@stigmer/protos 3.12.2 → 3.12.4

package/ai/stigmer/agentic/workflowexecution/v1/api_pb.d.ts

@@ -12,141 +12,36 @@ export declare const file_ai_stigmer_agentic_workflowexecution_v1_api: GenFile;
 /**
  * WorkflowExecution represents a single runtime invocation of a WorkflowInstance.
  *
- * @internal
- * WorkflowExecution is the "Execution" layer in the Template→Instance→Execution pattern.
- * It captures the complete lifecycle of a workflow run, from initial trigger through
- * task-by-task execution to final completion or failure.
- *
- * A WorkflowExecution:
- * - References a WorkflowInstance (which contains configuration and environment bindings)
- * - Is triggered by a user action, API call, webhook, or scheduled event
- * - Executes tasks sequentially or in parallel based on the workflow definition
- * - Tracks real-time progress through tasks and phases
- * - Captures outputs, errors, and execution metadata
- * - Is ephemeral - deleted after completion based on retention policies
- *
- * Execution Pattern:
- * Workflow "customer-onboarding" (template)
- *   → WorkflowInstance "acme-onboarding" (with prod-env)
- *     → WorkflowExecution "acme-onboarding-20250111-143022" (specific run)
- *       - Phase: IN_PROGRESS
- *       - Tasks: [validate_email: COMPLETED, create_account: IN_PROGRESS, send_welcome: PENDING]
- *       - Progress: 1/3 tasks completed
- *
- * Separation of Concerns:
- * - User Inputs (spec): workflow_instance_id, trigger_message, trigger_metadata, runtime_env
- * - System State (status): phase, tasks, output, timestamps, errors
- *
- * This separation ensures:
- * - Clear boundary between what users control (spec) and what the system manages (status)
- * - Status can be updated independently during execution without modifying user inputs
- * - Execution can be retried by creating a new WorkflowExecution with the same spec
- *
  * @generated from message ai.stigmer.agentic.workflowexecution.v1.WorkflowExecution
  */
 export type WorkflowExecution = Message<"ai.stigmer.agentic.workflowexecution.v1.WorkflowExecution"> & {
     /**
      * API version for this resource type.
      *
-     * @internal
-     * Format: 'agentic.stigmer.ai/v1'
-     * Validated as const to ensure version consistency across all workflow execution resources.
-     *
      * @generated from field: string api_version = 1;
      */
     apiVersion: string;
     /**
      * Resource kind identifier.
      *
-     * @internal
-     * Must be exactly 'WorkflowExecution' to match the message name.
-     * Validated as const for type safety and resource identification.
-     *
      * @generated from field: string kind = 2;
      */
     kind: string;
     /**
      * Resource metadata including name, organization, visibility, and labels.
      *
-     * @internal
-     * All workflow executions belong to an organization. Visibility (public/private)
-     * is typically PRIVATE for executions since they contain runtime data.
-     *
-     * Naming Pattern:
-     * - ID Format: "wfx_abc123xyz456" (auto-generated, unique)
-     * - Name Format: "{workflow_instance_name}-{timestamp}" (e.g., "prod-deploy-20250111-143022")
-     * - Slug Format: Same as name, URL-safe
-     *
-     * Labels and Tags:
-     * - Labels: workflow_instance_id, workflow_id, trigger_source
-     * - Tags: environment names, team names, execution metadata
-     *
      * @generated from field: ai.stigmer.commons.apiresource.ApiResourceMetadata metadata = 3;
      */
     metadata?: ApiResourceMetadata;
     /**
      * User-provided inputs and configuration for this workflow execution.
      *
-     * @internal
-     * Contains:
-     * - workflow_instance_id: Which WorkflowInstance to execute (required)
-     * - trigger_message: Input message or payload for the workflow (optional)
-     * - trigger_metadata: Metadata about who/what triggered the execution (optional)
-     * - runtime_env: Execution-specific environment variables and secrets (optional)
-     *
-     * The spec is immutable after creation - it represents the "inputs" for this execution.
-     * To retry with different inputs, create a new WorkflowExecution with updated spec.
