Merge pull request #1437 from trheyi/main

Enhance Execution Control and Error Handling in Robot Manager

Author: trheyi

agent/robot/api/execution.go

@@ -130,7 +130,12 @@ func PauseExecution(ctx *types.Context, execID string) error {
 		return err
 	}
 
-	return mgr.PauseExecution(ctx, execID)
+	if err := mgr.PauseExecution(ctx, execID); err != nil {
+		return err
+	}
+
+	// Update database status to paused
+	return getExecutionStore().UpdateStatus(context.Background(), execID, types.ExecPaused, "")
 }
 
 // ResumeExecution resumes a paused execution
@@ -144,7 +149,12 @@ func ResumeExecution(ctx *types.Context, execID string) error {
 		return err
 	}
 
-	return mgr.ResumeExecution(ctx, execID)
+	if err := mgr.ResumeExecution(ctx, execID); err != nil {
+		return err
+	}
+
+	// Update database status back to running
+	return getExecutionStore().UpdateStatus(context.Background(), execID, types.ExecRunning, "")
 }
 
 // StopExecution stops a running execution
@@ -158,7 +168,12 @@ func StopExecution(ctx *types.Context, execID string) error {
 		return err
 	}
 
-	return mgr.StopExecution(ctx, execID)
+	if err := mgr.StopExecution(ctx, execID); err != nil {
+		return err
+	}
+
+	// Update database status to cancelled
+	return getExecutionStore().UpdateStatus(context.Background(), execID, types.ExecCancelled, "User cancelled")
 }
 
 // ==================== Execution Status API ====================

agent/robot/executor/dryrun/executor.go

@@ -42,8 +42,18 @@ func NewWithConfig(config types.DryRunConfig) *Executor {
 	}
 }
 
-// Execute simulates robot execution without real Agent calls
+// Execute simulates robot execution without real Agent calls (auto-generates ID)
 func (e *Executor) Execute(ctx *robottypes.Context, robot *robottypes.Robot, trigger robottypes.TriggerType, data interface{}) (*robottypes.Execution, error) {
+	return e.ExecuteWithControl(ctx, robot, trigger, data, "", nil)
+}
+
+// ExecuteWithID simulates robot execution with a pre-generated execution ID (no control)
+func (e *Executor) ExecuteWithID(ctx *robottypes.Context, robot *robottypes.Robot, trigger robottypes.TriggerType, data interface{}, execID string) (*robottypes.Execution, error) {
+	return e.ExecuteWithControl(ctx, robot, trigger, data, execID, nil)
+}
+
+// ExecuteWithControl simulates robot execution with execution control
+func (e *Executor) ExecuteWithControl(ctx *robottypes.Context, robot *robottypes.Robot, trigger robottypes.TriggerType, data interface{}, execID string, control robottypes.ExecutionControl) (*robottypes.Execution, error) {
 	if robot == nil {
 		return nil, fmt.Errorf("robot cannot be nil")
 	}
@@ -54,9 +64,14 @@ func (e *Executor) Execute(ctx *robottypes.Context, robot *robottypes.Robot, tri
 		startPhaseIndex = 1 // Skip P0
 	}
 
+	// Use provided execID or generate new one
+	if execID == "" {
+		execID = fmt.Sprintf("dryrun_%d", time.Now().UnixNano())
+	}
+
 	// Create execution record
 	exec := &robottypes.Execution{
-		ID:          fmt.Sprintf("dryrun_%d", time.Now().UnixNano()),
+		ID:          execID,
 		MemberID:    robot.MemberID,
 		TeamID:      robot.TeamID,
 		TriggerType: trigger,
@@ -106,6 +121,24 @@ func (e *Executor) Execute(ctx *robottypes.Context, robot *robottypes.Robot, tri
 	// Execute phases with mock data
 	phases := robottypes.AllPhases[startPhaseIndex:]
 	for _, phase := range phases {
+		// Check if cancelled
+		select {
+		case <-ctx.Context.Done():
+			exec.Status = robottypes.ExecCancelled
+			exec.Error = "execution cancelled"
+			return exec, nil
+		default:
+		}
+
+		// Wait if paused
+		if control != nil {
+			if err := control.WaitIfPaused(); err != nil {
+				exec.Status = robottypes.ExecCancelled
+				exec.Error = "execution cancelled while paused"
+				return exec, nil
+			}
+		}
+
 		exec.Phase = phase
 
 		// Phase start callback

agent/robot/executor/sandbox/executor.go

@@ -48,8 +48,18 @@ func NewWithConfig(config types.SandboxConfig) *Executor {
 	}
 }
 
