metadata: dict[str, Any] = field(default_factory=dict)
+
+    def to_dict(self) -> dict[str, Any]:
+        return {
+            "template_id": self.template_id,
+            "template_version": self.template_version,
+            "inputs": dict(self.inputs),
+            "metadata": dict(self.metadata),
+        }
+
+
+@dataclass(frozen=True)
+class WorkflowRunResult:
+    run: WorkflowRun
+    success: bool
+    diagnostics: tuple[Diagnostic, ...] = ()
+    audit_event: AuditEvent | None = None
+    policy_decision: PolicyDecision | None = None
+    step_results: tuple[TransformationRunResult, ...] = ()
+
+    def __post_init__(self) -> None:
+        object.__setattr__(self, "diagnostics", tuple(self.diagnostics))
+        object.__setattr__(self, "step_results", tuple(self.step_results))
+
+    def to_dict(self) -> dict[str, Any]:
+        return {
+            "success": self.success,
+            "run": self.run.to_dict(),
+            "diagnostics": [item.to_dict() for item in self.diagnostics],
+            "audit_event": self.audit_event.to_dict() if self.audit_event else None,
+            "policy_decision": self.policy_decision.to_dict() if self.policy_decision else None,
+            "step_results": [item.to_dict() for item in self.step_results],
+        }
+
+
+class WorkflowService:
+    def __init__(
+        self,
+        repository: AssetRegistryRepository,
+        *,
+        transformation_service: TransformationService | None = None,
+        policy_gateway: PolicyGateway | None = None,
+    ) -> None:
+        self.repository = repository
+        self.policy_gateway = policy_gateway or AllowAllPolicyGateway()
+        self.transformation_service = transformation_service or TransformationService(
+            repository,
+            policy_gateway=self.policy_gateway,
+        )
+
+    def register_template(self, template: WorkflowTemplate, context: OperationContext) -> WorkflowTemplate:
+        self.repository.save_actor(context.actor)
+        diagnostics = _template_diagnostics(template)
+        if diagnostics:
+            raise ValidationError(
+                "Workflow template is invalid",
+                details={"diagnostics": [item.to_dict() for item in diagnostics]},
+            )
+        template = template.with_actor(context.actor.id)
+        decision = self._authorize(
+            context,
+            "workflow.template.register",
+            f"workflow_template:{template.template_id}",
+            resource_metadata={"template": template.to_dict()},
+        )
+        if not decision.allowed:
+            self._audit(
+                "workflow.template.register",
+                f"workflow_template:{template.template_id}",
+                AuditOutcome.DENIED,
+                context,
+                decision,
+            )
+            raise AuthorizationError(
+                "Operation denied by policy",
+                details={
+                    "action": "workflow.template.register",
+                    "resource": f"workflow_template:{template.template_id}",
+                    "correlation_id": context.correlation_id,
+                    "policy_decision": decision.to_dict(),
+                },
+            )
+        saved = self.repository.save_workflow_template(template)
+        self._audit(
+            "workflow.template.register",
+            f"workflow_template:{template.template_id}",
+            AuditOutcome.SUCCESS,
+            context,
+            decision,
+            details={"template_version": template.version, "template_hash": template.template_hash},
+        )
+        return saved
+
+    def get_template(self, template_id: str, *, version: str | None = None) -> WorkflowTemplate:
+        return self.repository.get_workflow_template(template_id, version=version)
+
+    def list_templates(self, *, template_id: str | None = None) -> tuple[WorkflowTemplate, ...]:
+        return tuple(self.repository.list_workflow_templates(template_id=template_id))
+
+    def queue_template(self, invocation: WorkflowInvocation, context: OperationContext) -> WorkflowRunResult:
+        template = self.repository.get_workflow_template(
+            invocation.template_id,
+            version=invocation.template_version,
+        )
+        self.repository.save_actor(context.actor)
+        decision = self._authorize(
+            context,
+            "workflow.run.execute",
+            f"workflow_template:{template.template_id}",
+            resource_metadata={"template": template.to_dict(), "invocation": invocation.to_dict()},
+        )
+        run = WorkflowRun(
+            template_id=template.template_id,
+            template_version=template.version,
+            input_bindings=dict(invocation.inputs),
+            actor_id=context.actor.id,
+            correlation_id=context.correlation_id,
+            policy_context={"run_execute": decision.to_dict()},
+            step_runs=_initial_step_runs(template),
+        )
+        self.repository.save_workflow_run(run)
+        self._audit(
+            "workflow.run.queued",
+            f"workflow_run:{run.run_id}",
+            AuditOutcome.SUCCESS,
+            context,
+            decision,
+            details={"template_id": template.template_id, "template_version": template.version},
+        )
+        if not decision.allowed:
+            diagnostic = _permission_diagnostic(decision)
+            failed = run.failed((_diagnostic_dict(diagnostic),))
+            self.repository.save_workflow_run(failed)
+            event = self._audit(
+                "workflow.run.execute",
+                f"workflow_run:{run.run_id}",
+                AuditOutcome.DENIED,
+                context,
+                decision,
+                details={"template_id": template.template_id},
+            )
+            return WorkflowRunResult(
+                run=failed,
+                success=False,
+                diagnostics=(diagnostic,),
+                audit_event=event,
+                policy_decision=decision,
+            )
+
+        input_diagnostics = _input_diagnostics(template, invocation)
+        if input_diagnostics:
