openclaw/src/auto-reply/reply/agent-runner-execution.test.ts

342 lines
11 KiB
TypeScript

import { afterEach, beforeEach, describe, expect, it, vi } from "vitest";
import { LiveSessionModelSwitchError } from "../../agents/live-model-switch.js";
import type { TemplateContext } from "../templating.js";
import type { GetReplyOptions } from "../types.js";
import type { FollowupRun } from "./queue.js";
import type { TypingSignaler } from "./typing-mode.js";
const state = vi.hoisted(() => ({
runEmbeddedPiAgentMock: vi.fn(),
runWithModelFallbackMock: vi.fn(),
}));
vi.mock("../../agents/pi-embedded.js", () => ({
runEmbeddedPiAgent: (params: unknown) => state.runEmbeddedPiAgentMock(params),
}));
vi.mock("../../agents/model-fallback.js", () => ({
runWithModelFallback: (params: unknown) => state.runWithModelFallbackMock(params),
isFallbackSummaryError: (err: unknown) =>
err instanceof Error &&
err.name === "FallbackSummaryError" &&
Array.isArray((err as { attempts?: unknown[] }).attempts),
}));
vi.mock("../../agents/model-selection.js", () => ({
isCliProvider: () => false,
}));
vi.mock("../../agents/cli-runner.js", () => ({
runCliAgent: vi.fn(),
}));
vi.mock("../../agents/cli-session.js", () => ({
getCliSessionId: vi.fn(),
}));
vi.mock("../../agents/bootstrap-budget.js", () => ({
resolveBootstrapWarningSignaturesSeen: () => [],
}));
vi.mock("../../agents/pi-embedded-helpers.js", () => ({
BILLING_ERROR_USER_MESSAGE: "billing",
isCompactionFailureError: () => false,
isContextOverflowError: () => false,
isBillingErrorMessage: () => false,
isLikelyContextOverflowError: () => false,
isRateLimitErrorMessage: () => false,
isTransientHttpError: () => false,
sanitizeUserFacingText: (text?: string) => text ?? "",
}));
vi.mock("../../config/sessions.js", () => ({
resolveGroupSessionKey: vi.fn(() => null),
resolveSessionTranscriptPath: vi.fn(),
updateSessionStore: vi.fn(),
}));
vi.mock("../../globals.js", () => ({
logVerbose: vi.fn(),
}));
vi.mock("../../infra/agent-events.js", () => ({
emitAgentEvent: vi.fn(),
registerAgentRunContext: vi.fn(),
}));
vi.mock("../../runtime.js", () => ({
defaultRuntime: {
error: vi.fn(),
},
}));
vi.mock("../../utils/message-channel.js", () => ({
isMarkdownCapableMessageChannel: () => true,
resolveMessageChannel: () => "whatsapp",
isInternalMessageChannel: () => false,
}));
vi.mock("../heartbeat.js", () => ({
stripHeartbeatToken: (text: string) => ({
text,
didStrip: false,
shouldSkip: false,
}),
}));
vi.mock("./agent-runner-utils.js", () => ({
buildEmbeddedRunExecutionParams: (params: { provider: string; model: string }) => ({
embeddedContext: {},
senderContext: {},
runBaseParams: {
provider: params.provider,
model: params.model,
},
}),
resolveModelFallbackOptions: vi.fn(() => ({})),
}));
vi.mock("./reply-delivery.js", () => ({
createBlockReplyDeliveryHandler: vi.fn(),
}));
vi.mock("./reply-media-paths.runtime.js", () => ({
createReplyMediaPathNormalizer: () => (payload: unknown) => payload,
}));
async function getRunAgentTurnWithFallback() {
return (await import("./agent-runner-execution.js")).runAgentTurnWithFallback;
}
type FallbackRunnerParams = {
run: (provider: string, model: string) => Promise<unknown>;
};
type EmbeddedAgentParams = {
onToolResult?: (payload: { text?: string; mediaUrls?: string[] }) => Promise<void> | void;
};
function createMockTypingSignaler(): TypingSignaler {
return {
mode: "message",
shouldStartImmediately: false,
shouldStartOnMessageStart: true,
shouldStartOnText: true,
shouldStartOnReasoning: false,
signalRunStart: vi.fn(async () => {}),
signalMessageStart: vi.fn(async () => {}),
signalTextDelta: vi.fn(async () => {}),
signalReasoningDelta: vi.fn(async () => {}),
signalToolStart: vi.fn(async () => {}),
};
}
function createFollowupRun(): FollowupRun {
return {
prompt: "hello",
summaryLine: "hello",
enqueuedAt: Date.now(),
run: {
agentId: "agent",
agentDir: "/tmp/agent",
sessionId: "session",
sessionKey: "main",
messageProvider: "whatsapp",
sessionFile: "/tmp/session.jsonl",
workspaceDir: "/tmp",
config: {},
skillsSnapshot: {},
provider: "anthropic",
model: "claude",
thinkLevel: "low",
verboseLevel: "off",
elevatedLevel: "off",
bashElevated: {
enabled: false,
allowed: false,
defaultLevel: "off",
},
timeoutMs: 1_000,
blockReplyBreak: "message_end",
},
} as unknown as FollowupRun;
}
describe("runAgentTurnWithFallback", () => {
beforeEach(() => {
state.runEmbeddedPiAgentMock.mockReset();
state.runWithModelFallbackMock.mockReset();
state.runWithModelFallbackMock.mockImplementation(async (params: FallbackRunnerParams) => ({
result: await params.run("anthropic", "claude"),
provider: "anthropic",
model: "claude",
attempts: [],
}));
});
afterEach(() => {
vi.clearAllMocks();
});
it("forwards media-only tool results without typing text", async () => {
const onToolResult = vi.fn();
