Files
PaperClipAI/server/src/__tests__/browser-use-connection.test.ts
T
DottaandPaperclip 30b753f614 perf(budgets): scope accounting and deliver enforcement asynchronously (#17046)
## Thinking Path

> - Paperclip manages AI agents and their paid work.
> - Budgets must prevent new work and stop reported overruns.
> - Cost settlement also ran serial lifecycle reconciliation for every
eligible company agent.
> - That made request latency grow with unrelated idle agents and
delayed results on stop delivery.
> - This change separates atomic admission and accounting from durable
asynchronous enforcement.
> - Scoped live usage detects overruns without writing provisional
ledger charges.
> - The result is constant query scope for ordinary settlement and
recoverable stop delivery.

## Linked Issues or Issue Description

**What happened?**

Settlement and budget checks entered company-wide lifecycle
reconciliation. A company with 1,000 unrelated idle agents performed
1,001 accounting transactions for one ordinary cost evaluation. User
requests could also wait for provider termination. Running usage
checkpoints did not enforce reported monetary exposure.

**Expected behavior**

Keep the final reservation gate authoritative. Commit charges, scope
blocks, and enforcement intent before returning. Deliver stop requests
after commit. Read only applicable policies and execution scopes for
ordinary charges. Stop long-running work when supported reported or
priced usage establishes an overrun.

**Steps to reproduce**

1. Create a disposable company with one company budget and idle agents.
2. Evaluate a charge below the budget with 0, 17, and 1,000 unrelated
idle agents.
3. Compare query counts, company accounting transactions, and duration.
4. Use the controlled partial-usage and delayed-stop fixtures to check
enforcement and request responsiveness.

**Paperclip version or commit**

The measured baseline is `6c9b87d0922f1ebfc43e8b0c36889ac5ba7b9837` from
`master`.

Related public work: Refs #16976, #2006, #3539, #13763, #14296, #16853,
#13075. These cover the affected service charge, cancellation, and
accounting paths. This change removes their shared reconciliation
overhead.

## What Changed

- Batch applicable policy exposure under the existing company accounting
lock. Preserve exact fractional money and zero-reservation dispatch
fences.
- Commit a durable final budget dispatch claim before provider I/O.
Retain a separate final issue/run ownership transaction through
synchronous handoff, without holding the company accounting lock or
awaiting provider work. Preserve native recovery lease checks and
duplicate handoff fencing. Browser handoff rechecks current policy; it
reuses a consumed rate slot only for the identical policy generation.
- Separate ledger spend, live observed spend, and effective held
capacity. Use the maximum of reservation and live observation for each
unfinished operation.
- Deliver budget cancellation through the existing coordinator.
Revalidate generations and persist queued/running stop fences before
external termination. Bound stop concurrency to four.
- Apply agent holds only to the affected agent. Project budgets leave
other projects running. Project real company transitions in bounded
lifecycle batches.
- Replace periodic broad reconciliation with durable pending work,
startup repair, receipt replay, and concrete retry/window deadlines.
Retain drain and idle-safety checks.
- Enforce accepted cumulative usage checkpoints without provisional
charges. Final accounting atomically replaces live exposure and settles
reservations.
- Add operator-gated budget spans, regression coverage, map updates, and
a real-PostgreSQL comparison harness. No schema or first-party telemetry
change.