-     *
-     * Example:
-     * spec {
-     *   workflow_instance_id: "wfi_customer-onboarding-prod"
-     *   trigger_message: "New customer: customer-email@example.com"
-     *   trigger_metadata: {
-     *     "source": "api"
-     *     "caller_id": "usr-john-doe"
-     *     "timestamp": "2025-01-11T14:30:22Z"
-     *   }
-     *   runtime_env: {
-     *     "CUSTOMER_EMAIL": { value: "customer-email@example.com" }
-     *     "WEBHOOK_URL": { secret_ref: "sec-webhook-callback" }
-     *   }
-     * }
-     *
      * @generated from field: ai.stigmer.agentic.workflowexecution.v1.WorkflowExecutionSpec spec = 4;
      */
     spec?: WorkflowExecutionSpec;
     /**
      * System-managed execution state and results.
      *
-     * @internal
-     * Contains:
-     * - phase: Current lifecycle phase (PENDING → IN_PROGRESS → COMPLETED/FAILED/CANCELLED)
-     * - tasks: List of workflow tasks with their execution state (source of truth for progress)
-     * - output: Final workflow output (JSON structure, only for COMPLETED executions)
-     * - error: Error message (only for FAILED executions)
-     * - started_at: Timestamp when execution started
-     * - completed_at: Timestamp when execution finished (COMPLETED/FAILED/CANCELLED)
-     * - temporal_workflow_id: Correlation ID for workflow engine
-     *
-     * The status is continuously updated by the workflow execution engine as the workflow progresses.
-     * Users can read status but cannot modify it - it reflects the actual execution state.
-     *
-     * Progress Tracking:
-     * - Total tasks: tasks.length
-     * - Completed tasks: count(tasks where status in [COMPLETED, FAILED, SKIPPED])
-     * - Progress percentage: (completed_tasks / total_tasks) * 100
-     *
-     * Example (in-progress execution):
-     * status {
-     *   phase: EXECUTION_IN_PROGRESS
-     *   tasks: [
-     *     { task_id: "task-1", task_name: "validate_email", status: WORKFLOW_TASK_COMPLETED }
-     *     { task_id: "task-2", task_name: "create_account", status: WORKFLOW_TASK_IN_PROGRESS }
-     *     { task_id: "task-3", task_name: "send_welcome", status: WORKFLOW_TASK_PENDING }
-     *   ]
-     *   started_at: "2025-01-11T14:30:22Z"
-     * }
-     *
      * @generated from field: ai.stigmer.agentic.workflowexecution.v1.WorkflowExecutionStatus status = 5;
      */
     status?: WorkflowExecutionStatus;
@@ -159,251 +54,84 @@ export declare const WorkflowExecutionSchema: GenMessage<WorkflowExecution>;
 /**
  * WorkflowExecutionStatus contains all system-managed execution state and results.
  *
- * @internal
- * Everything populated during or after execution goes here, not in spec.
- * This message follows the standard pattern: audit information at field 99, custom fields at 1-98.
- *
- * Status Lifecycle:
- * 1. Created: audit.created_at set, phase = EXECUTION_PENDING
- * 2. Started: started_at set, phase = EXECUTION_IN_PROGRESS, tasks populated
- * 3. Progress: tasks updated as workflow runner sends status updates
- * 4. Completed: completed_at set, phase = EXECUTION_COMPLETED, output populated
- * 5. Failed: completed_at set, phase = EXECUTION_FAILED, error populated
- * 6. Cancelled: completed_at set, phase = EXECUTION_CANCELLED
- *
- * The status is read-only to users and continuously updated by the workflow execution engine.
- *
- * Progress Tracking Pattern (Matches Agent Execution):
- * - Agent execution: messages[], tool_calls[], sub_agent_executions[] (transformed from LLM streaming events)
- * - Workflow execution: tasks[] (transformed from workflow task execution)
- * - Both store structured state, not raw streaming events
- * - Both use updateStatus RPC for progressive updates during execution
- *
  * @generated from message ai.stigmer.agentic.workflowexecution.v1.WorkflowExecutionStatus
  */
 export type WorkflowExecutionStatus = Message<"ai.stigmer.agentic.workflowexecution.v1.WorkflowExecutionStatus"> & {
     /**
      * Standard audit information including timestamps and created-by identity.
      *
-     * @internal
-     * Always at field 99 for consistency across all Stigmer API resources.
-     *
-     * Contains:
-     * - created_at: When this execution was created (ISO 8601 timestamp)
-     * - updated_at: Last time status was updated (ISO 8601 timestamp)
-     * - created_by: User or system that created this execution
-     *
      * @generated from field: ai.stigmer.commons.apiresource.ApiResourceAudit audit = 99;
      */
     audit?: ApiResourceAudit;
     /**
      * Current execution lifecycle phase.
