feat: surface workflow diagnostics

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
This commit is contained in:
David Kaya
2026-04-01 18:50:56 +02:00
co-authored by Copilot
parent b434dd86b4
commit 11b36827f5
12 changed files with 527 additions and 4 deletions
+1
View File
@@ -224,6 +224,7 @@ The protocol also carries **turn-scoped lifecycle events** alongside output delt
- **Session compaction events**: start and complete, with token-reduction metrics when infinite sessions trigger context trimming
- **Session usage events**: current token count and context-window limit from `session.usage_info` for context-bar rendering
- **Pending-messages-modified events**: emitted when mid-turn steering changes the pending message queue
- **Workflow diagnostic events**: normalized warnings and errors from Agent Framework (`WorkflowWarningEvent`, `WorkflowErrorEvent`, `ExecutorFailedEvent`) with optional executor or subworkflow metadata for richer debugging surfaces
These events flow through a single `onTurnScopedEvent` callback on the `runTurn` command, avoiding per-event-type callback proliferation. The main process maps each event to a `SessionEventRecord` and pushes it to the renderer, where lightweight state maps (activity, usage, turn-event log) consume them without touching the persisted workspace.
@@ -488,6 +488,19 @@ public sealed class PendingMessagesModifiedEventDto : SidecarEventDto
public string? AgentName { get; init; }
}
public sealed class WorkflowDiagnosticEventDto : SidecarEventDto
{
public string SessionId { get; init; } = string.Empty;
public string Severity { get; init; } = string.Empty;
public string DiagnosticKind { get; init; } = string.Empty;
public string Message { get; init; } = string.Empty;
public string? AgentId { get; init; }
public string? AgentName { get; init; }
public string? ExecutorId { get; init; }
public string? SubworkflowId { get; init; }
public string? ExceptionType { get; init; }
}
public sealed class SessionsListedEventDto : SidecarEventDto
{
public IReadOnlyList<CopilotSessionInfoDto> Sessions { get; init; } = [];
@@ -211,6 +211,12 @@ public sealed class CopilotWorkflowRunner : ITurnWorkflowRunner
return false;
}
if (TryCreateWorkflowDiagnosticEvent(command, evt, state, out WorkflowDiagnosticEventDto? diagnostic))
{
await onEvent(diagnostic).ConfigureAwait(false);
return false;
}
if (evt is AgentResponseUpdateEvent update)
{
await HandleAgentResponseUpdateAsync(command, update, state, onDelta, onEvent).ConfigureAwait(false);
@@ -341,6 +347,94 @@ public sealed class CopilotWorkflowRunner : ITurnWorkflowRunner
}
}
private static bool TryCreateWorkflowDiagnosticEvent(
RunTurnCommandDto command,
WorkflowEvent evt,
CopilotTurnExecutionState state,
out WorkflowDiagnosticEventDto diagnostic)
{
diagnostic = default!;
switch (evt)
{
case ExecutorFailedEvent executorFailed:
{
AgentIdentity? agent = AgentIdentityResolver.TryResolveObservedAgentIdentity(
command.Pattern,
executorFailed.ExecutorId,
state.ActiveAgent,
out AgentIdentity resolvedAgent)
? resolvedAgent
: null;
Exception? exception = executorFailed.Data;
diagnostic = new WorkflowDiagnosticEventDto
{
Type = "workflow-diagnostic",
RequestId = command.RequestId,
SessionId = command.SessionId,
Severity = "error",
DiagnosticKind = "executor-failed",
Message = ResolveDiagnosticMessage(exception, "Executor failed."),
AgentId = agent?.AgentId,
AgentName = agent?.AgentName,
ExecutorId = executorFailed.ExecutorId,
ExceptionType = exception?.GetBaseException().GetType().Name,
};
return true;
}
case WorkflowWarningEvent workflowWarning:
diagnostic = new WorkflowDiagnosticEventDto
{
Type = "workflow-diagnostic",
RequestId = command.RequestId,
SessionId = command.SessionId,
Severity = "warning",
DiagnosticKind = workflowWarning is SubworkflowWarningEvent
? "subworkflow-warning"
: "workflow-warning",
Message = ResolveDiagnosticMessage(workflowWarning.Data as string, "Workflow warning."),