-// Execute runs robot execution within sandbox constraints
+// Execute runs robot execution within sandbox constraints (auto-generates ID)
 func (e *Executor) Execute(ctx *robottypes.Context, robot *robottypes.Robot, trigger robottypes.TriggerType, data interface{}) (*robottypes.Execution, error) {
+	return e.ExecuteWithControl(ctx, robot, trigger, data, "", nil)
+}
+
+// ExecuteWithID runs robot execution within sandbox constraints with a pre-generated execution ID (no control)
+func (e *Executor) ExecuteWithID(ctx *robottypes.Context, robot *robottypes.Robot, trigger robottypes.TriggerType, data interface{}, execID string) (*robottypes.Execution, error) {
+	return e.ExecuteWithControl(ctx, robot, trigger, data, execID, nil)
+}
+
+// ExecuteWithControl runs robot execution within sandbox constraints with execution control
+func (e *Executor) ExecuteWithControl(ctx *robottypes.Context, robot *robottypes.Robot, trigger robottypes.TriggerType, data interface{}, execID string, control robottypes.ExecutionControl) (*robottypes.Execution, error) {
 	if robot == nil {
 		return nil, fmt.Errorf("robot cannot be nil")
 	}
@@ -67,9 +77,14 @@ func (e *Executor) Execute(ctx *robottypes.Context, robot *robottypes.Robot, tri
 		startPhaseIndex = 1
 	}
 
+	// Use provided execID or generate new one
+	if execID == "" {
+		execID = fmt.Sprintf("sandbox_%d", time.Now().UnixNano())
+	}
+
 	// Create execution record
 	exec := &robottypes.Execution{
-		ID:          fmt.Sprintf("sandbox_%d", time.Now().UnixNano()),
+		ID:          execID,
 		MemberID:    robot.MemberID,
 		TeamID:      robot.TeamID,
 		TriggerType: trigger,
@@ -99,7 +114,7 @@ func (e *Executor) Execute(ctx *robottypes.Context, robot *robottypes.Robot, tri
 	// Execute phases with sandbox constraints
 	phases := robottypes.AllPhases[startPhaseIndex:]
 	for _, phase := range phases {
-		// Check timeout
+		// Check timeout or cancellation
 		select {
 		case <-execCtx.Done():
 			exec.Status = robottypes.ExecFailed
@@ -108,6 +123,15 @@ func (e *Executor) Execute(ctx *robottypes.Context, robot *robottypes.Robot, tri
 		default:
 		}
 
+		// Wait if paused
+		if control != nil {
+			if err := control.WaitIfPaused(); err != nil {
+				exec.Status = robottypes.ExecCancelled
+				exec.Error = "execution cancelled while paused"
+				return exec, nil
+			}
+		}
+
 		exec.Phase = phase
 
 		if e.config.OnPhaseStart != nil {

agent/robot/executor/standard/executor.go

@@ -45,8 +45,19 @@ func NewWithConfig(config types.Config) *Executor {
 	}
 }
 
-// Execute runs a robot through all applicable phases with real Agent calls
+// Execute runs a robot through all applicable phases with real Agent calls (auto-generates ID)
 func (e *Executor) Execute(ctx *robottypes.Context, robot *robottypes.Robot, trigger robottypes.TriggerType, data interface{}) (*robottypes.Execution, error) {
+	return e.ExecuteWithControl(ctx, robot, trigger, data, "", nil)
+}
+
+// ExecuteWithID runs a robot through all applicable phases with a pre-generated execution ID (no control)
+func (e *Executor) ExecuteWithID(ctx *robottypes.Context, robot *robottypes.Robot, trigger robottypes.TriggerType, data interface{}, execID string) (*robottypes.Execution, error) {
+	return e.ExecuteWithControl(ctx, robot, trigger, data, execID, nil)
+}
+
+// ExecuteWithControl runs a robot through all applicable phases with execution control
+// control: optional, allows pause/resume functionality during execution
+func (e *Executor) ExecuteWithControl(ctx *robottypes.Context, robot *robottypes.Robot, trigger robottypes.TriggerType, data interface{}, execID string, control robottypes.ExecutionControl) (*robottypes.Execution, error) {
 	if robot == nil {
 		return nil, fmt.Errorf("robot cannot be nil")
 	}
@@ -57,10 +68,15 @@ func (e *Executor) Execute(ctx *robottypes.Context, robot *robottypes.Robot, tri
 		startPhaseIndex = 1 // Skip P0 (Inspiration)
 	}
 