      *
-     * @internal
-     * Phase Transitions:
-     * PENDING → IN_PROGRESS → COMPLETED
-     *        ↓              ↘ FAILED
-     *        ↓              ↘ CANCELLED
-     *
-     * The phase is used for:
-     * - UI status indicators (progress bars, badges)
-     * - Filtering executions (show only failed, show only in-progress)
-     * - Alerting and notifications (notify on failure)
-     * - Retry logic (retry failed executions)
-     *
-     * Validation: Must be a defined enum value (no unspecified/unknown).
-     *
      * @generated from field: ai.stigmer.agentic.workflowexecution.v1.ExecutionPhase phase = 1;
      */
     phase: ExecutionPhase;
     /**
      * Workflow tasks with their individual execution state.
      *
-     * @internal
-     * Tasks represent the atomic units of work within the workflow.
-     * Each task has:
-     * - task_id: Unique identifier within this execution
-     * - task_name: Human-readable name (e.g., "validate_email", "send_notification")
-     * - task_type: Type of task (agent_invocation, api_call, approval, etc.)
-     * - status: Current task status (PENDING, IN_PROGRESS, COMPLETED, FAILED, SKIPPED)
-     * - input: Task input parameters (JSON structure)
-     * - output: Task output results (JSON structure, only for COMPLETED tasks)
-     * - error: Error message (only for FAILED tasks)
-     * - started_at: When task started executing
-     * - completed_at: When task finished (COMPLETED/FAILED/SKIPPED)
-     *
-     * Tasks are populated as the workflow progresses through its task graph.
-     * The order in this list reflects the execution order (sequential or parallel).
-     *
-     * Progress Calculation:
-     * - Total tasks: tasks.length
-     * - Completed tasks: count(tasks where status in [COMPLETED, FAILED, SKIPPED])
-     * - Progress percentage: (completed_tasks / total_tasks) * 100
-     * - Current task: tasks.find(status == IN_PROGRESS)
-     *
      * @generated from field: repeated ai.stigmer.agentic.workflowexecution.v1.WorkflowTask tasks = 2;
      */
     tasks: WorkflowTask[];
     /**
      * Final workflow output, populated only when phase is EXECUTION_COMPLETED.
      *
-     * @internal
-     * The output structure is workflow-specific and defined by the Workflow template.
-     * Common patterns:
-     * - API response data (e.g., created user ID, order confirmation)
-     * - Aggregated results from multiple tasks
-     * - Links to generated artifacts (reports, files)
-     * - Summary statistics (items processed, duration)
-     *
      * @generated from field: google.protobuf.Struct output = 3;
      */
     output?: JsonObject;
     /**
      * Error message, populated only when phase is EXECUTION_FAILED.
      *
-     * @internal
-     * Contains a human-readable description of what went wrong.
-     *
-     * Error message includes:
-     * - What failed (which task or workflow step)
-     * - Why it failed (validation error, API error, timeout, etc.)
-     * - How to fix it (if known)
-     *
-     * For detailed debugging, inspect tasks[].error for task-specific error messages.
-     *
      * @generated from field: string error = 4;
      */
     error: string;
     /**
      * ISO 8601 timestamp when execution started processing.
      *
-     * @internal
-     * Set when phase transitions from PENDING to IN_PROGRESS.
-     * Format: "YYYY-MM-DDTHH:MM:SSZ" (UTC timezone)
-     *
      * @generated from field: string started_at = 5;
      */
     startedAt: string;
     /**
      * ISO 8601 timestamp when execution reached a terminal state.
      *
-     * @internal
-     * Set when the workflow reaches a terminal state:
-     * - EXECUTION_COMPLETED: Successfully finished
-     * - EXECUTION_FAILED: Failed during execution
-     * - EXECUTION_CANCELLED: Cancelled by user or system
-     *
-     * Not set for PENDING or IN_PROGRESS executions.
-     * Format: "YYYY-MM-DDTHH:MM:SSZ" (UTC timezone)
-     *
      * @generated from field: string completed_at = 6;
      */
     completedAt: string;
     /**
      * Correlation ID for the underlying workflow engine.