+            failed = run.failed(tuple(_diagnostic_dict(item) for item in input_diagnostics))
+            self.repository.save_workflow_run(failed)
+            event = self._audit(
+                "workflow.run.failed",
+                f"workflow_run:{run.run_id}",
+                AuditOutcome.FAILED,
+                context,
+                decision,
+                details={"diagnostics": [item.to_dict() for item in input_diagnostics]},
+            )
+            return WorkflowRunResult(
+                run=failed,
+                success=False,
+                diagnostics=tuple(input_diagnostics),
+                audit_event=event,
+                policy_decision=decision,
+            )
+
+        return WorkflowRunResult(run=run, success=True, policy_decision=decision)
+
+    def invoke_template(self, invocation: WorkflowInvocation, context: OperationContext) -> WorkflowRunResult:
+        queued = self.queue_template(invocation, context)
+        if not queued.success:
+            return queued
+        return self.resume_run(queued.run.run_id, context)
+
+    def resume_run(self, run_id: str, context: OperationContext) -> WorkflowRunResult:
+        run = self.repository.get_workflow_run(run_id)
+        if run.status in (
+            WorkflowRunStatus.COMPLETED,
+            WorkflowRunStatus.CANCELED,
+            WorkflowRunStatus.RETRIED,
+        ):
+            diagnostic = Diagnostic(
+                severity="error",
+                code="workflow.run_not_resumable",
+                message="Workflow run cannot be resumed from its current status",
+                details={"run_id": run_id, "status": run.status.value},
+            )
+            return WorkflowRunResult(run=run, success=False, diagnostics=(diagnostic,))
+        template = self.repository.get_workflow_template(run.template_id, version=run.template_version)
+        return self._execute_run(template, run, context)
+
+    def retry_run(self, run_id: str, context: OperationContext) -> WorkflowRunResult:
+        previous = self.repository.get_workflow_run(run_id)
+        marked = previous.retried()
+        self.repository.save_workflow_run(marked)
+        retry = previous.retry(actor_id=context.actor.id, correlation_id=context.correlation_id)
+        retry = replace(retry, step_runs=tuple(_reset_step_run(item) for item in previous.step_runs))
+        self.repository.save_actor(context.actor)
+        self.repository.save_workflow_run(retry)
+        decision = PolicyDecision.allow(
+            context.actor.id,
+            "workflow.run.retry",
+            f"workflow_run:{run_id}",
+            context={"previous_run_id": run_id, "retry_run_id": retry.run_id},
+        )
+        self._audit(
+            "workflow.run.retried",
+            f"workflow_run:{run_id}",
+            AuditOutcome.SUCCESS,
+            context,
+            decision,
+            details={"retry_run_id": retry.run_id},
+        )
+        return self.resume_run(retry.run_id, context)
+
+    def cancel_run(self, run_id: str, context: OperationContext, *, reason: str | None = None) -> WorkflowRun:
+        run = self.repository.get_workflow_run(run_id)
+        diagnostic = Diagnostic(
+            severity="warning",
+            code="workflow.run_canceled",
+            message="Workflow run was canceled",
+            details={"reason": reason} if reason else {},
+        )
+        canceled_steps = tuple(
+            step.canceled((_diagnostic_dict(diagnostic),))
+            if step.status in (WorkflowStepRunStatus.QUEUED, WorkflowStepRunStatus.WAITING)
+            else step
+            for step in run.step_runs
+        )
+        canceled = replace(run, step_runs=canceled_steps).canceled((_diagnostic_dict(diagnostic),))
+        self.repository.save_actor(context.actor)
+        decision = self._authorize(
+            context,
+            "workflow.run.cancel",
+            f"workflow_run:{run_id}",
+            resource_metadata={"reason": reason, "status": run.status.value},
+        )
+        if not decision.allowed:
+            self._audit(
+                "workflow.run.canceled",
+                f"workflow_run:{run_id}",
+                AuditOutcome.DENIED,
+                context,
+                decision,
+                details={"reason": reason} if reason else {},
+            )
+            raise AuthorizationError(
+                "Operation denied by policy",
+                details={
+                    "action": "workflow.run.cancel",
+                    "resource": f"workflow_run:{run_id}",
+                    "correlation_id": context.correlation_id,
+                    "policy_decision": decision.to_dict(),
+                },
+            )
+        self.repository.save_workflow_run(canceled)
+        self._audit(
+            "workflow.run.canceled",
+            f"workflow_run:{run_id}",
+            AuditOutcome.SUCCESS,
+            context,
+            decision,
+            details={"reason": reason} if reason else {},
+        )
+        return canceled
+
+    def _execute_run(
+        self,
+        template: WorkflowTemplate,
+        run: WorkflowRun,
+        context: OperationContext,
+    ) -> WorkflowRunResult:
+        decision = PolicyDecision.from_dict(run.policy_context["run_execute"])
+        run = run.running()
+        self.repository.save_workflow_run(run)
+        self._audit(
+            "workflow.run.started",
+            f"workflow_run:{run.run_id}",
+            AuditOutcome.SUCCESS,
+            context,
+            decision,
+            details={"template_id": template.template_id, "template_version": template.version},
+        )
+
+        step_results: list[TransformationRunResult] = []
+        attempted_step_ids: set[str] = set()
+        progress = True
+        while progress:
+            progress = False
+            step_runs_by_id = {item.step_id: item for item in run.step_runs}
+            for step in template.steps:
+                current = step_runs_by_id[step.step_id]
+                if current.status in (
+                    WorkflowStepRunStatus.COMPLETED,
+                    WorkflowStepRunStatus.SKIPPED,