+	// Use provided execID or generate new one
+	if execID == "" {
+		execID = utils.NewID()
+	}
+
 	// Create execution (Job system removed, using ExecutionStore only)
 	input := types.BuildTriggerInput(trigger, data)
 	exec := &robottypes.Execution{
-		ID:          utils.NewID(),
+		ID:          execID,
 		MemberID:    robot.MemberID,
 		TeamID:      robot.TeamID,
 		TriggerType: trigger,
@@ -174,7 +190,31 @@ func (e *Executor) Execute(ctx *robottypes.Context, robot *robottypes.Robot, tri
 	// Execute phases
 	phases := robottypes.AllPhases[startPhaseIndex:]
 	for _, phase := range phases {
-		if err := e.runPhase(ctx, exec, phase, data); err != nil {
+		if err := e.runPhase(ctx, exec, phase, data, control); err != nil {
+			// Check if execution was cancelled
+			if err == robottypes.ErrExecutionCancelled {
+				exec.Status = robottypes.ExecCancelled
+				exec.Error = "execution cancelled by user"
+				now := time.Now()
+				exec.EndTime = &now
+
+				// Update UI field for cancellation with i18n
+				e.updateUIFields(ctx, exec, "", getLocalizedMessage(locale, "cancelled"))
+
+				log.With(log.F{
+					"execution_id": exec.ID,
+					"member_id":    exec.MemberID,
+					"phase":        string(phase),
+				}).Info("Execution cancelled by user")
+
+				// Persist cancelled status
+				if !e.config.SkipPersistence && e.store != nil {
+					_ = e.store.UpdateStatus(ctx.Context, exec.ID, robottypes.ExecCancelled, "execution cancelled by user")
+				}
+				return exec, nil
+			}
+
+			// Normal failure case
 			exec.Status = robottypes.ExecFailed
 			exec.Error = err.Error()
 
@@ -228,7 +268,21 @@ func (e *Executor) Execute(ctx *robottypes.Context, robot *robottypes.Robot, tri
 }
 
 // runPhase executes a single phase
-func (e *Executor) runPhase(ctx *robottypes.Context, exec *robottypes.Execution, phase robottypes.Phase, data interface{}) error {
+func (e *Executor) runPhase(ctx *robottypes.Context, exec *robottypes.Execution, phase robottypes.Phase, data interface{}, control robottypes.ExecutionControl) error {
+	// Check if context is cancelled before starting this phase
+	select {
+	case <-ctx.Context.Done():
+		return robottypes.ErrExecutionCancelled
+	default:
+	}
+
+	// Wait if execution is paused (blocks until resumed or cancelled)
+	if control != nil {
+		if err := control.WaitIfPaused(); err != nil {
+			return err // Returns ErrExecutionCancelled if cancelled while paused
+		}
+	}
+
 	exec.Phase = phase
 
 	log.With(log.F{
@@ -422,6 +476,7 @@ var uiMessages = map[string]map[string]string{
 		"sending_delivery":    "Sending delivery...",
 		"learning_from_exec":  "Learning from execution...",
 		"completed":           "Completed",
+		"cancelled":           "Cancelled",
 		"failed_prefix":       "Failed at ",
 		"task_prefix":         "Task",
 		// Phase names for failure messages
@@ -445,6 +500,7 @@ var uiMessages = map[string]map[string]string{
 		"sending_delivery":    "正在发送...",
 		"learning_from_exec":  "学习执行经验...",
 		"completed":           "已完成",
+		"cancelled":           "已取消",
 		"failed_prefix":       "失败于",
 		"task_prefix":         "任务",
 		// Phase names for failure messages

agent/robot/executor/types/types.go

@@ -12,12 +12,22 @@ import (
 //   - DryRun:   Plan-only mode, simulates execution without Agent calls
 //   - Sandbox:  Isolated execution with resource limits and safety controls
 type Executor interface {
-	// Execute runs a robot through all applicable phases
+	// ExecuteWithControl runs a robot through all applicable phases with execution control
 	// ctx: Execution context with auth and logging
 	// robot: Robot configuration and state
 	// trigger: What triggered this execution (clock, human, event)
 	// data: Trigger-specific data (human input, event payload, etc.)
+	// execID: Pre-generated execution ID (empty string to auto-generate)
+	// control: Optional execution control for pause/resume functionality
 	// Returns: Execution record with all phase outputs
+	ExecuteWithControl(ctx *robottypes.Context, robot *robottypes.Robot, trigger robottypes.TriggerType, data interface{}, execID string, control robottypes.ExecutionControl) (*robottypes.Execution, error)
+
+	// ExecuteWithID runs a robot through all applicable phases with a pre-generated execution ID
+	// This is a convenience wrapper around ExecuteWithControl without control
+	ExecuteWithID(ctx *robottypes.Context, robot *robottypes.Robot, trigger robottypes.TriggerType, data interface{}, execID string) (*robottypes.Execution, error)
+
+	// Execute runs a robot through all applicable phases (auto-generates execution ID)
+	// This is a convenience wrapper around ExecuteWithControl
 	Execute(ctx *robottypes.Context, robot *robottypes.Robot, trigger robottypes.TriggerType, data interface{}) (*robottypes.Execution, error)
 