      *
-     * @internal
-     * This is the workflow ID in Temporal, used for:
-     * - Correlation between Stigmer and Temporal (for debugging)
-     * - Querying Temporal directly (for advanced troubleshooting)
-     * - Signaling or cancelling Temporal workflows
-     *
-     * Format: Typically "{workflow_instance_id}-{execution_id}" or a UUID
-     * Example: "wfi_prod-deploy-wfx_abc123xyz456" or "temporal-wf-uuid-12345"
-     *
-     * This field is optional and only relevant when Temporal is used as the execution engine.
-     * Other workflow engines (Step Functions, Argo, etc.) may use different correlation IDs.
-     *
      * @generated from field: string temporal_workflow_id = 7;
      */
     temporalWorkflowId: string;
     /**
      * Pending approvals from child agent tool executions.
      *
-     * @internal
-     * Populated when workflow tasks invoke agents that enter
-     * EXECUTION_WAITING_FOR_APPROVAL phase. This surfaces all approval
-     * requests at the workflow level for UI visibility.
-     *
-     * Guarded Update Protocol:
-     * This field is only modified when UpdateStatusInput.update_pending_approvals
-     * is explicitly set to true. Normal event emissions (which don't concern
-     * approvals) leave this field untouched, preventing race conditions between
-     * concurrent status writers.
-     *
-     * Only call-agent-status manages this field:
-     * - Non-empty list: child agent(s) need approval
-     * - Empty list + update_pending_approvals=true: all approvals resolved
-     *
-     * Parallel Agents:
-     * When multiple child agents run in parallel, entries from different children
-     * accumulate in this list. Each entry's child_agent_execution_id distinguishes
-     * the source.
-     *
      * @generated from field: repeated ai.stigmer.agentic.workflowexecution.v1.WorkflowPendingApproval pending_approvals = 9;
      */
     pendingApprovals: WorkflowPendingApproval[];
     /**
      * Cumulative cost across all tasks in micro-USD (1 USD = 1,000,000 micros).
      *
-     * @internal
-     * Updated by the runner alongside each status update. Reflects the
-     * budget tracker's accumulated cost at the time of the last status write.
-     * Used by getExecutionSummary for fast aggregation without scanning events.
-     *
-     * Consistent with WorkflowBudget.max_cost_micros and the billing domain's
-     * micro-USD convention (CostStamp.provider_cost_micros, etc.).
-     *
-     * @since Cost Data Pipeline
-     *
      * @generated from field: int64 total_cost_micros = 10;
      */
     totalCostMicros: bigint;
     /**
      * Cumulative input tokens consumed across all LLM and agent tasks.
      *
-     * @internal
-     * Input tokens represent prompt/context tokens sent to the model.
-     * Updated alongside total_cost_micros from the budget tracker.
-     *
-     * @since Cost Data Pipeline
-     *
      * @generated from field: int64 total_input_tokens = 11;
      */
     totalInputTokens: bigint;
     /**
      * Cumulative output tokens consumed across all LLM and agent tasks.
      *
-     * @internal
-     * Output tokens represent completion/generation tokens returned by the model.
-     * Updated alongside total_cost_micros from the budget tracker.
-     *
-     * @since Cost Data Pipeline
-     *
      * @generated from field: int64 total_output_tokens = 12;
      */
     totalOutputTokens: bigint;
     /**
      * SHA-256 hash identifying which workflow version was used for this execution.
      *
-     * @internal
-     * Pinned at execution creation time from Workflow.status.version_hash.
-     * Immutable after creation — represents the exact workflow definition this
-     * execution ran (or will run, if still pending).
-     *
-     * Consumers:
-     * - Runner: fetches the version-specific CNCF YAML via getVersion() during
-     *   hydration, ensuring the execution runs the intended definition even if
-     *   the workflow has been updated since creation.
-     * - Execution viewer: fetches the version entry to render the correct graph
-     *   for historical executions, eliminating the version mismatch problem.
-     *
-     * Empty for executions created before workflow versioning was introduced.
-     * In that case, consumers fall back to fetching the current workflow
-     * definition (legacy behavior with mismatch warning).
-     *
-     * @since Workflow Versioning
-     *
      * @generated from field: string workflow_version_hash = 13;
      */
     workflowVersionHash: string;
@@ -411,23 +139,6 @@ export type WorkflowExecutionStatus = Message<"ai.stigmer.agentic.workflowexecut
      * Pending file reviews from child agent executions, surfaced for
      * workflow-level visibility and in-place keep/discard decisioning.