state.runEmbeddedPiAgentMock.mockImplementationOnce(async (params: EmbeddedAgentParams) => {
await params.onToolResult?.({ mediaUrls: ["/tmp/generated.png"] });
return { payloads: [{ text: "final" }], meta: {} };
});
const runAgentTurnWithFallback = await getRunAgentTurnWithFallback();
const pendingToolTasks = new Set<Promise<void>>();
const typingSignals = createMockTypingSignaler();
const result = await runAgentTurnWithFallback({
commandBody: "hello",
followupRun: createFollowupRun(),
sessionCtx: {
Provider: "whatsapp",
MessageSid: "msg",
} as unknown as TemplateContext,
opts: {
onToolResult,
} satisfies GetReplyOptions,
typingSignals,
blockReplyPipeline: null,
blockStreamingEnabled: false,
resolvedBlockStreamingBreak: "message_end",
applyReplyToMode: (payload) => payload,
shouldEmitToolResult: () => true,
shouldEmitToolOutput: () => false,
pendingToolTasks,
resetSessionAfterCompactionFailure: async () => false,
resetSessionAfterRoleOrderingConflict: async () => false,
isHeartbeat: false,
sessionKey: "main",
getActiveSessionEntry: () => undefined,
resolvedVerboseLevel: "off",
});
await Promise.all(pendingToolTasks);
expect(result.kind).toBe("success");
expect(typingSignals.signalTextDelta).not.toHaveBeenCalled();
expect(onToolResult).toHaveBeenCalledTimes(1);
expect(onToolResult.mock.calls[0]?.[0]).toMatchObject({
mediaUrls: ["/tmp/generated.png"],
});
expect(onToolResult.mock.calls[0]?.[0]?.text).toBeUndefined();
});
it("does not show a rate-limit countdown for mixed-cause fallback exhaustion", async () => {
state.runWithModelFallbackMock.mockRejectedValueOnce(
Object.assign(
new Error(
"All models failed (2): anthropic/claude: 429 (rate_limit) | openai/gpt-5.2: 402 (billing)",
),
{
name: "FallbackSummaryError",
attempts: [
{ provider: "anthropic", model: "claude", error: "429", reason: "rate_limit" },
{ provider: "openai", model: "gpt-5.2", error: "402", reason: "billing" },
],
soonestCooldownExpiry: Date.now() + 60_000,
},
),
);
const runAgentTurnWithFallback = await getRunAgentTurnWithFallback();
const result = await runAgentTurnWithFallback({
commandBody: "hello",
followupRun: createFollowupRun(),
sessionCtx: {
Provider: "whatsapp",
MessageSid: "msg",
} as unknown as TemplateContext,
opts: {},
typingSignals: createMockTypingSignaler(),
blockReplyPipeline: null,
blockStreamingEnabled: false,
resolvedBlockStreamingBreak: "message_end",
applyReplyToMode: (payload) => payload,
shouldEmitToolResult: () => true,
shouldEmitToolOutput: () => false,
pendingToolTasks: new Set(),
resetSessionAfterCompactionFailure: async () => false,
resetSessionAfterRoleOrderingConflict: async () => false,
isHeartbeat: false,
sessionKey: "main",
getActiveSessionEntry: () => undefined,
resolvedVerboseLevel: "off",
});
expect(result.kind).toBe("final");
if (result.kind === "final") {
expect(result.payload.text).toContain("Agent failed before reply");
expect(result.payload.text).not.toContain("Rate-limited");
}
});
it("restarts the active prompt when a live model switch is requested", async () => {
let fallbackInvocation = 0;
state.runWithModelFallbackMock.mockImplementation(
async (params: { run: (provider: string, model: string) => Promise<unknown> }) => ({
result: await params.run(
fallbackInvocation === 0 ? "anthropic" : "openai",
fallbackInvocation === 0 ? "claude" : "gpt-5.4",
),
provider: fallbackInvocation === 0 ? "anthropic" : "openai",
model: fallbackInvocation++ === 0 ? "claude" : "gpt-5.4",
attempts: [],
}),
);
state.runEmbeddedPiAgentMock
.mockImplementationOnce(async () => {
throw new LiveSessionModelSwitchError({
provider: "openai",
model: "gpt-5.4",
});
})
.mockImplementationOnce(async () => {
return {
payloads: [{ text: "switched" }],
meta: {
agentMeta: {
sessionId: "session",
provider: "openai",
model: "gpt-5.4",
},
},
};
});
const runAgentTurnWithFallback = await getRunAgentTurnWithFallback();
const followupRun = createFollowupRun();
const result = await runAgentTurnWithFallback({
commandBody: "hello",
followupRun,
sessionCtx: {
Provider: "whatsapp",
MessageSid: "msg",
} as unknown as TemplateContext,
opts: {},
typingSignals: createMockTypingSignaler(),
blockReplyPipeline: null,
blockStreamingEnabled: false,
resolvedBlockStreamingBreak: "message_end",
applyReplyToMode: (payload) => payload,
shouldEmitToolResult: () => true,
shouldEmitToolOutput: () => false,
pendingToolTasks: new Set(),
resetSessionAfterCompactionFailure: async () => false,
resetSessionAfterRoleOrderingConflict: async () => false,
isHeartbeat: false,
sessionKey: "main",
getActiveSessionEntry: () => undefined,
resolvedVerboseLevel: "off",
});
expect(result.kind).toBe("success");
expect(state.runEmbeddedPiAgentMock).toHaveBeenCalledTimes(2);
expect(followupRun.run.provider).toBe("openai");
expect(followupRun.run.model).toBe("gpt-5.4");
});
});