+                    WorkflowStepRunStatus.CANCELED,
+                ):
+                    continue
+                if current.status == WorkflowStepRunStatus.FAILED and current.step_id in attempted_step_ids:
+                    continue
+                dependency_diagnostic = _dependency_diagnostic(step, step_runs_by_id)
+                if dependency_diagnostic is not None:
+                    if _dependency_terminal(step, step_runs_by_id):
+                        current = _step_failure_by_behavior(step, current, dependency_diagnostic)
+                        run = run.with_step_run(current)
+                        self.repository.save_workflow_run(run)
+                        progress = True
+                    continue
+
+                current = current.running()
+                run = run.with_step_run(current)
+                self.repository.save_workflow_run(run)
+                self._audit(
+                    "workflow.step.started",
+                    f"workflow_run:{run.run_id}:step:{step.step_id}",
+                    AuditOutcome.SUCCESS,
+                    context,
+                    decision,
+                    details={"operation_id": step.operation_id, "kind": step.kind},
+                )
+                attempted_step_ids.add(step.step_id)
+                executed, step_result = self._execute_step(step, run, context)
+                if step_result is not None:
+                    step_results.append(step_result)
+                if executed.status == WorkflowStepRunStatus.FAILED and step.failure_behavior == "continue":
+                    executed = replace(executed, status=WorkflowStepRunStatus.SKIPPED)
+                run = run.with_step_run(executed)
+                self.repository.save_workflow_run(run)
+                self._audit(
+                    "workflow.step.completed" if executed.status == WorkflowStepRunStatus.COMPLETED else "workflow.step.failed",
+                    f"workflow_run:{run.run_id}:step:{step.step_id}",
+                    AuditOutcome.SUCCESS if executed.status == WorkflowStepRunStatus.COMPLETED else AuditOutcome.FAILED,
+                    context,
+                    decision,
+                    details={
+                        "operation_id": step.operation_id,
+                        "status": executed.status.value,
+                        "diagnostics": list(executed.diagnostics),
+                    },
+                )
+                progress = True
+                if executed.status == WorkflowStepRunStatus.FAILED and step.failure_behavior != "continue":
+                    progress = False
+                    break
+
+        final = _finalize_run(run)
+        self.repository.save_workflow_run(final)
+        event = self._audit(
+            _workflow_completion_operation(final),
+            f"workflow_run:{final.run_id}",
+            _workflow_audit_outcome(final),
+            context,
+            decision,
+            details={
+                "template_id": template.template_id,
+                "status": final.status.value,
+                "output_asset_ids": list(final.output_asset_ids),
+                "diagnostics": list(final.diagnostics),
+            },
+        )
+        return WorkflowRunResult(
+            run=final,
+            success=final.status == WorkflowRunStatus.COMPLETED,
+            diagnostics=tuple(Diagnostic(**item) for item in final.diagnostics),
+            audit_event=event,
+            policy_decision=decision,
+            step_results=tuple(step_results),
+        )
+
+    def _execute_step(
+        self,
+        step: WorkflowStepDefinition,
+        run: WorkflowRun,
+        context: OperationContext,
+    ) -> tuple[WorkflowStepRun, TransformationRunResult | None]:
+        current = {item.step_id: item for item in run.step_runs}[step.step_id]
+        precondition_diagnostics = _precondition_diagnostics(step, run)
+        if precondition_diagnostics:
+            return current.failed(tuple(_diagnostic_dict(item) for item in precondition_diagnostics)), None
+        if step.kind != "transformation":
+            diagnostic = Diagnostic(
+                severity="error",
+                code="workflow.step_kind_unsupported",
+                message="Workflow step kind is not executable by the MVP runner",
+                details={"step_id": step.step_id, "kind": step.kind},
+            )
+            return current.failed((_diagnostic_dict(diagnostic),)), None
+        if not step.operation_id:
+            diagnostic = Diagnostic(
+                severity="error",
+                code="workflow.operation_missing",
+                message="Transformation workflow step requires an operation ID",
+                details={"step_id": step.step_id},
+            )
+            return current.failed((_diagnostic_dict(diagnostic),)), None
+
+        source_asset_ids, resolve_diagnostics = _resolve_source_asset_ids(step, run)
+        if resolve_diagnostics:
+            return current.failed(tuple(_diagnostic_dict(item) for item in resolve_diagnostics)), None
+
+        result = self.transformation_service.execute_transformation(
+            TransformationRequest(
+                operation_id=step.operation_id,
+                source_asset_ids=source_asset_ids,
+                parameters=dict(step.parameters),
+                output_asset_id=self._available_output_asset_id(
+                    _resolve_value(step.outputs.get("asset_id"), run),
+                    run,
+                ),
+                output_title=_resolve_value(step.outputs.get("title"), run),
+                output_asset_type=str(step.outputs.get("asset_type", "derived_artifact")),
+                output_media_type=step.outputs.get("media_type"),
+                metadata=dict(step.metadata),
+            ),
+            context,
+        )
+        if result.success and result.run is not None:
+            return (
+                current.completed(
+                    transformation_run_id=result.run.run_id,
+                    output_asset_ids=result.run.output_asset_ids,
+                ),
+                result,