 	// Metrics and control

agent/robot/manager/integration_concurrent_test.go

@@ -545,12 +545,22 @@ type trackingExecutor struct {
 }
 
 func (e *trackingExecutor) Execute(ctx *types.Context, robot *types.Robot, trigger types.TriggerType, data interface{}) (*types.Execution, error) {
+	return e.ExecuteWithControl(ctx, robot, trigger, data, "", nil)
+}
+
+func (e *trackingExecutor) ExecuteWithID(ctx *types.Context, robot *types.Robot, trigger types.TriggerType, data interface{}, execID string) (*types.Execution, error) {
+	return e.ExecuteWithControl(ctx, robot, trigger, data, execID, nil)
+}
+
+func (e *trackingExecutor) ExecuteWithControl(ctx *types.Context, robot *types.Robot, trigger types.TriggerType, data interface{}, execID string, control types.ExecutionControl) (*types.Execution, error) {
 	if robot == nil {
 		return nil, types.ErrRobotNotFound
 	}
 
-	// Use unique ID for each execution to properly track quota
-	execID := fmt.Sprintf("exec_%d", time.Now().UnixNano())
+	// Use provided execID or generate unique ID for each execution to properly track quota
+	if execID == "" {
+		execID = fmt.Sprintf("exec_%d", time.Now().UnixNano())
+	}
 	exec := &types.Execution{
 		ID:          execID,
 		MemberID:    robot.MemberID,
@@ -604,12 +614,22 @@ type triggerTrackingExecutor struct {
 }
 
 func (e *triggerTrackingExecutor) Execute(ctx *types.Context, robot *types.Robot, trigger types.TriggerType, data interface{}) (*types.Execution, error) {
+	return e.ExecuteWithControl(ctx, robot, trigger, data, "", nil)
+}
+
+func (e *triggerTrackingExecutor) ExecuteWithID(ctx *types.Context, robot *types.Robot, trigger types.TriggerType, data interface{}, execID string) (*types.Execution, error) {
+	return e.ExecuteWithControl(ctx, robot, trigger, data, execID, nil)
+}
+
+func (e *triggerTrackingExecutor) ExecuteWithControl(ctx *types.Context, robot *types.Robot, trigger types.TriggerType, data interface{}, execID string, control types.ExecutionControl) (*types.Execution, error) {
 	if robot == nil {
 		return nil, types.ErrRobotNotFound
 	}
 
-	// Use unique ID for each execution to properly track quota
-	execID := fmt.Sprintf("exec_trigger_%s_%d", string(trigger), time.Now().UnixNano())
+	// Use provided execID or generate unique ID for each execution to properly track quota
+	if execID == "" {
+		execID = fmt.Sprintf("exec_trigger_%s_%d", string(trigger), time.Now().UnixNano())
+	}
 	exec := &types.Execution{
 		ID:          execID,
 		MemberID:    robot.MemberID,

agent/robot/manager/integration_control_test.go

@@ -495,12 +495,24 @@ type slowExecutor struct {
 }
 
 func (e *slowExecutor) Execute(ctx *types.Context, robot *types.Robot, trigger types.TriggerType, data interface{}) (*types.Execution, error) {
+	return e.ExecuteWithControl(ctx, robot, trigger, data, "", nil)
+}
+
+func (e *slowExecutor) ExecuteWithID(ctx *types.Context, robot *types.Robot, trigger types.TriggerType, data interface{}, execID string) (*types.Execution, error) {
+	return e.ExecuteWithControl(ctx, robot, trigger, data, execID, nil)
+}
+
+func (e *slowExecutor) ExecuteWithControl(ctx *types.Context, robot *types.Robot, trigger types.TriggerType, data interface{}, execID string, control types.ExecutionControl) (*types.Execution, error) {
 	if robot == nil {
 		return nil, types.ErrRobotNotFound
 	}
 
+	// Use provided execID or generate one
+	if execID == "" {
+		execID = "exec_slow_" + robot.MemberID
+	}
 	exec := &types.Execution{
-		ID:          "exec_slow_" + robot.MemberID,
+		ID:          execID,
 		MemberID:    robot.MemberID,
 		TeamID:      robot.TeamID,
 		TriggerType: trigger,