SubworkflowId = workflowWarning is SubworkflowWarningEvent subworkflowWarning
? subworkflowWarning.SubWorkflowId
: null,
};
return true;
case WorkflowErrorEvent workflowError:
{
Exception? exception = workflowError.Exception;
diagnostic = new WorkflowDiagnosticEventDto
{
Type = "workflow-diagnostic",
RequestId = command.RequestId,
SessionId = command.SessionId,
Severity = "error",
DiagnosticKind = workflowError is SubworkflowErrorEvent
? "subworkflow-error"
: "workflow-error",
Message = ResolveDiagnosticMessage(exception, "Workflow failed."),
SubworkflowId = workflowError is SubworkflowErrorEvent subworkflowError
? subworkflowError.SubworkflowId
: null,
ExceptionType = exception?.GetBaseException().GetType().Name,
};
return true;
}
default:
return false;
}
}
private static string ResolveDiagnosticMessage(Exception? exception, string fallback)
{
return ResolveDiagnosticMessage(
exception?.GetBaseException().Message,
fallback);
}
private static string ResolveDiagnosticMessage(string? message, string fallback)
{
return string.IsNullOrWhiteSpace(message) ? fallback : message;
}
private static bool IsHandoffFunctionName(string? candidate)
{
return !string.IsNullOrWhiteSpace(candidate)
@@ -798,6 +798,117 @@ public sealed class CopilotWorkflowRunnerTests
});
}
[Fact]
public async Task HandleWorkflowEventAsync_EmitsExecutorFailedDiagnostic()
{
RunTurnCommandDto command = CreateApprovalCommand();
CopilotTurnExecutionState state = new(command);
List<WorkflowDiagnosticEventDto> diagnostics = [];
MethodInfo handleWorkflowEvent = typeof(CopilotWorkflowRunner).GetMethod(
"HandleWorkflowEventAsync",
BindingFlags.NonPublic | BindingFlags.Static)!;
Task<bool> handleTask = (Task<bool>)handleWorkflowEvent.Invoke(
null,
[
command,
new ExecutorFailedEvent("agent-1", new InvalidOperationException("Tool crashed.")),
Array.Empty<ChatMessage>(),
state,
(Func<TurnDeltaEventDto, Task>)(_ => Task.CompletedTask),
(Func<SidecarEventDto, Task>)(sidecarEvent =>
{
diagnostics.Add(Assert.IsType<WorkflowDiagnosticEventDto>(sidecarEvent));
return Task.CompletedTask;
}),
])!;
bool shouldEndTurn = await handleTask;
Assert.False(shouldEndTurn);
WorkflowDiagnosticEventDto diagnostic = Assert.Single(diagnostics);
Assert.Equal("workflow-diagnostic", diagnostic.Type);
Assert.Equal("error", diagnostic.Severity);
Assert.Equal("executor-failed", diagnostic.DiagnosticKind);
Assert.Equal("Tool crashed.", diagnostic.Message);
Assert.Equal("agent-1", diagnostic.AgentId);
Assert.Equal("Primary", diagnostic.AgentName);
Assert.Equal("agent-1", diagnostic.ExecutorId);
Assert.Equal("InvalidOperationException", diagnostic.ExceptionType);
}
[Fact]
public async Task HandleWorkflowEventAsync_EmitsWorkflowWarningDiagnostic()
{
RunTurnCommandDto command = CreateApprovalCommand();
CopilotTurnExecutionState state = new(command);
List<WorkflowDiagnosticEventDto> diagnostics = [];
MethodInfo handleWorkflowEvent = typeof(CopilotWorkflowRunner).GetMethod(
"HandleWorkflowEventAsync",
BindingFlags.NonPublic | BindingFlags.Static)!;
Task<bool> handleTask = (Task<bool>)handleWorkflowEvent.Invoke(
null,
[
command,
new WorkflowWarningEvent("Token budget is nearly exhausted."),
Array.Empty<ChatMessage>(),
state,
(Func<TurnDeltaEventDto, Task>)(_ => Task.CompletedTask),
(Func<SidecarEventDto, Task>)(sidecarEvent =>
{
diagnostics.Add(Assert.IsType<WorkflowDiagnosticEventDto>(sidecarEvent));
return Task.CompletedTask;
}),
])!;
bool shouldEndTurn = await handleTask;
Assert.False(shouldEndTurn);
WorkflowDiagnosticEventDto diagnostic = Assert.Single(diagnostics);
Assert.Equal("warning", diagnostic.Severity);
Assert.Equal("workflow-warning", diagnostic.DiagnosticKind);
Assert.Equal("Token budget is nearly exhausted.", diagnostic.Message);
Assert.Null(diagnostic.SubworkflowId);
Assert.Null(diagnostic.ExceptionType);
}
[Fact]
public async Task HandleWorkflowEventAsync_EmitsSubworkflowErrorDiagnostic()