+            )
+        diagnostics = result.diagnostics or (
+            Diagnostic(
+                severity="error",
+                code="workflow.transformation_failed",
+                message="Transformation step failed without diagnostics",
+                details={"step_id": step.step_id, "operation_id": step.operation_id},
+            ),
+        )
+        transformation_run_id = result.run.run_id if result.run else None
+        return (
+            replace(current, transformation_run_id=transformation_run_id).failed(
+                tuple(_diagnostic_dict(item) for item in diagnostics)
+            ),
+            result,
+        )
+
+    def _available_output_asset_id(self, output_asset_id: Any, run: WorkflowRun) -> str | None:
+        if output_asset_id is None:
+            return None
+        output_asset_id = str(output_asset_id)
+        try:
+            self.repository.get_asset(output_asset_id)
+        except NotFoundError:
+            return output_asset_id
+        return f"{output_asset_id}-{run.run_id}"
+
+    def _authorize(
+        self,
+        context: OperationContext,
+        action: str,
+        resource: str,
+        *,
+        resource_metadata: dict[str, Any] | None = None,
+    ) -> PolicyDecision:
+        self.repository.save_actor(context.actor)
+        try:
+            return self.policy_gateway.authorize(
+                context,
+                action,
+                resource,
+                resource_metadata=resource_metadata,
+            )
+        except Exception as exc:
+            return PolicyDecision.fail_closed(
+                context.actor.id,
+                action,
+                resource,
+                reason=str(exc) or "Workflow policy gateway failed",
+                context={"resource_metadata": resource_metadata or {}, "gateway_error": type(exc).__name__},
+            )
+
+    def _audit(
+        self,
+        operation: str,
+        target: str,
+        outcome: AuditOutcome,
+        context: OperationContext,
+        policy_decision: PolicyDecision,
+        *,
+        details: dict[str, Any] | None = None,
+    ) -> AuditEvent:
+        event = AuditEvent.from_context(
+            operation,
+            target,
+            outcome,
+            context,
+            policy_decision=policy_decision,
+            details=details,
+        )
+        return self.repository.save_audit_event(event)
+
+
+def _initial_step_runs(template: WorkflowTemplate) -> tuple[WorkflowStepRun, ...]:
+    return tuple(
+        WorkflowStepRun(step_id=step.step_id, operation_id=step.operation_id)
+        for step in template.steps
+    )
+
+
+def _reset_step_run(step_run: WorkflowStepRun) -> WorkflowStepRun:
+    return WorkflowStepRun(step_id=step_run.step_id, operation_id=step_run.operation_id)
+
+
+def _template_diagnostics(template: WorkflowTemplate) -> tuple[Diagnostic, ...]:
+    diagnostics: list[Diagnostic] = []
+    step_ids = [step.step_id for step in template.steps]
+    duplicate_ids = sorted({step_id for step_id in step_ids if step_ids.count(step_id) > 1})
+    for step_id in duplicate_ids:
+        diagnostics.append(
+            Diagnostic(
+                severity="error",
+                code="workflow.step_id_duplicate",
+                message="Workflow template contains duplicate step IDs",
+                details={"template_id": template.template_id, "step_id": step_id},
+            )
+        )
+    known = set(step_ids)
+    for step in template.steps:
+        if step.failure_behavior not in ("fail_workflow", "continue"):
+            diagnostics.append(
+                Diagnostic(
+                    severity="error",
+                    code="workflow.failure_behavior_unsupported",
+                    message="Workflow step declares unsupported failure behavior",
+                    details={"step_id": step.step_id, "failure_behavior": step.failure_behavior},
+                )
+            )
+        for dependency in step.depends_on:
+            if dependency not in known:
+                diagnostics.append(
+                    Diagnostic(
+                        severity="error",
+                        code="workflow.dependency_missing",
+                        message="Workflow step depends on an unknown step",
+                        details={"step_id": step.step_id, "dependency": dependency},
+                    )
+                )
+    cycle = _dependency_cycle(template.steps)
+    if cycle:
+        diagnostics.append(
+            Diagnostic(
+                severity="error",
+                code="workflow.dependency_cycle",
+                message="Workflow template contains a dependency cycle",
+                details={"cycle": cycle},
+            )
+        )
+    return tuple(diagnostics)
+
+
+def _dependency_cycle(steps: tuple[WorkflowStepDefinition, ...]) -> list[str]:
+    graph = {step.step_id: tuple(step.depends_on) for step in steps}
+    visiting: set[str] = set()
+    visited: set[str] = set()
+    stack: list[str] = []
+
+    def visit(step_id: str) -> list[str] | None:
+        if step_id in visited:
+            return None
+        if step_id in visiting:
+            if step_id in stack:
+                return stack[stack.index(step_id):] + [step_id]
+            return [step_id]
+        visiting.add(step_id)
+        stack.append(step_id)
+        for dependency in graph.get(step_id, ()):
+            result = visit(dependency)
+            if result:
+                return result
+        visiting.remove(step_id)
+        visited.add(step_id)
+        stack.pop()
+        return None
+
+    for step_id in graph:
+        result = visit(step_id)
+        if result:
+            return result
+    return []
+
+
+def _input_diagnostics(template: WorkflowTemplate, invocation: WorkflowInvocation) -> list[Diagnostic]:
+    diagnostics: list[Diagnostic] = []
+    specs = {item.name: item for item in template.inputs}
+    for spec in template.inputs:
+        if spec.required and spec.name not in invocation.inputs:
+            diagnostics.append(
+                Diagnostic(
+                    severity="error",
