🙌 OpenHands: AI-Driven Development
feat: add Debug with OpenHands button to failed automation runs (#1603)
Hiep Le8a4be25a7a55
diff --git a/.pr/app2499-installed-mcp-card-configured-name.png b/.pr/app2499-installed-mcp-card-configured-name.pngdeleted file mode 100644index 7345378db5..0000000000Binary files a/.pr/app2499-installed-mcp-card-configured-name.png and /dev/null differdiff --git a/.pr/app2499-installed-mcp-card-fresh-configured-name.png b/.pr/app2499-installed-mcp-card-fresh-configured-name.pngdeleted file mode 100644index e3e93cf295..0000000000Binary files a/.pr/app2499-installed-mcp-card-fresh-configured-name.png and /dev/null differdiff --git a/.pr/app2499-installed-mcp-card-marketplace-or-url-fallback.png b/.pr/app2499-installed-mcp-card-marketplace-or-url-fallback.pngdeleted file mode 100644index cc19411c25..0000000000Binary files a/.pr/app2499-installed-mcp-card-marketplace-or-url-fallback.png and /dev/null differdiff --git a/.pr/app2499-installed-mcp-section-overview.png b/.pr/app2499-installed-mcp-section-overview.pngdeleted file mode 100644index 25cdd0a906..0000000000Binary files a/.pr/app2499-installed-mcp-section-overview.png and /dev/null differdiff --git a/__tests__/components/automations/detail/debug-automation-button.test.tsx b/__tests__/components/automations/detail/debug-automation-button.test.tsxnew file mode 100644index 0000000000..c57e9c02e4--- /dev/null+++ b/__tests__/components/automations/detail/debug-automation-button.test.tsx@@ -0,0 +1,159 @@+import { describe, it, expect, vi, beforeEach, afterEach } from "vitest";+import { render, screen, fireEvent, waitFor } from "@testing-library/react";+import { QueryClient, QueryClientProvider } from "@tanstack/react-query";++import { DebugAutomationButton } from "#/components/features/automations/detail/debug-automation-button";+import AgentServerConversationService from "#/api/conversation-service/agent-server-conversation-service.api";+import ProfilesService from "#/api/profiles-service/profiles-service.api";+import { ActiveBackendProvider } from "#/contexts/active-backend-context";+import {+ NavigationProvider,+ type NavigationContextValue,+} from "#/context/navigation-context";+import {+ __resetActiveStoreForTests,+ setActiveSelection,+ setRegisteredBackends,+} from "#/api/backend-registry/active-store";+import type { Backend } from "#/api/backend-registry/types";+import {+ AutomationRunStatus,+ type Automation,+ type AutomationRun,+} from "#/types/automation";++// Mock the underlying services (not the useCreateConversation hook), so the+// real hook logic — id mapping + onSuccess navigation — is exercised.+vi.mock(+ "#/api/conversation-service/agent-server-conversation-service.api",+ () => ({+ default: { createConversation: vi.fn() },+ }),+);++vi.mock("#/api/profiles-service/profiles-service.api", () => ({+ default: { listProfiles: vi.fn() },+}));++type CreatedConversation = Awaited<+ ReturnType<typeof AgentServerConversationService.createConversation>+>;++const localBackend: Backend = {+ id: "local-1",+ name: "Local",+ host: "http://localhost:8000",+ apiKey: "session-key",+ kind: "local",+};++const failedRun: AutomationRun = {+ id: "run-9",+ status: AutomationRunStatus.FAILED,+ conversation_id: "conv-1",+ bash_command_id: "cmd-1",+ error_detail: "Process exited with code 1",+ started_at: "2026-01-01T10:00:00Z",+ completed_at: "2026-01-01T10:02:00Z",+};++const automation: Automation = {+ id: "auto-1",+ name: "Daily Jira digest",+ prompt: "Summarize new Jira issues",+ trigger: { type: "cron", schedule: "0 9 * * *" },+ enabled: true,+ created_at: "2026-01-01T00:00:00Z",+ updated_at: "2026-01-01T00:00:00Z",+};++function renderButton(stderr = "HTTP Error 410: Gone") {+ const navigate = vi.fn();+ const navValue: NavigationContextValue = {+ currentPath: "/automations/auto-1",+ conversationId: null,+ isNavigating: false,+ navigate,+ };+ const queryClient = new QueryClient({+ defaultOptions: { queries: { retry: false }, mutations: { retry: false } },+ });+ render(+ <QueryClientProvider client={queryClient}>+ <ActiveBackendProvider>+ <NavigationProvider value={navValue}>+ <DebugAutomationButton+ run={failedRun}+ automation={automation}+ stderr={stderr}+ />+ </NavigationProvider>+ </ActiveBackendProvider>+ </QueryClientProvider>,+ );+ return { navigate };+}++beforeEach(() => {+ vi.clearAllMocks();+ __resetActiveStoreForTests();+ setRegisteredBackends([localBackend]);+ setActiveSelection({ backendId: localBackend.id });+ vi.mocked(ProfilesService.listProfiles).mockResolvedValue({+ profiles: [],+ active_profile: null,+ });+});++afterEach(() => {+ __resetActiveStoreForTests();+});++describe("DebugAutomationButton", () => {+ it("seeds a new conversation with the run's error details and navigates to it", async () => {+ // Arrange — a null app_conversation_id resolves to a `task-<id>` route.+ vi.mocked(+ AgentServerConversationService.createConversation,+ ).mockResolvedValue({+ id: "abc123",+ app_conversation_id: null,+ agent_server_url: null,+ } as CreatedConversation);+ const { navigate } = renderButton("HTTP Error 410: Gone");++ // Act+ fireEvent.click(screen.getByTestId("debug-automation-button"));++ // Assert — the first arg is the initial user message (the debug prompt).+ await waitFor(() =>+ expect(+ AgentServerConversationService.createConversation,+ ).toHaveBeenCalledTimes(1),+ );+ const query = vi.mocked(AgentServerConversationService.createConversation)+ .mock.calls[0][0];+ expect(query).toContain("HTTP Error 410: Gone");+ expect(query).toContain("Daily Jira digest");++ await waitFor(() =>+ expect(navigate).toHaveBeenCalledWith("/conversations/task-abc123"),+ );+ });++ it("disables the button while the conversation is being created", async () => {+ // Arrange — a never-settling promise keeps the mutation pending.+ vi.mocked(+ AgentServerConversationService.createConversation,+ ).mockReturnValue(new Promise<CreatedConversation>(() => {}));+ renderButton();++ const button = screen.getByTestId("debug-automation-button");++ // Act+ fireEvent.click(button);++ // Assert+ await waitFor(() => expect(button).toBeDisabled());+ expect(button).toHaveAttribute("aria-busy", "true");+ });+});diff --git a/__tests__/components/automations/detail/run-logs-modal.test.tsx b/__tests__/components/automations/detail/run-logs-modal.test.tsxindex 5b30f3dc7c..20a500b0de 100644--- a/__tests__/components/automations/detail/run-logs-modal.test.tsx+++ b/__tests__/components/automations/detail/run-logs-modal.test.tsx@@ -2,6 +2,7 @@ import { describe, it, expect, vi, beforeEach, afterEach } from "vitest";import { render, screen, fireEvent } from "@testing-library/react";import { I18nKey } from "#/i18n/declaration";import { RunLogsModal } from "#/components/features/automations/detail/run-logs-modal";+import { AutomationRunStatus, type AutomationRun } from "#/types/automation";const { useBashCommandLogsMock } = vi.hoisted(() => ({useBashCommandLogsMock: vi.fn(),@@ -11,6 +12,17 @@ vi.mock("#/hooks/query/use-bash-command-logs", () => ({useBashCommandLogs: useBashCommandLogsMock,}));+// The debug button owns its own query/navigation stack; stub it so these tests+// only verify whether the modal chooses to render it.