{
RunTurnCommandDto command = CreateApprovalCommand();
CopilotTurnExecutionState state = new(command);
List<WorkflowDiagnosticEventDto> diagnostics = [];
MethodInfo handleWorkflowEvent = typeof(CopilotWorkflowRunner).GetMethod(
"HandleWorkflowEventAsync",
BindingFlags.NonPublic | BindingFlags.Static)!;
Task<bool> handleTask = (Task<bool>)handleWorkflowEvent.Invoke(
null,
[
command,
new SubworkflowErrorEvent("subworkflow-review", new InvalidOperationException("Reviewer agent failed.")),
Array.Empty<ChatMessage>(),
state,
(Func<TurnDeltaEventDto, Task>)(_ => Task.CompletedTask),
(Func<SidecarEventDto, Task>)(sidecarEvent =>
{
diagnostics.Add(Assert.IsType<WorkflowDiagnosticEventDto>(sidecarEvent));
return Task.CompletedTask;
}),
])!;
bool shouldEndTurn = await handleTask;
Assert.False(shouldEndTurn);
WorkflowDiagnosticEventDto diagnostic = Assert.Single(diagnostics);
Assert.Equal("error", diagnostic.Severity);
Assert.Equal("subworkflow-error", diagnostic.DiagnosticKind);
Assert.Equal("Reviewer agent failed.", diagnostic.Message);
Assert.Equal("subworkflow-review", diagnostic.SubworkflowId);
Assert.Equal("InvalidOperationException", diagnostic.ExceptionType);
}
[Fact]
public void RequiresToolCallApproval_HonorsAutoApprovedToolNames()
{
@@ -232,6 +232,77 @@ public sealed class SidecarProtocolHostTests
});
}
[Fact]
public async Task RunTurnCommand_ReturnsWorkflowDiagnosticEventsAndCompletion()
{
SidecarProtocolHost host = new(
new PatternValidator(),
new FakeWorkflowRunner(async (command, onDelta, onActivity, onApproval, onUserInput, onMcpOAuthRequired, onExitPlanMode, cancellationToken) =>
{
await onActivity(new WorkflowDiagnosticEventDto
{
Type = "workflow-diagnostic",
RequestId = command.RequestId,
SessionId = command.SessionId,
Severity = "error",
DiagnosticKind = "executor-failed",
Message = "Tool crashed.",
AgentId = "agent-1",
AgentName = "Primary",
ExecutorId = "agent-1",
ExceptionType = "InvalidOperationException",
});
return [];
}));
IReadOnlyList<JsonElement> events = await RunHostAsync(
new RunTurnCommandDto
{
Type = "run-turn",
RequestId = "turn-diagnostic",
SessionId = "session-1",
ProjectPath = "C:\\workspace\\project",
Pattern = new PatternDefinitionDto
{
Id = "pattern-1",
Name = "Single Agent",
Mode = "single",
Availability = "available",
Agents =
[
CreateAgent(name: "Primary"),
],
},
Messages = [],
},
host);
Assert.Collection(
events,
diagnosticEvent =>
{
Assert.Equal("workflow-diagnostic", diagnosticEvent.GetProperty("type").GetString());
Assert.Equal("turn-diagnostic", diagnosticEvent.GetProperty("requestId").GetString());
Assert.Equal("session-1", diagnosticEvent.GetProperty("sessionId").GetString());
Assert.Equal("error", diagnosticEvent.GetProperty("severity").GetString());
Assert.Equal("executor-failed", diagnosticEvent.GetProperty("diagnosticKind").GetString());
Assert.Equal("Tool crashed.", diagnosticEvent.GetProperty("message").GetString());
Assert.Equal("agent-1", diagnosticEvent.GetProperty("executorId").GetString());
},
completionEvent =>
{
Assert.Equal("turn-complete", completionEvent.GetProperty("type").GetString());
Assert.Equal("session-1", completionEvent.GetProperty("sessionId").GetString());
Assert.False(completionEvent.GetProperty("cancelled").GetBoolean());
},
commandCompleteEvent =>
{
Assert.Equal("command-complete", commandCompleteEvent.GetProperty("type").GetString());
Assert.Equal("turn-diagnostic", commandCompleteEvent.GetProperty("requestId").GetString());
});
}
[Fact]
public async Task RunTurnCommand_DeserializesInteractionMode()
{
+15
View File
@@ -2504,6 +2504,21 @@ export class AryxAppService extends EventEmitter<AppServiceEvents> {
agentName: event.agentName,
});
return;
case 'workflow-diagnostic':
this.emitSessionEvent({
sessionId,
kind: 'workflow-diagnostic',
occurredAt,
agentId: event.agentId,
agentName: event.agentName,
diagnosticSeverity: event.severity,