+                    code="workflow.input_missing",
+                    message="Workflow invocation is missing a required input",
+                    details={"template_id": template.template_id, "input": spec.name, "kind": spec.kind.value},
+                )
+            )
+    for name, binding in invocation.inputs.items():
+        spec = specs.get(name)
+        declared_kind = binding.get("kind") if isinstance(binding, dict) else None
+        if spec is not None and declared_kind is not None and declared_kind != spec.kind.value:
+            diagnostics.append(
+                Diagnostic(
+                    severity="error",
+                    code="workflow.input_kind_mismatch",
+                    message="Workflow invocation input kind does not match the template declaration",
+                    details={
+                        "template_id": template.template_id,
+                        "input": name,
+                        "expected": spec.kind.value,
+                        "actual": declared_kind,
+                    },
+                )
+            )
+    return diagnostics
+
+
+def _dependency_diagnostic(
+    step: WorkflowStepDefinition,
+    step_runs_by_id: dict[str, WorkflowStepRun],
+) -> Diagnostic | None:
+    for dependency in step.depends_on:
+        dependency_run = step_runs_by_id[dependency]
+        if dependency_run.status != WorkflowStepRunStatus.COMPLETED:
+            return Diagnostic(
+                severity="error",
+                code="workflow.dependency_not_completed",
+                message="Workflow step dependency is not completed",
+                details={
+                    "step_id": step.step_id,
+                    "dependency": dependency,
+                    "dependency_status": dependency_run.status.value,
+                },
+            )
+    return None
+
+
+def _dependency_terminal(
+    step: WorkflowStepDefinition,
+    step_runs_by_id: dict[str, WorkflowStepRun],
+) -> bool:
+    return any(
+        step_runs_by_id[dependency].status
+        in (
+            WorkflowStepRunStatus.FAILED,
+            WorkflowStepRunStatus.SKIPPED,
+            WorkflowStepRunStatus.CANCELED,
+        )
+        for dependency in step.depends_on
+    )
+
+
+def _step_failure_by_behavior(
+    step: WorkflowStepDefinition,
+    step_run: WorkflowStepRun,
+    diagnostic: Diagnostic,
+) -> WorkflowStepRun:
+    if step.failure_behavior == "continue":
+        return step_run.skipped((_diagnostic_dict(diagnostic),))
+    return step_run.failed((_diagnostic_dict(diagnostic),))
+
+
+def _precondition_diagnostics(step: WorkflowStepDefinition, run: WorkflowRun) -> list[Diagnostic]:
+    diagnostics: list[Diagnostic] = []
+    for precondition in step.preconditions:
+        precondition_type = precondition.get("type")
+        if precondition_type == "input_present":
+            name = precondition.get("name")
+            if name not in run.input_bindings:
+                diagnostics.append(
+                    Diagnostic(
+                        severity="error",
+                        code="workflow.precondition_failed",
+                        message="Workflow step precondition failed",
+                        details={"step_id": step.step_id, "type": precondition_type, "name": name},
+                    )
+                )
+        elif precondition_type:
+            diagnostics.append(
+                Diagnostic(
+                    severity="error",
+                    code="workflow.precondition_unsupported",
+                    message="Workflow step precondition type is unsupported",
+                    details={"step_id": step.step_id, "type": precondition_type},
+                )
+            )
+    return diagnostics
+
+
+def _resolve_source_asset_ids(
+    step: WorkflowStepDefinition,
+    run: WorkflowRun,
+) -> tuple[tuple[str, ...], list[Diagnostic]]:
+    source_asset_ids = _resolve_value(step.inputs.get("source_asset_ids", ()), run)
+    try:
+        return _normalise_asset_ids(source_asset_ids), []
+    except ValueError as exc:
+        return (), [
+            Diagnostic(
+                severity="error",
+                code="workflow.source_asset_resolution_failed",
+                message="Workflow step source assets could not be resolved",
+                details={"step_id": step.step_id, "error": str(exc), "binding": source_asset_ids},
+            )
+        ]
+
+
+def _resolve_value(value: Any, run: WorkflowRun) -> Any:
+    if isinstance(value, str):
+        if value.startswith("$inputs."):
+            return run.input_bindings.get(value.removeprefix("$inputs."))
+        if value.startswith("$steps."):
+            parts = value.split(".")
+            if len(parts) == 3:
+                step_id = parts[1]
+                attr = parts[2]
+                step_run = {item.step_id: item for item in run.step_runs}.get(step_id)
+                if step_run is None:
+                    return None
+                if attr == "output_asset_ids":
+                    return step_run.output_asset_ids
+                if attr == "output_asset_id":
+                    return step_run.output_asset_ids[0] if step_run.output_asset_ids else None
+                if attr == "transformation_run_id":
+                    return step_run.transformation_run_id
+            return None
+        return value
+    if isinstance(value, list):
+        return [_resolve_value(item, run) for item in value]
+    if isinstance(value, tuple):
+        return tuple(_resolve_value(item, run) for item in value)
+    if isinstance(value, dict):
+        return {key: _resolve_value(item, run) for key, item in value.items()}
+    return value
+
+
+def _normalise_asset_ids(value: Any) -> tuple[str, ...]:
+    if value is None:
+        return ()
+    if isinstance(value, str):
+        return (value,)
+    if isinstance(value, dict):
+        if "asset_id" in value:
+            return _normalise_asset_ids(value["asset_id"])
+        if "asset_ids" in value:
+            return _normalise_asset_ids(value["asset_ids"])
+        raise ValueError("asset binding dictionary must include asset_id or asset_ids")