## Verification

- Full workspace `pnpm -r typecheck` passes on the final ownership fix,
rebased onto `bbcf5dfe98`. Production `pnpm build` passes on final head
`4c5f4de8b1`. Map, module-boundary, token, and Storybook catalog checks
pass.
- Focused real-PostgreSQL coverage includes the 292-case rebased
accounting/dispatch run, 63 browser cases, 52 run-control/cancellation
cases, 42 policy-generation cases, 77 lifecycle/commentary cases, and
the exact Muse confirmation regression. Native executor unit coverage
passes all 768 cases. Final ownership coverage passes: 74
adapter/use-case cases, nine public regression cases, and the additional
deleted-project race. These exercise competing ownership, budget changes
and charges between phases, zero/nonzero duplicate claims, rollback,
recovery contention, read-only capacity rechecks, and controlled HTTP
startup after commit. Another 74 pure/startup/tracing cases pass. The
complete CI matrix and both 5/5 reviews passed on `1ac9b58`; its
external security scan identified a stale-owner handoff race. The final
fix restores synchronous issue/run ownership fencing after the
accounting transaction commits. Final head `8e84f55ce8` has fresh
Greptile 5/5 with zero new comments, Paperclip review 5/5, and a passing
security scan. The complete latest-head [CI
matrix](https://github.com/paperclipai/paperclip/actions/runs/39535189808)
passes, including budget, browser, Runner, build, and canary jobs. The
unchanged Runner close fixture passed on its sole same-head job retry.
- Broad local `pnpm test:run` invocations were stopped before required
source updates (rebase and review fixes). They are not claimed as full
passes. Final full-suite coverage comes from the complete latest-head CI
matrix; local focused coverage and the incomplete broad invocations are
reported separately.
- Reproduce performance with `pnpm --filter @paperclipai/server exec tsx
../scripts/benchmark-budget-enforcement.ts
--baseline-ref=6c9b87d0922f1ebfc43e8b0c36889ac5ba7b9837`; repeat with
`--reverse`. The harness uses disposable real PostgreSQL, 20 warm
samples per variant/count, and records raw timings, SQL/lock counts,
agent reads, unchanged agent state, and query plans. Final
production-source measurement revision:
`eff713a4f72c7f4b7a4002a701cb481d4bfccd95`; the extended harness is
committed in `4c5f4de8b1`.

| Unrelated idle agents | Baseline median / p95 | New median / p95 |
Baseline SQL / locks | New SQL / locks | New agent reads |
| --- | ---: | ---: | ---: | ---: | ---: |
| 0 | 1.80 / 2.33 ms | 2.99 / 3.88 ms | 5 / 1 | 7 / 1 | 0 |
| 17 | 15.53 / 19.55 ms | 2.24 / 2.60 ms | 90 / 18 | 7 / 1 | 0 |
| 1,000 | 978.35 / 1,790.43 ms | 2.16 / 2.52 ms | 5,005 / 1,001 | 7 / 1
| 0 |

- These measurements use local PostgreSQL 18.1 and Node 26.4.0 arm64.
They isolate the cost-evaluation core and complete admission path,
without hosted-region or provider latency. Prior-revision reversed-order
controls also preserved the constant query counts; the final
measurements above use the final production source.
- Complete reserve → final ownership gate → immediate offline provider
result is measured separately, with one responsible agent and one active
operation at each unrelated-agent count. Admission p95 baseline/new: 0
agents **11.54 / 29.77 ms**; 17 agents **9.31 / 27.62 ms**; 1,000 agents
**24.51 / 74.05 ms**. Baseline admission was already scoped; the
stronger final fence adds bounded overhead. New counts stay constant at
53 Drizzle statements, three transactions, two company-accounting locks,
and five agent-reading statements bound to the responsible agent. Zero
unrelated-agent reads, exactly one provider callback per sample, and no
leftover runs or holds. All admission scenarios stay below 250 ms. No
external provider calls are included.
- The partial-usage/delayed-termination fixture passed 20/20: receipt
persistence p95 13 ms, receipt return to observed durable Stop fence p95
28 ms, complete receipt-start-to-fence p95 50 ms, unrelated database
request p95 1 ms while termination was blocked. After releasing the
controlled termination fixture, acknowledgement p95 was 14 ms. These are
separately measured phases; fixture acknowledgement is not live-provider
termination latency.
- Models and provider deadlines are unchanged. No deployment, live
inference, or staging acceptance is claimed. A quiet 30-minute follow-up
waits for the merged revision on staging, then repeats classifier and
controlled budget-stop acceptance in staging, recording hosted
bookkeeping, queue delay, and actual provider acknowledgement
separately.
- Development failures retained: finalizer notification regressions
(fixed; 27/27 affected cases passed), invalid browser admission fixture
assumptions (fixed; browser coverage now passes 63/63), and first-head
lifecycle/dispatch CI regressions (reproduced and repaired). A
prior-head Cases E2E test hit its unchanged announcement-dismiss handler
race while opening a filter; the same journey passed on `1ac9b58`. The
external security review then found a stale-owner race between
accounting commit and provider entry; final verification covers a
competing run in that gap while retaining provider I/O outside the
accounting lock. Early PostgreSQL startup attempts hit local semaphore
pressure; serial reruns succeeded. The final-head Runner lane had one
failure among 1,823 tests: the unchanged `timeout-attachment` fixture
failed at its first process close after 6.28 seconds. The sole same-head
job retry passed all 1,819 active tests (four skipped), including that
fixture in 1.85 seconds. Three isolated local probes could not run
correctly because the Rust fake-provider fixture binary was absent;
cleanup masked the original startup failure. CI builds that fixture
through its normal lane. These probes are inconclusive and are not
counted as passes.