diagnosticKind: event.diagnosticKind,
diagnosticMessage: event.message,
executorId: event.executorId,
subworkflowId: event.subworkflowId,
exceptionType: event.exceptionType,
});
return;
case 'assistant-usage':
this.emitSessionEvent({
sessionId,
+3 -1
View File
@@ -15,6 +15,7 @@ import type {
AssistantUsageEvent,
AssistantIntentEvent,
ReasoningDeltaEvent,
WorkflowDiagnosticEvent,
} from '@shared/contracts/sidecar';
import type { ChatMessageRecord } from '@shared/domain/session';
@@ -27,7 +28,8 @@ export type TurnScopedEvent =
| PendingMessagesModifiedEvent
| AssistantUsageEvent
| AssistantIntentEvent
| ReasoningDeltaEvent;
| ReasoningDeltaEvent
| WorkflowDiagnosticEvent;
export interface RunTurnPendingCommand {
kind: 'run-turn';
+1
View File
@@ -454,6 +454,7 @@ export class SidecarClient {
case 'session-usage':
case 'session-compaction':
case 'pending-messages-modified':
case 'workflow-diagnostic':
case 'assistant-usage':
case 'assistant-intent':
case 'reasoning-delta':
+23
View File
@@ -384,6 +384,28 @@ export interface PendingMessagesModifiedEvent {
agentName?: string;
}
export type WorkflowDiagnosticSeverity = 'warning' | 'error';
export type WorkflowDiagnosticKind =
| 'workflow-warning'
| 'workflow-error'
| 'executor-failed'
| 'subworkflow-warning'
| 'subworkflow-error';
export interface WorkflowDiagnosticEvent {
type: 'workflow-diagnostic';
requestId: string;
sessionId: string;
severity: WorkflowDiagnosticSeverity;
diagnosticKind: WorkflowDiagnosticKind;
message: string;
agentId?: string;
agentName?: string;
executorId?: string;
subworkflowId?: string;
exceptionType?: string;
}
export interface CopilotSessionInfo {
copilotSessionId: string;
managedByAryx: boolean;
@@ -563,6 +585,7 @@ export type SidecarEvent =
| SessionUsageEvent
| SessionCompactionEvent
| PendingMessagesModifiedEvent
| WorkflowDiagnosticEvent
| ApprovalRequestedEvent
| UserInputRequestedEvent
| McpOauthRequiredEvent
+16 -2
View File
@@ -1,7 +1,12 @@
import type { SessionRunRecord } from '@shared/domain/runTimeline';
import type { ChatMessageKind } from '@shared/domain/session';
import type { QuotaSnapshot, ToolCallFileChangePreview } from '@shared/contracts/sidecar';
import type {
QuotaSnapshot,
ToolCallFileChangePreview,
WorkflowDiagnosticKind,
WorkflowDiagnosticSeverity,
} from '@shared/contracts/sidecar';
export type SessionActivityType = 'thinking' | 'tool-calling' | 'handoff' | 'completed';
@@ -19,7 +24,8 @@ export type SessionEventKind =
| 'session-usage'
| 'session-compaction'
| 'pending-messages-modified'
| 'assistant-usage';
| 'assistant-usage'
| 'workflow-diagnostic';
export type SubagentEventKind = 'started' | 'completed' | 'failed' | 'selected' | 'deselected';
@@ -86,4 +92,12 @@ export interface SessionEventRecord {
usageDuration?: number;
usageTotalNanoAiu?: number;
usageQuotaSnapshots?: Record<string, QuotaSnapshot>;
// Workflow diagnostic fields
diagnosticSeverity?: WorkflowDiagnosticSeverity;
diagnosticKind?: WorkflowDiagnosticKind;
diagnosticMessage?: string;
executorId?: string;
subworkflowId?: string;
exceptionType?: string;
}
@@ -0,0 +1,82 @@
import { describe, expect, mock, test } from 'bun:test';
import type { SessionEventRecord } from '@shared/domain/event';
import type { WorkflowDiagnosticEvent } from '@shared/contracts/sidecar';
mock.module('electron', () => {
const electronMock = {
app: {
isPackaged: false,
getAppPath: () => 'C:\\workspace\\personal\\repositories\\aryx',
getPath: () => 'C:\\workspace\\personal\\repositories\\aryx\\tests\\fixtures',
},
dialog: {
showOpenDialog: async () => ({ canceled: true, filePaths: [] }),
},
shell: {
openPath: async () => '',
},
};
return {
...electronMock,
default: electronMock,
};
});
mock.module('keytar', () => ({
default: {
getPassword: async () => null,
setPassword: async () => undefined,
deletePassword: async () => false,
},
}));
const { AryxAppService } = await import('@main/AryxAppService');