+vi.mock(+ "#/components/features/automations/detail/debug-automation-button",+ () => ({+ DebugAutomationButton: ({ run }: { run: AutomationRun }) => (+ <div data-testid="debug-automation-button-stub" data-run-id={run.id} />+ ),+ }),+);+function makeHookResult(overrides: Partial<ReturnType<typeof baseResult>> = {},) {@@ -274,3 +286,47 @@ describe("RunLogsModal", () => {expect(onClose).toHaveBeenCalledTimes(1);});});++describe("RunLogsModal — Debug with OpenHands button", () => {+ const makeRun = (status: AutomationRunStatus): AutomationRun => ({+ id: "run-1",+ status,+ conversation_id: "conv-1",+ bash_command_id: "cmd-1",+ error_detail: status === AutomationRunStatus.FAILED ? "boom" : null,+ started_at: "2026-01-01T10:00:00Z",+ completed_at: "2026-01-01T10:02:00Z",+ });++ it("renders the debug button for a failed run", () => {+ useBashCommandLogsMock.mockReturnValue(makeHookResult());+ render(+ <RunLogsModal+ isOpen+ conversationId="conv-1"+ bashCommandId="cmd-1"+ onClose={() => {}}+ run={makeRun(AutomationRunStatus.FAILED)}+ />,+ );+ expect(+ screen.getByTestId("debug-automation-button-stub"),+ ).toBeInTheDocument();+ });++ it("does not render the debug button for a successful run", () => {+ useBashCommandLogsMock.mockReturnValue(makeHookResult());+ render(+ <RunLogsModal+ isOpen+ conversationId="conv-1"+ bashCommandId="cmd-1"+ onClose={() => {}}+ run={makeRun(AutomationRunStatus.COMPLETED)}+ />,+ );+ expect(+ screen.queryByTestId("debug-automation-button-stub"),+ ).not.toBeInTheDocument();+ });+});diff --git a/__tests__/utils/automation-debug-prompt.test.ts b/__tests__/utils/automation-debug-prompt.test.tsnew file mode 100644index 0000000000..3af5012e83--- /dev/null+++ b/__tests__/utils/automation-debug-prompt.test.ts@@ -0,0 +1,89 @@+import { describe, it, expect } from "vitest";+import { buildAutomationDebugPrompt } from "#/utils/automation-debug-prompt";++describe("buildAutomationDebugPrompt", () => {+ it("includes the automation name, its instructions, the stderr, and the run id", () => {+ // Arrange+ const input = {+ automationName: "Daily Jira digest",+ automationPrompt: "Summarize new Jira issues each morning",+ errorDetail: "Process exited with code 1",+ stderr: "Traceback ...\nHTTP Error 410: Gone",+ runId: "run-42",+ };++ // Act+ const prompt = buildAutomationDebugPrompt(input);++ // Assert+ expect(prompt).toContain("Daily Jira digest");+ expect(prompt).toContain("Summarize new Jira issues each morning");+ expect(prompt).toContain("HTTP Error 410: Gone");+ expect(prompt).toContain("run-42");+ // The live stderr is preferred over the run-level error_detail.+ expect(prompt).not.toContain("Process exited with code 1");+ });++ it("falls back to error_detail when stderr is blank", () => {+ // Arrange — sandbox gone, so no stderr was fetched.+ const prompt = buildAutomationDebugPrompt({+ automationName: "Daily Jira digest",+ automationPrompt: "Summarize new Jira issues",+ errorDetail: "Process exited with code 1",+ stderr: " \n ",+ runId: "run-42",+ });++ // Assert+ expect(prompt).toContain("Process exited with code 1");+ expect(prompt).not.toContain("No error output");+ });++ it("omits the instructions section when the automation prompt is absent", () => {+ // Act+ const prompt = buildAutomationDebugPrompt({+ automationName: "Daily Jira digest",+ automationPrompt: null,+ errorDetail: "boom",+ stderr: "",+ runId: "run-42",+ });++ // Assert+ expect(prompt).not.toContain("What the automation was set up to do");+ });++ it("reports missing output when neither stderr nor error_detail is available", () => {+ // Act+ const prompt = buildAutomationDebugPrompt({+ automationName: "Daily Jira digest",+ automationPrompt: "Summarize new Jira issues",+ errorDetail: null,+ stderr: "",+ runId: "run-42",+ });++ // Assert+ expect(prompt).toContain("No error output was captured");+ });++ it("truncates oversized stderr, keeping the tail", () => {+ // Arrange — an error stream well over the 4000-char cap, with distinct+ // head and tail markers so we can prove which end survives.