+    if isinstance(value, (list, tuple)):
+        asset_ids: list[str] = []
+        for item in value:
+            asset_ids.extend(_normalise_asset_ids(item))
+        return tuple(asset_ids)
+    raise ValueError(f"unsupported asset binding type: {type(value).__name__}")
+
+
+def _finalize_run(run: WorkflowRun) -> WorkflowRun:
+    output_asset_ids = tuple(
+        output_asset_id
+        for step_run in run.step_runs
+        if step_run.status == WorkflowStepRunStatus.COMPLETED
+        for output_asset_id in step_run.output_asset_ids
+    )
+    diagnostics = tuple(
+        diagnostic
+        for step_run in run.step_runs
+        if step_run.status
+        in (
+            WorkflowStepRunStatus.FAILED,
+            WorkflowStepRunStatus.SKIPPED,
+            WorkflowStepRunStatus.CANCELED,
+            WorkflowStepRunStatus.WAITING,
+        )
+        for diagnostic in step_run.diagnostics
+    )
+    statuses = {step_run.status for step_run in run.step_runs}
+    if not run.step_runs or statuses == {WorkflowStepRunStatus.COMPLETED}:
+        return run.completed(output_asset_ids=output_asset_ids)
+    if WorkflowStepRunStatus.FAILED in statuses or WorkflowStepRunStatus.CANCELED in statuses:
+        if output_asset_ids:
+            return run.partially_completed(output_asset_ids=output_asset_ids, diagnostics=diagnostics)
+        return run.failed(diagnostics)
+    if WorkflowStepRunStatus.SKIPPED in statuses:
+        if output_asset_ids:
+            return run.partially_completed(output_asset_ids=output_asset_ids, diagnostics=diagnostics)
+        return run.failed(diagnostics)
+    return run.waiting(diagnostics)
+
+
+def _workflow_completion_operation(run: WorkflowRun) -> str:
+    if run.status == WorkflowRunStatus.COMPLETED:
+        return "workflow.run.completed"
+    if run.status == WorkflowRunStatus.PARTIALLY_COMPLETED:
+        return "workflow.run.partially_completed"
+    if run.status == WorkflowRunStatus.WAITING:
+        return "workflow.run.waiting"
+    return "workflow.run.failed"
+
+
+def _workflow_audit_outcome(run: WorkflowRun) -> AuditOutcome:
+    if run.status == WorkflowRunStatus.COMPLETED:
+        return AuditOutcome.SUCCESS
+    if run.status == WorkflowRunStatus.PARTIALLY_COMPLETED:
+        return AuditOutcome.PARTIAL
+    if run.status == WorkflowRunStatus.WAITING:
+        return AuditOutcome.REVIEW_REQUIRED
+    return AuditOutcome.FAILED
+
+
+def _permission_diagnostic(decision: PolicyDecision) -> Diagnostic:
+    return Diagnostic(
+        severity="error",
+        code="workflow.permission_denied",
+        message="Workflow operation denied by policy",
+        details={"policy_decision": decision.to_dict()},
+    )
+
+
+def _diagnostic_dict(diagnostic: Diagnostic) -> dict[str, Any]:
+    return diagnostic.to_dict()

tests/test_workflow_service.py

@@ -0,0 +1,345 @@
+from pathlib import Path
+
+import pytest
+
+from kontextual_engine import (
+    Actor,
+    ActorType,
+    AssetRegistryService,
+    AssetRepresentation,
+    Classification,
+    InMemoryAssetRegistryRepository,
+    OperationContext,
+    RepresentationKind,
+    Sensitivity,
+    SQLiteAssetRegistryRepository,
+    TransformationRunStatus,
+    TransformationService,
+    ValidationError,
+    WorkflowInputDefinition,
+    WorkflowInputKind,
+    WorkflowInvocation,
+    WorkflowRunStatus,
+    WorkflowService,
+    WorkflowStepDefinition,
+    WorkflowStepRunStatus,
+    WorkflowTemplate,
+)
+
+
+def test_workflow_template_registration_persists_input_kinds_and_rejects_bad_dependencies() -> None:
+    repo = InMemoryAssetRegistryRepository()
+    service = WorkflowService(repo)
+    context = operation_context()
+
+    template = service.register_template(rich_input_template(), context)
+
+    assert template.created_by == "user-test"
+    assert [item.kind for item in repo.get_workflow_template("workflow-rich-inputs").inputs] == [
+        WorkflowInputKind.ASSET,
+        WorkflowInputKind.COLLECTION,
+        WorkflowInputKind.QUERY,
+        WorkflowInputKind.SOURCE_EVENT,
+        WorkflowInputKind.PAYLOAD,
+    ]
+    assert repo.list_audit_events(target="workflow_template:workflow-rich-inputs")[0].operation == (
+        "workflow.template.register"
+    )
+
+    with pytest.raises(ValidationError) as missing_dependency:
+        service.register_template(
+            WorkflowTemplate(
+                template_id="workflow-bad",
+                name="Bad Workflow",
+                steps=(
+                    WorkflowStepDefinition(
+                        step_id="late",
+                        operation_id="structured_view",
+                        depends_on=("missing",),
+                    ),
+                ),
+            ),
+            context,
+        )
+    assert missing_dependency.value.details["diagnostics"][0]["code"] == "workflow.dependency_missing"
+
+    with pytest.raises(ValidationError) as cycle:
+        service.register_template(
+            WorkflowTemplate(
+                template_id="workflow-cycle",
+                name="Cycle",
+                steps=(
+                    WorkflowStepDefinition(
+                        step_id="a",
+                        operation_id="structured_view",
+                        depends_on=("b",),
+                    ),
+                    WorkflowStepDefinition(
+                        step_id="b",
+                        operation_id="structured_view",
+                        depends_on=("a",),
+                    ),
+                ),
+            ),
+            context,
+        )
+    assert cycle.value.details["diagnostics"][0]["code"] == "workflow.dependency_cycle"
+
+
+def test_workflow_invocation_executes_dependent_transformations_in_order() -> None:
+    repo = InMemoryAssetRegistryRepository()
+    registry = AssetRegistryService(repo)
+    context = operation_context()
+    create_source_asset(registry, context, asset_id="asset-source")
+    workflow = WorkflowService(
+        repo,
+        transformation_service=TransformationService(repo, asset_service=registry),
+    )
+    workflow.register_template(dependent_transformation_template(), context)