describe('AryxAppService workflow diagnostics', () => {
test('maps turn-scoped workflow diagnostics to session events', async () => {
const service = new AryxAppService();
const captured: SessionEventRecord[] = [];
const workflowDiagnostic: WorkflowDiagnosticEvent = {
type: 'workflow-diagnostic',
requestId: 'turn-1',
sessionId: 'session-1',
severity: 'error',
diagnosticKind: 'executor-failed',
message: 'Tool crashed.',
agentId: 'agent-1',
agentName: 'Primary',
executorId: 'agent-1',
exceptionType: 'InvalidOperationException',
};
const internals = service as unknown as {
emitSessionEvent: (event: SessionEventRecord) => void;
handleTurnScopedEvent: (
workspace: unknown,
sessionId: string,
event: WorkflowDiagnosticEvent,
) => void | Promise<void>;
};
internals.emitSessionEvent = (event) => {
captured.push(event);
};
await internals.handleTurnScopedEvent({}, 'session-1', workflowDiagnostic);
expect(captured).toHaveLength(1);
expect(captured[0]).toMatchObject({
sessionId: 'session-1',
kind: 'workflow-diagnostic',
agentId: 'agent-1',
agentName: 'Primary',
diagnosticSeverity: 'error',
diagnosticKind: 'executor-failed',
diagnosticMessage: 'Tool crashed.',
executorId: 'agent-1',
exceptionType: 'InvalidOperationException',
});
expect(captured[0]?.occurredAt).toEqual(expect.any(String));
});
});
+97 -1
View File
@@ -2,7 +2,7 @@ import { EventEmitter } from 'node:events';
import { describe, expect, mock, test } from 'bun:test';
import type { SidecarCapabilities } from '@shared/contracts/sidecar';
import type { RunTurnCommand, SidecarCapabilities, WorkflowDiagnosticEvent } from '@shared/contracts/sidecar';
class FakeReadableStream extends EventEmitter {
setEncoding(_encoding: BufferEncoding): void {}
@@ -142,4 +142,100 @@ describe('SidecarClient', () => {
spawnedProcesses[1]!.completeExit();
await finalDispose;
});
test('routes workflow diagnostic events through the turn-scoped callback', async () => {
spawnedProcesses.length = 0;
const client = new SidecarClient();
const diagnostics: WorkflowDiagnosticEvent[] = [];
const command: RunTurnCommand = {
type: 'run-turn',
requestId: 'turn-1',
sessionId: 'session-1',
projectPath: 'C:\\workspace\\project',
pattern: {
id: 'pattern-1',
name: 'Single Agent',
description: '',
mode: 'single',
availability: 'available',
maxIterations: 1,
agents: [
{
id: 'agent-1',
name: 'Primary',
description: '',
instructions: 'Help with the request.',
model: 'gpt-5.4',
},
],
createdAt: '2026-04-01T00:00:00.000Z',
updatedAt: '2026-04-01T00:00:00.000Z',
},
messages: [],
};
const turn = client.runTurn(
command,
async () => undefined,
async () => undefined,
async () => undefined,
async () => undefined,
async () => undefined,
async () => undefined,
async () => undefined,
async (event) => {
if (event.type === 'workflow-diagnostic') {
diagnostics.push(event);
}
},
);
await Promise.resolve();
expect(spawnedProcesses).toHaveLength(1);
spawnedProcesses[0]!.emitStdout(
`${JSON.stringify({
type: 'workflow-diagnostic',
requestId: command.requestId,
sessionId: command.sessionId,
severity: 'error',
diagnosticKind: 'executor-failed',
message: 'Tool crashed.',
agentId: 'agent-1',
agentName: 'Primary',
executorId: 'agent-1',
exceptionType: 'InvalidOperationException',
} satisfies WorkflowDiagnosticEvent)}\n`,
);
spawnedProcesses[0]!.emitStdout(
`${JSON.stringify({
type: 'turn-complete',
requestId: command.requestId,
sessionId: command.sessionId,
messages: [],
cancelled: false,
})}\n`,
);
spawnedProcesses[0]!.emitStdout(
`${JSON.stringify({
type: 'command-complete',
requestId: command.requestId,
})}\n`,
);
await expect(turn).resolves.toEqual([]);
expect(diagnostics).toEqual([
expect.objectContaining({
type: 'workflow-diagnostic',
severity: 'error',
diagnosticKind: 'executor-failed',
message: 'Tool crashed.',
executorId: 'agent-1',
}),
]);
const dispose = client.dispose();
spawnedProcesses[0]!.completeExit();
await dispose;
});
});