+ const stderr = `HEAD_MARKER${"x".repeat(5000)}TAIL_MARKER`;++ // Act+ const prompt = buildAutomationDebugPrompt({+ automationName: "Daily Jira digest",+ automationPrompt: "Summarize new Jira issues",+ errorDetail: null,+ stderr,+ runId: "run-42",+ });++ // Assert — the tail (the actual error) is kept, the head is dropped.+ expect(prompt).toContain("TAIL_MARKER");+ expect(prompt).toContain("…(truncated)…");+ expect(prompt).not.toContain("HEAD_MARKER");+ });+});diff --git a/src/components/features/automations/detail/activity-log-item.tsx b/src/components/features/automations/detail/activity-log-item.tsxindex 715e31ed46..c56735389d 100644--- a/src/components/features/automations/detail/activity-log-item.tsx+++ b/src/components/features/automations/detail/activity-log-item.tsx@@ -2,12 +2,17 @@ import { useState } from "react";import { useTranslation } from "react-i18next";import { I18nKey } from "#/i18n/declaration";import TerminalIcon from "#/icons/terminal.svg?react";-import { AutomationRunStatus, type AutomationRun } from "#/types/automation";+import {+ AutomationRunStatus,+ type Automation,+ type AutomationRun,+} from "#/types/automation";import { RunStatusBadge } from "./run-status-badge";import { RunLogsModal } from "./run-logs-modal";interface ActivityLogItemProps {run: AutomationRun;+ automation?: Automation;}function formatRunTimestamp(dateStr: string, locale: string): string {@@ -33,7 +38,7 @@ function getConversationUrl(conversationId: string): string {return `/conversations/${conversationId}`;}-export function ActivityLogItem({ run }: ActivityLogItemProps) {+export function ActivityLogItem({ run, automation }: ActivityLogItemProps) {const { t, i18n } = useTranslation("openhands");const hasConversation = !!run.conversation_id;const hasBashCommand = !!run.bash_command_id;@@ -126,6 +131,8 @@ export function ActivityLogItem({ run }: ActivityLogItemProps) {bashCommandId={run.bash_command_id}isOpen={logsOpen}onClose={() => setLogsOpen(false)}+ run={run}+ automation={automation}/>)}</>diff --git a/src/components/features/automations/detail/activity-log-section.tsx b/src/components/features/automations/detail/activity-log-section.tsxindex d6c552506e..37712273ef 100644--- a/src/components/features/automations/detail/activity-log-section.tsx+++ b/src/components/features/automations/detail/activity-log-section.tsx@@ -4,18 +4,19 @@ import { I18nKey } from "#/i18n/declaration";import { useAutomationRuns } from "#/hooks/query/use-automation-detail";import ActivityIcon from "#/icons/activity.svg?react";import { ActivityLogItem } from "./activity-log-item";+import type { Automation } from "#/types/automation";interface ActivityLogSectionProps {- automationId: string;+ automation: Automation;}const PAGE_SIZE = 20;-export function ActivityLogSection({ automationId }: ActivityLogSectionProps) {+export function ActivityLogSection({ automation }: ActivityLogSectionProps) {const { t } = useTranslation("openhands");const [limit, setLimit] = useState(PAGE_SIZE);const { data, isLoading } = useAutomationRuns({- id: automationId,+ id: automation.id,limit,offset: 0,});@@ -60,7 +61,7 @@ export function ActivityLogSection({ automationId }: ActivityLogSectionProps) {key={run.id}className={index > 0 ? "border-t border-[var(--oh-border)]" : ""}>- <ActivityLogItem run={run} />+ <ActivityLogItem run={run} automation={automation} /></div>))}diff --git a/src/components/features/automations/detail/debug-automation-button.tsx b/src/components/features/automations/detail/debug-automation-button.tsxnew file mode 100644index 0000000000..f3c6248b16--- /dev/null+++ b/src/components/features/automations/detail/debug-automation-button.tsx@@ -0,0 +1,66 @@+import { useTranslation } from "react-i18next";+import { I18nKey } from "#/i18n/declaration";+import { BrandButton } from "#/components/features/settings/brand-button";+import { useCreateConversation } from "#/hooks/mutation/use-create-conversation";+import { useNavigation } from "#/context/navigation-context";+import { buildAutomationDebugPrompt } from "#/utils/automation-debug-prompt";+import type { Automation, AutomationRun } from "#/types/automation";++interface DebugAutomationButtonProps {+ /** The failed run being debugged. */+ run: AutomationRun;+ /** The parent automation, used to add context to the debug prompt. */+ automation?: Automation;+ /** The run's stderr, already computed by the enclosing logs modal. */+ stderr: string;+}++/**+ * Starts a new OpenHands conversation seeded with the failed run's error+ * details so the agent begins debugging immediately. Kept as its own component+ * (rather than inlined into the logs modal) so `useCreateConversation` — which+ * requires a QueryClientProvider — only mounts for failed runs.+ */+export function DebugAutomationButton({+ run,+ automation,+ stderr,+}: DebugAutomationButtonProps) {+ const { t } = useTranslation("openhands");+ const { navigate } = useNavigation();+ const createConversation = useCreateConversation();++ const handleClick = () => {+ if (createConversation.isPending) return;++ const query = buildAutomationDebugPrompt({+ automationName: automation?.name,+ automationPrompt: automation?.prompt,+ errorDetail: run.error_detail,+ stderr,+ runId: run.id,+ });++ createConversation.mutate(+ { query },+ {+ onSuccess: (data) => {+ navigate(`/conversations/${data.conversation_id}`);+ },+ },+ );+ };++ return (+ <BrandButton+ testId="debug-automation-button"+ type="button"+ variant="primary"+ onClick={handleClick}+ isDisabled={createConversation.isPending}+ aria-busy={createConversation.isPending}+ >+ {t(I18nKey.AUTOMATIONS$DETAIL$DEBUG_WITH_OPENHANDS)}+ </BrandButton>+ );+}diff --git a/src/components/features/automations/detail/run-logs-modal.tsx b/src/components/features/automations/detail/run-logs-modal.tsxindex 88df46c947..edda217f9e 100644--- a/src/components/features/automations/detail/run-logs-modal.tsx+++ b/src/components/features/automations/detail/run-logs-modal.tsx@@ -9,6 +9,12 @@ import {import type { BashOutput } from "@openhands/typescript-client";import { cn } from "#/utils/utils";import { modalTitleLgMediumClassName } from "#/utils/modal-classes";+import {+ AutomationRunStatus,+ type Automation,+ type AutomationRun,+} from "#/types/automation";+import { DebugAutomationButton } from "./debug-automation-button";/*** Localized empty-state message key for each `SandboxIssue` reason.@@ -31,6 +37,10 @@ interface RunLogsModalProps {bashCommandId: string | null;isOpen: boolean;onClose: () => void;+ /** The run these logs belong to; enables the debug action for failed runs. */+ run?: AutomationRun;+ /** The parent automation, used to add context to the debug prompt. */+ automation?: Automation;}function concatStream(outputs: BashOutput[], key: "stdout" | "stderr"): string {@@ -52,6 +62,8 @@ export function RunLogsModal({bashCommandId,isOpen,onClose,+ run,+ automation,}: RunLogsModalProps) {const { t } = useTranslation("openhands");const [activeTab, setActiveTab] = useState<LogTab>("stdout");@@ -231,6 +243,16 @@ export function RunLogsModal({</pre>)}</div>++ {run?.status === AutomationRunStatus.FAILED && (+ <div className="mt-4 flex justify-end">+ <DebugAutomationButton+ run={run}+ automation={automation}+ stderr={stderr}+ />+ </div>+ )}</div></div>);diff --git a/src/i18n/translation.json b/src/i18n/translation.jsonindex 