      *
-     * @internal
-     * A reference-only sibling of pending_approvals (field 9): each entry names a
-     * child agent execution and the change_set ids it currently has
-     * AWAITING_REVIEW. The heavy content (diffs, blobs) is NOT copied here — it
-     * stays single-sourced on the child's
-     * AgentExecution.status.file_change_sets and is read on demand by the UI.
-     *
-     * Guarded, per-child merge protocol:
-     * Modified only when WorkflowExecutionUpdateStatusInput.update_pending_file_reviews
-     * is true, scoped by pending_update_child_agent_execution_id. The write
-     * replaces the entry for that one child and preserves every sibling child's
-     * entry, so parallel child agents never clobber each other. A scoped write with
-     * an empty change_set list removes that child's entry. Only call-agent-status
-     * manages this field.
-     *
-     * @since Workflow-Parent File Review
-     *
      * @generated from field: repeated ai.stigmer.agentic.workflowexecution.v1.WorkflowPendingFileReview pending_file_reviews = 14;
      */
     pendingFileReviews: WorkflowPendingFileReview[];
@@ -440,11 +151,6 @@ export declare const WorkflowExecutionStatusSchema: GenMessage<WorkflowExecution
 /**
  * WorkflowPendingApproval pairs approval details with routing information for workflow-level forwarding.
  *
- * @internal
- * PendingApproval is an agentexecution domain type — it describes what tool
- * needs approval. WorkflowPendingApproval adds the workflow-level concern:
- * which child agent execution the approval should be forwarded to.
- *
  * @generated from message ai.stigmer.agentic.workflowexecution.v1.WorkflowPendingApproval
  */
 export type WorkflowPendingApproval = Message<"ai.stigmer.agentic.workflowexecution.v1.WorkflowPendingApproval"> & {
@@ -457,13 +163,6 @@ export type WorkflowPendingApproval = Message<"ai.stigmer.agentic.workflowexecut
     /**
      * ID of the child agent execution to forward the approval decision to.
      *
-     * @internal
-     * Set by the runner when surfacing child agent approvals
-     * at the workflow level. WorkflowExecution.SubmitApproval uses this
-     * to route the decision to the correct AgentExecution.SubmitApproval RPC.
-     *
-     * Format: AgentExecution.metadata.id (e.g., "aex_abc123xyz456")
-     *
      * @generated from field: string child_agent_execution_id = 2;
      */
     childAgentExecutionId: string;
@@ -477,38 +176,18 @@ export declare const WorkflowPendingApprovalSchema: GenMessage<WorkflowPendingAp
  * WorkflowPendingFileReview references a child agent execution's file-review
  * gate for workflow-level surfacing. Reference-only — never a copy of the diff.
  *
- * @internal
- * FileChangeSet is an agentexecution domain concept owned by the child
- * AgentExecution. WorkflowPendingFileReview adds only the workflow-level
- * concern: which child holds a file-review gate, and which of its change sets
- * are awaiting a decision. The UI reads the child's file_change_sets to render
- * the diff; WorkflowExecution.submitFileDecision forwards the verdict to the
- * child. This mirrors how WorkflowPendingApproval references a child's tool
- * approval without embedding the tool call.
- *
  * @generated from message ai.stigmer.agentic.workflowexecution.v1.WorkflowPendingFileReview
  */
 export type WorkflowPendingFileReview = Message<"ai.stigmer.agentic.workflowexecution.v1.WorkflowPendingFileReview"> & {
     /**
      * ID of the child agent execution that holds the file-review gate.
      *
-     * @internal
-     * WorkflowExecution.submitFileDecision uses this to route the decision to the
-     * correct AgentExecution.submitFileDecision RPC.
-     * Format: AgentExecution.metadata.id (e.g., "aex_abc123xyz456")
-     *
      * @generated from field: string child_agent_execution_id = 1;
      */
     childAgentExecutionId: string;
     /**
      * IDs of the child's change sets currently AWAITING_REVIEW.
      *
-     * @internal
-     * Each matches a FileChangeSet.id on the child's
-     * AgentExecution.status.file_change_sets. Empty means the child has no
-     * outstanding review (used transiently to clear this child's entry under the
-     * per-child merge protocol).
-     *
      * @generated from field: repeated string change_set_id = 2;
      */
     changeSetId: string[];
@@ -521,209 +200,90 @@ export declare const WorkflowPendingFileReviewSchema: GenMessage<WorkflowPending
 /**
  * WorkflowTask represents a single task within a workflow execution.