+
+    result = workflow.invoke_template(
+        WorkflowInvocation(
+            template_id="workflow-dependent",
+            inputs={"source": {"kind": "asset", "asset_id": "asset-source"}},
+        ),
+        context,
+    )
+
+    assert result.success is True
+    assert result.run.status == WorkflowRunStatus.COMPLETED
+    step_runs = {step.step_id: step for step in result.run.step_runs}
+    assert step_runs["view"].status == WorkflowStepRunStatus.COMPLETED
+    assert step_runs["view_again"].status == WorkflowStepRunStatus.COMPLETED
+    assert result.run.output_asset_ids == ("asset-view-1", "asset-view-2")
+    first_transformation = repo.get_transformation_run(step_runs["view"].transformation_run_id)
+    second_transformation = repo.get_transformation_run(step_runs["view_again"].transformation_run_id)
+    assert first_transformation.source_asset_ids == ("asset-source",)
+    assert second_transformation.source_asset_ids == ("asset-view-1",)
+    assert second_transformation.status == TransformationRunStatus.COMPLETED
+    assert [
+        event.operation
+        for event in repo.list_audit_events(target=f"workflow_run:{result.run.run_id}")
+    ] == [
+        "workflow.run.queued",
+        "workflow.run.started",
+        "workflow.run.completed",
+    ]
+
+
+def test_workflow_queue_cancel_resume_and_retry_do_not_require_storage_edits() -> None:
+    repo = InMemoryAssetRegistryRepository()
+    registry = AssetRegistryService(repo)
+    context = operation_context()
+    create_source_asset(registry, context, asset_id="asset-source")
+    workflow = WorkflowService(
+        repo,
+        transformation_service=TransformationService(repo, asset_service=registry),
+    )
+    workflow.register_template(single_step_template(), context)
+
+    queued = workflow.queue_template(
+        WorkflowInvocation(
+            template_id="workflow-single",
+            inputs={"source": {"kind": "asset", "asset_id": "asset-source"}},
+        ),
+        context,
+    )
+    canceled = workflow.cancel_run(queued.run.run_id, context, reason="operator pause")
+    resumed = workflow.resume_run(canceled.run_id, context)
+    completed = workflow.invoke_template(
+        WorkflowInvocation(
+            template_id="workflow-single",
+            inputs={"source": {"kind": "asset", "asset_id": "asset-source"}},
+        ),
+        context,
+    )
+    retry = workflow.retry_run(completed.run.run_id, context)
+
+    assert queued.run.status == WorkflowRunStatus.QUEUED
+    assert canceled.status == WorkflowRunStatus.CANCELED
+    assert {step.status for step in canceled.step_runs} == {WorkflowStepRunStatus.CANCELED}
+    assert resumed.success is False
+    assert resumed.diagnostics[0].code == "workflow.run_not_resumable"
+    assert repo.get_workflow_run(completed.run.run_id).status == WorkflowRunStatus.RETRIED
+    assert retry.run.retry_of_run_id == completed.run.run_id
+    assert retry.run.attempt == 2
+    assert retry.run.status == WorkflowRunStatus.COMPLETED
+
+
+def test_workflow_continue_failure_finishes_partially_completed() -> None:
+    repo = InMemoryAssetRegistryRepository()
+    registry = AssetRegistryService(repo)
+    context = operation_context()
+    create_source_asset(registry, context, asset_id="asset-source")
+    workflow = WorkflowService(
+        repo,
+        transformation_service=TransformationService(repo, asset_service=registry),
+    )
+    workflow.register_template(partial_template(), context)
+
+    result = workflow.invoke_template(
+        WorkflowInvocation(
+            template_id="workflow-partial",
+            inputs={"source": {"kind": "asset", "asset_id": "asset-source"}},
+        ),
+        context,
+    )
+
+    step_runs = {step.step_id: step for step in result.run.step_runs}
+    assert result.success is False
+    assert result.run.status == WorkflowRunStatus.PARTIALLY_COMPLETED
+    assert result.run.output_asset_ids == ("asset-view-success",)
+    assert step_runs["markdown"].status == WorkflowStepRunStatus.SKIPPED
+    assert step_runs["view"].status == WorkflowStepRunStatus.COMPLETED
+    assert result.diagnostics[0].code == "transformation.operation_not_executable"
+    assert repo.list_assets(asset_type="derived_artifact")[0].id == "asset-view-success"
+
+
+def test_sqlite_workflow_templates_and_runs_survive_reinstantiation(tmp_path: Path) -> None:
+    db_path = tmp_path / "registry.sqlite"
+    repo = SQLiteAssetRegistryRepository(db_path)
+    registry = AssetRegistryService(repo)
+    context = operation_context()
+    create_source_asset(registry, context, asset_id="asset-source")
+    workflow = WorkflowService(
+        repo,
+        transformation_service=TransformationService(repo, asset_service=registry),
+    )
+    workflow.register_template(single_step_template(), context)
+
+    result = workflow.invoke_template(
+        WorkflowInvocation(
+            template_id="workflow-single",
+            inputs={"source": {"kind": "asset", "asset_id": "asset-source"}},
+        ),
+        context,
+    )
+
+    reloaded = SQLiteAssetRegistryRepository(db_path)
+    run = reloaded.get_workflow_run(result.run.run_id)
+
+    assert reloaded.get_workflow_template("workflow-single").template_id == "workflow-single"
+    assert run.status == WorkflowRunStatus.COMPLETED
+    assert run.step_runs[0].status == WorkflowStepRunStatus.COMPLETED
+    assert reloaded.list_workflow_runs(template_id="workflow-single")[0].run_id == result.run.run_id
+    assert reloaded.list_representations(asset_id="asset-single-output")[0].kind == RepresentationKind.DERIVED
+
+
+def rich_input_template() -> WorkflowTemplate:
+    return WorkflowTemplate(
+        template_id="workflow-rich-inputs",
+        name="Rich Inputs",
+        inputs=(
+            WorkflowInputDefinition("source", WorkflowInputKind.ASSET),
+            WorkflowInputDefinition("bundle", WorkflowInputKind.COLLECTION, required=False),
+            WorkflowInputDefinition("query", WorkflowInputKind.QUERY, required=False),
+            WorkflowInputDefinition("event", WorkflowInputKind.SOURCE_EVENT, required=False),