23a31d5fcf..da5ec91a06 100644--- a/src/i18n/translation.json+++ b/src/i18n/translation.json@@ -13667,6 +13667,23 @@"uk": "Переглянути журнали","ca": "Mostra els registres"},+ "AUTOMATIONS$DETAIL$DEBUG_WITH_OPENHANDS": {+ "en": "Debug with OpenHands",+ "ja": "OpenHandsでデバッグ",+ "zh-CN": "使用 OpenHands 调试",+ "zh-TW": "使用 OpenHands 偵錯",+ "ko-KR": "OpenHands로 디버그",+ "no": "Feilsøk med OpenHands",+ "it": "Esegui il debug con OpenHands",+ "pt": "Depurar com OpenHands",+ "es": "Depurar con OpenHands",+ "ar": "التصحيح باستخدام OpenHands",+ "fr": "Déboguer avec OpenHands",+ "tr": "OpenHands ile hata ayıkla",+ "de": "Mit OpenHands debuggen",+ "uk": "Налагодити за допомогою OpenHands",+ "ca": "Depura amb OpenHands"+ },"AUTOMATIONS$DETAIL$LOGS_TITLE": {"en": "Logs","ja": "ログ",diff --git a/src/routes/automation-detail.tsx b/src/routes/automation-detail.tsxindex 562cc96eaa..76c6d57810 100644--- a/src/routes/automation-detail.tsx+++ b/src/routes/automation-detail.tsx@@ -189,7 +189,7 @@ export default function AutomationDetail() {createdAt={automation.created_at}lastRunAt={automation.last_triggered_at}/>- <ActivityLogSection automationId={automation.id} />+ <ActivityLogSection automation={automation} /><DeleteConfirmationModalautomationName={automation.name}isOpen={showDeleteModal}diff --git a/src/utils/automation-debug-prompt.ts b/src/utils/automation-debug-prompt.tsnew file mode 100644index 0000000000..05a1455464--- /dev/null+++ b/src/utils/automation-debug-prompt.ts@@ -0,0 +1,65 @@+export interface AutomationDebugPromptInput {+ /** Human-readable automation name, if known. */+ automationName?: string;+ /** The automation's instructions (what it was set up to do). */+ automationPrompt?: string | null;+ /** Run-level error string; the durable fallback when stderr is unavailable. */+ errorDetail?: string | null;+ /** The run's captured stderr (the traceback shown in the Error tab). */+ stderr?: string | null;+ /** The failed run's id, included as a reference. */+ runId: string;+}++// Keep the seeded first message bounded. A traceback's most useful part is the+// tail (the actual error), so when the captured output is huge we keep the end.+const MAX_ERROR_CHARS = 4000;++function keepTail(text: string, max: number): string {+ if (text.length <= max) return text;+ return `…(truncated)…\n${text.slice(text.length - max)}`;+}++/**+ * Build the first message for a "Debug with OpenHands" conversation started+ * from a failed automation run. This is an instruction to the agent (not+ * user-facing UI copy), so it is intentionally in English and not localized.+ */+export function buildAutomationDebugPrompt({+ automationName,+ automationPrompt,+ errorDetail,+ stderr,+ runId,+}: AutomationDebugPromptInput): string {+ // Prefer the live stderr the user is looking at; fall back to the run-level+ // error_detail when the sandbox is gone and stderr couldn't be fetched.+ const rawError = stderr?.trim() || errorDetail?.trim() || "";+ const errorSection = rawError+ ? keepTail(rawError, MAX_ERROR_CHARS)+ : "No error output was captured for this run.";++ const trimmedName = automationName?.trim();+ const intro = trimmedName+ ? `The scheduled automation "${trimmedName}" failed during a run.`+ : "A scheduled automation failed during a run.";++ const lines: string[] = [+ `${intro} Please investigate the error and fix the root cause.`,+ ];++ const trimmedPrompt = automationPrompt?.trim();+ if (trimmedPrompt) {+ lines.push("", "What the automation was set up to do:", trimmedPrompt);+ }++ lines.push(+ "",+ `Error output from the failed run (run ${runId}):`,+ errorSection,+ "",+ "Please diagnose why it failed and propose or implement a fix.",+ );++ return lines.join("\n");+}