  *
- * @internal
- * Tasks are the atomic units of work in a workflow. Each task:
- * - Has a specific type (agent invocation, API call, approval, conditional, etc.)
- * - Receives input parameters from the workflow context or previous tasks
- * - Executes its specific logic (invoke an agent, call an API, wait for approval)
- * - Produces output that can be used by subsequent tasks
- * - Tracks its own execution state (pending, in-progress, completed, failed, skipped)
- *
- * Task Execution Flow:
- * 1. Task is in PENDING state when workflow starts
- * 2. When task dependencies are met, task_status changes to IN_PROGRESS
- * 3. Task executes its logic (invoke agent, call API, etc.)
- * 4. Task completes successfully (COMPLETED) or fails (FAILED) or is skipped (SKIPPED)
- *
  * @generated from message ai.stigmer.agentic.workflowexecution.v1.WorkflowTask
  */
 export type WorkflowTask = Message<"ai.stigmer.agentic.workflowexecution.v1.WorkflowTask"> & {
     /**
      * Unique task identifier within this workflow execution.
      *
-     * @internal
-     * Format: Typically "task-{number}" or a descriptive slug
-     * Examples: "task-1", "task-validate-email", "task-send-notification"
-     *
-     * The task_id is unique within this WorkflowExecution but may repeat across
-     * different executions of the same WorkflowInstance (same task, different run).
-     *
      * @generated from field: string task_id = 1;
      */
     taskId: string;
     /**
      * Human-readable task name.
      *
-     * @internal
-     * Describes what this task does in plain language.
-     * Used in UI to show task progress and in logs for debugging.
-     *
-     * Naming conventions:
-     * - Use verb phrases (validate, create, send, wait)
-     * - Be specific about what's being operated on
-     * - Keep it concise (under 50 characters)
-     *
      * @generated from field: string task_name = 2;
      */
     taskName: string;
     /**
      * Type of task (agent invocation, API call, approval, etc.).
      *
-     * @internal
-     * Determines how the task is executed by the workflow engine.
-     * The task_type influences:
-     * - How task.input is structured (different types expect different input schemas)
-     * - How task.output is produced (different types produce different outputs)
-     * - How errors are handled (retry policies, timeout behaviors)
-     *
-     * Validation: Must be a defined enum value (no unspecified).
-     *
      * @generated from field: ai.stigmer.agentic.workflowexecution.v1.WorkflowTaskType task_type = 3;
      */
     taskType: WorkflowTaskType;
     /**
      * Task input parameters, structured as JSON.
      *
-     * @internal
-     * Contains the configuration and data needed for this task to execute.
-     * The structure varies by task_type.
-     *
-     * Input can reference:
-     * - Workflow inputs: {{workflow.input.field_name}}
-     * - Previous task outputs: {{tasks.task-1.output.field_name}}
-     * - Environment variables: {{env.VARIABLE_NAME}}
-     *
      * @generated from field: google.protobuf.Struct input = 4;
      */
     input?: JsonObject;
     /**
      * Task output results, populated only when status is WORKFLOW_TASK_COMPLETED.
      *
-     * @internal
-     * Contains the data produced by this task after successful execution.
-     * Output can be referenced by subsequent tasks using: {{tasks.this-task-id.output.field_name}}
-     *
      * @generated from field: google.protobuf.Struct output = 5;
      */
     output?: JsonObject;
     /**
      * Current task execution status.
      *
-     * @internal
-     * Status Transitions:
-     * PENDING → IN_PROGRESS → COMPLETED
-     *        ↓              ↘ FAILED
-     *        ↓              ↘ SKIPPED (if conditional)
-     *
-     * Validation: Must be a defined enum value (no unspecified).
-     *
      * @generated from field: ai.stigmer.agentic.workflowexecution.v1.WorkflowTaskStatus status = 6;
      */
     status: WorkflowTaskStatus;
     /**
      * ISO 8601 timestamp when the task started executing.
      *
-     * @internal
-     * Set when task status changes from PENDING to IN_PROGRESS.
-     * Format: "YYYY-MM-DDTHH:MM:SSZ" (UTC timezone)
-     *
      * @generated from field: string started_at = 7;
      */
     startedAt: string;
     /**
      * ISO 8601 timestamp when the task reached a terminal state.