## Risks

- This changes accounting, lifecycle, and cancellation ordering. Durable
generations, source/sequence checks, ownership leases, and company
locking remain required for correctness.
- Provider stop acknowledgement can be slow or fail. Stop intent remains
pending until existing termination proof succeeds; cancelled row status
alone does not acknowledge delivery.
- Mid-run limits depend on provider usage reports or supported price
calculations. Providers that report only at completion cannot provide
exact mid-call detection.
- Signals are process-local wake hints for the supported
single-active-writer model. Durable state and startup recovery remain
authoritative.
- There are no migrations, new queue tables, or cached balances. Repeat
the classifier and controlled budget-stop acceptance after normal
rollout.

## Model Used

OpenAI Codex, based on GPT-6, with reasoning, tool use, code execution,
and parallel coding agents. The host does not expose the exact backend
model revision or context-window size, so those values are not asserted.

## Checklist

- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [x] All Paperclip CI gates are green
- [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge

---------

Co-authored-by: Paperclip <noreply@paperclip.ing>
2026-10-21 07:23:00 -05:00

1728 lines
79 KiB
TypeScript

import { budgetService, computeObservedSpend } from "../services/budgets.js";
import { buildPaperclipRuntimeMcpServers } from "../services/heartbeat.js";
import { resolveNativeRuntimeMcpSnapshot } from "../services/native-runtime/runtime-context.js";
import express from "express";
import http from "supertest";
import { browserUseRoutes } from "../routes/browser-use.js";
import { errorHandler } from "../middleware/index.js";
import { randomUUID } from "node:crypto";
import { readFile } from "node:fs/promises";
import { and, eq, sql } from "drizzle-orm";
import { afterAll, beforeAll, describe, expect, it, vi } from "vitest";
import {
agents,
approvals,
toolActionRequests,
toolInvocations,
toolPolicies,
companies,
companyMemberships,
connectionGrants,
createDb,
heartbeatRuns,
issues,
projects,
toolConnections,
toolApplications,
toolCatalogEntries,
toolConnectionInstalls,
toolProfiles,
toolProfileBindings,
toolProfileEntries,
browserUseSessions,
browserUseRuns,
browserUseBrowsers,
costEvents,
financeEvents,
budgetPolicies,
budgetReservations,
} from "@paperclipai/db";
import { toolAccessService } from "../services/tool-access.js";
import { createToolGatewayService } from "../services/tool-gateway.js";
import { browserUseViewports } from "../services/browser-use-viewport.js";
import { createDeliveryWorkCoordinator } from "../services/delivery-work-coordinator.js";
import { registerBrowserUseCleanup } from "../services/browser-use-work.js";
import { browserUseService } from "../services/browser-use.js";
import { applyConnectorSkills, prepareConnectorSkillDelivery, resolveConnectorAssignments } from "../services/connector-runtime.js";
import { registerAssignedMcpGateway } from "../services/native-runtime/assigned-mcp-tools.js";
import { listPaperclipSkillEntries } from "@paperclipai/adapter-utils/server-utils";
import { fileURLToPath } from "node:url";
import path from "node:path";
import {
getEmbeddedPostgresTestSupport,
startEmbeddedPostgresTestDatabase,
} from "./helpers/embedded-postgres.js";
const support = await getEmbeddedPostgresTestSupport();
const actor = { actorType: "user" as const, actorId: "browser-reviewer" };
(support.supported ? describe : describe.skip)(
"Browser Use connection and lifecycle",
() => {
let db: ReturnType<typeof createDb>;
let temp: Awaited<ReturnType<typeof startEmbeddedPostgresTestDatabase>>;
const originalApiUrl = process.env.PAPERCLIP_API_URL;
beforeAll(async () => {
process.env.PAPERCLIP_API_URL = "http://127.0.0.1:3100";
temp = await startEmbeddedPostgresTestDatabase("paperclip-browser-use-");