+            WorkflowInputDefinition("payload", WorkflowInputKind.PAYLOAD, required=False),
+        ),
+        steps=(
+            WorkflowStepDefinition(
+                step_id="view",
+                operation_id="structured_view",
+                inputs={"source_asset_ids": "$inputs.source"},
+                outputs={"asset_id": "asset-rich-view"},
+            ),
+        ),
+    )
+
+
+def dependent_transformation_template() -> WorkflowTemplate:
+    return WorkflowTemplate(
+        template_id="workflow-dependent",
+        name="Dependent Transformations",
+        inputs=(WorkflowInputDefinition("source", WorkflowInputKind.ASSET),),
+        steps=(
+            WorkflowStepDefinition(
+                step_id="view",
+                operation_id="structured_view",
+                inputs={"source_asset_ids": "$inputs.source"},
+                outputs={"asset_id": "asset-view-1", "title": "First View"},
+            ),
+            WorkflowStepDefinition(
+                step_id="view_again",
+                operation_id="structured_view",
+                depends_on=("view",),
+                inputs={"source_asset_ids": "$steps.view.output_asset_ids"},
+                outputs={"asset_id": "asset-view-2", "title": "Second View"},
+            ),
+        ),
+    )
+
+
+def single_step_template() -> WorkflowTemplate:
+    return WorkflowTemplate(
+        template_id="workflow-single",
+        name="Single Step",
+        inputs=(WorkflowInputDefinition("source", WorkflowInputKind.ASSET),),
+        steps=(
+            WorkflowStepDefinition(
+                step_id="view",
+                operation_id="structured_view",
+                inputs={"source_asset_ids": "$inputs.source"},
+                outputs={"asset_id": "asset-single-output", "title": "Single Output"},
+            ),
+        ),
+    )
+
+
+def partial_template() -> WorkflowTemplate:
+    return WorkflowTemplate(
+        template_id="workflow-partial",
+        name="Partial Workflow",
+        inputs=(WorkflowInputDefinition("source", WorkflowInputKind.ASSET),),
+        steps=(
+            WorkflowStepDefinition(
+                step_id="markdown",
+                operation_id="markdown_transform",
+                inputs={"source_asset_ids": "$inputs.source"},
+                outputs={"asset_id": "asset-markdown-output"},
+                failure_behavior="continue",
+            ),
+            WorkflowStepDefinition(
+                step_id="view",
+                operation_id="structured_view",
+                inputs={"source_asset_ids": "$inputs.source"},
+                outputs={"asset_id": "asset-view-success"},
+            ),
+        ),
+    )
+
+
+def create_source_asset(
+    registry: AssetRegistryService,
+    context: OperationContext,
+    *,
+    asset_id: str,
+) -> None:
+    registry.create_asset(
+        "Source",
+        Classification(
+            asset_type="document",
+            sensitivity=Sensitivity.INTERNAL,
+            owner="Platform Knowledge",
+        ),
+        context,
+        asset_id=asset_id,
+        representations=[
+            AssetRepresentation.from_content(
+                asset_id,
+                RepresentationKind.SOURCE,
+                "text/markdown",
+                "# Source\n",
+            )
+        ],
+    )
+
+
+def operation_context() -> OperationContext:
+    actor = Actor.create(
+        ActorType.HUMAN,
+        actor_id="user-test",
+        display_name="Test User",
+        groups=["engineering"],
+    )
+    return OperationContext.create(actor, correlation_id="corr-test")

workplans/KONT-WP-0008-transformations-workflow-jobs.md

@@ -47,12 +47,13 @@ audit events.
 
 ## Implementation Status
 
-The first foundation slice is implemented for transformation operations,
+The first foundation slices are implemented for transformation operations,
-transformation run persistence, and derived artifact lineage. See
+transformation run persistence, derived artifact lineage, workflow templates,
-`docs/transformation-implementation.md`.
+and the MVP durable job runner. See `docs/transformation-implementation.md`
+and `docs/workflow-jobs-implementation.md`.
 
-Workflow templates, job runner orchestration, review gates, exception queues,
+Review gates, exception queues, and richer workflow audit reconstruction remain
-and richer workflow audit reconstruction remain open in this workplan.
+open in this workplan.
 
 ## O8.1 - Implement transformation operation registry
 
@@ -144,7 +145,7 @@ Implemented:
 
 ```task
 id: KONT-WP-0008-T004
-status: todo
+status: done
 priority: high
 state_hub_task_id: "2c55c5dd-f07b-466b-85a5-f229e41fd124"
 ```
@@ -159,11 +160,22 @@ Acceptance:
 - Workflow inputs can be assets, collections, queries, source events, or
   submitted payloads.
 
+Implemented:
+
+- `WorkflowTemplate`, `WorkflowInputDefinition`, and
+  `WorkflowStepDefinition` capture reusable templates, dependencies, inputs,
+  outputs, preconditions, policy checks, and failure behavior.
+- Template registration validates missing dependencies, duplicate step IDs,
+  unsupported failure behavior, and dependency cycles.
+- Workflow invocations preserve asset, collection, query, source event, and
+  payload input bindings; the MVP runner resolves asset bindings for
+  transformation-backed steps.
+
 ## O8.5 - Implement job runner status retry resume and cancel behavior
 
 ```task
 id: KONT-WP-0008-T005
-status: todo
+status: done
 priority: high
 state_hub_task_id: "5f4d6c88-904d-4369-90d5-eaa4d27e3010"
 ```
@@ -177,6 +189,15 @@ Acceptance:
 - Safe retry, resume, and cancellation behavior is defined per operation.
 - Recovery actions do not require direct storage edits.
 
+Implemented:
+
+- `WorkflowService` can queue, invoke, resume, retry, and cancel workflow
+  runs programmatically.
+- The MVP runner executes transformation-backed steps in dependency order and
+  persists workflow run/step state in memory and SQLite repositories.
+- Retry creates fresh governed outputs when fixed template output IDs would
+  collide with existing assets.
+
 ## O8.6 - Implement review gates human tasks and exception queues
 
 ```task