      *
-     * @internal
-     * Set when task reaches COMPLETED, FAILED, or SKIPPED.
-     * Not set for PENDING or IN_PROGRESS tasks.
-     * Format: "YYYY-MM-DDTHH:MM:SSZ" (UTC timezone)
-     *
      * @generated from field: string completed_at = 8;
      */
     completedAt: string;
     /**
      * Error message, populated only when status is WORKFLOW_TASK_FAILED.
      *
-     * @internal
-     * Contains a human-readable description of why the task failed.
-     *
-     * Error message includes:
-     * - What operation failed (API call, agent invocation, etc.)
-     * - Error type (validation error, network error, timeout, etc.)
-     * - Error details (status code, exception message, stacktrace)
-     * - How to fix it (if known)
-     *
      * @generated from field: string error = 9;
      */
     error: string;
     /**
      * Task-specific metadata as arbitrary JSON.
      *
-     * @internal
-     * Contains task-specific information that doesn't fit in other fields.
-     * Used for:
-     * - Retry count (how many times this task was retried)
-     * - Agent execution ID (for WORKFLOW_TASK_AGENT_INVOCATION)
-     * - API response headers (for WORKFLOW_TASK_API_CALL)
-     * - Approval history (who approved, when, comments)
-     * - Performance metrics (execution time, memory usage)
-     *
      * @generated from field: google.protobuf.Struct metadata = 10;
      */
     metadata?: JsonObject;
     /**
      * Artifact IDs produced by this task.
      *
-     * @internal
-     * Populated when the task's output (or portions of it) is auto-promoted
-     * to the artifact store because it exceeds the size threshold (256KB),
-     * or when the workflow author explicitly declares artifact persistence.
-     *
-     * Each entry is an Artifact ID (format: "art_{unique-suffix}") that can
-     * be resolved via the Artifact.get() and Artifact.getDownloadUrl() RPCs.
-     *
-     * When artifact_ids is non-empty, the task's output field contains
-     * artifact references ({"_artifact_ref": "art_xxx", ...}) instead of
-     * the original inline data. Consumers (execution viewer, SDK hooks)
-     * detect these references and resolve them via the Artifact APIs.
-     *
-     * @since T07 (Artifact Store)
-     *
      * @generated from field: repeated string artifact_ids = 11;
      */
     artifactIds: string[];
     /**
      * Cost incurred by this task in micro-USD (1 USD = 1,000,000 micros).
      *
-     * @internal
-     * Non-zero for cost-incurring task kinds (llm_call, agent_call).
-     * Zero for non-LLM tasks (transform, validate, emit_event, etc.).
-     * Set by the runner when the task completes.
-     *
-     * @since Cost Data Pipeline
-     *
      * @generated from field: int64 cost_micros = 12;
      */
     costMicros: bigint;
     /**
      * Input (prompt/context) tokens consumed by this task.
      *
-     * @internal
-     * Non-zero for LLM-backed tasks. Zero for non-LLM tasks.
-     *
-     * @since Cost Data Pipeline
-     *
      * @generated from field: int64 input_tokens = 13;
      */
     inputTokens: bigint;
     /**
      * Output (completion/generation) tokens produced by this task.
      *
-     * @internal
-     * Non-zero for LLM-backed tasks. Zero for non-LLM tasks.
-     *
-     * @since Cost Data Pipeline
-     *
      * @generated from field: int64 output_tokens = 14;
      */
     outputTokens: bigint;
@@ -737,20 +297,6 @@ export type WorkflowTask = Message<"ai.stigmer.agentic.workflowexecution.v1.Work
      * persisted before this field existed — consumers treat empty as a
      * generic review.
      *
-     * @internal
-     * Written by the runner's task status accumulator on the
-     * waiting_approval transition and retained after the gate resolves
-     * (the record of what kind of review was performed). The user-input
-     * length constraint (max 63 chars) is enforced at the source field,
-     * HumanInputTaskConfig.ui_hint; this system-written copy is trusted,
-     * matching the other string fields on this message.
-     *
-     * Not to be confused with WorkflowExecutionStatus.pending_approvals,
-     * which carries forwarded child-agent *tool* approvals — workflow-native
-     * human_input gates live only here, on tasks[].
-     *
-     * @since Review Payloads (stigmer/stigmer#234)
-     *
      * @generated from field: string ui_hint = 15;
      */
     uiHint: string;