db = createDb(temp.connectionString);
}, 30000);
afterAll(async () => {
await temp?.cleanup();
if (originalApiUrl === undefined) delete process.env.PAPERCLIP_API_URL;
else process.env.PAPERCLIP_API_URL = originalApiUrl;
});
async function fixture(personal = false) {
const [company] = await db
.insert(companies)
.values({
name: "Browser fixture",
issuePrefix: `BU${randomUUID().slice(0, 6)}`,
})
.returning();
await db.insert(companyMemberships).values({
companyId: company.id,
principalType: "user",
principalId: actor.actorId,
membershipRole: "admin",
status: "active",
});
const providerSession = randomUUID(),
providerBrowser = randomUUID();
let providerRun = randomUUID();
let acceptedTask = "", loseReply = false, rejectCode = 0, recoveryPages = 0;
let createGate: (() => Promise<void>) | undefined;
let totalCostUsd = "0.15";
let status = "running",
stopped = false,
browserReady = false,
invalidKey = false,
failCreate = false,
paginatedEvents = false;
const request = vi.fn(async (url: string, init: RequestInit) => {
expect(new Headers(init.headers).get("X-Browser-Use-API-Key")).toBe(
"bu_fixture_secret",
);
if (invalidKey)
return new Response("bu_fixture_secret private-error", {
status: 401,
});
const path = new URL(url).pathname.replace("/api/v4", "");
if (path === "/profiles")
return Response.json({ items: [], totalItems: 0 });
if (path === "/runs" && init.method === "GET") {
const cursor = Number(new URL(url).searchParams.get("cursor") ?? 0);
return Response.json(cursor < recoveryPages
? { runs: [], hasMore: true, nextCursor: String(cursor + 1) }
: { runs: acceptedTask ? [{ id: providerRun, sessionId: providerSession, task: acceptedTask, status }] : [], hasMore: false });
}
if (path === "/runs" && init.method === "POST" && rejectCode)
return new Response("Rejected", { status: rejectCode, headers: { "Retry-After": "1" } });
if (path === "/runs" && init.method === "POST" && failCreate)
throw new Error("lost create response");
if (path === "/runs" && init.method === "POST") {
await createGate?.();
providerRun = randomUUID();
status = "running";
acceptedTask = JSON.parse(String(init.body)).task;
if (loseReply) { browserReady = true; throw new Error("lost accepted create response"); }
return Response.json({
id: providerRun,
sessionId: providerSession,
status,
model: "fixture",
});
}
if (path.endsWith("/cancel")) {
status = "cancelled";
return Response.json({});
}
if (path.endsWith("/status")) return Response.json({ status });
if (path.endsWith("/events") && paginatedEvents) {
const next = Number(new URL(url).searchParams.get("after")) + 1;
return Response.json({
events: [{ id: next, type: "agent.step", data: {} }],
nextAfter: next,
hasMore: next < 9,
});
}
if (path.endsWith("/events"))
return Response.json({
events: browserReady
? [
{
id: 1,
type: "browser.ready",
data: {
live_view_url: "https://live.browser-use.com/canary",
},
},
]
: [],
nextAfter: browserReady ? 1 : 0,
hasMore: false,
});
if (path === `/runs/${providerRun}`)
return Response.json({
status,
result: "Finished https://live.browser-use.com/canary",
totalCostUsd,
model: "fixture",
});
if (path === "/browsers")
return Response.json({
items: browserReady
? [
{
id: providerBrowser,
status: stopped ? "stopped" : "active",
},
]
: [],
totalItems: browserReady ? 1 : 0,
});
if (path === `/browsers/${providerBrowser}`) {
if (init.method === "PATCH") stopped = true;
return Response.json({
id: providerBrowser,
agentSessionId: providerSession,
status: stopped ? "stopped" : "active",
liveUrl: "https://live.browser-use.com/canary",
cdpUrl: `https://${providerBrowser}.cdp.browser-use.com`,
timeoutAt: new Date(Date.now() + 3600000).toISOString(),
});
}
throw new Error(`Unexpected fixture path ${path}`);
});
const access = toolAccessService(db, {
remoteHttpRequest: request,
remoteHttpEndpointLookup: async () => [
{ address: "8.8.8.8", family: 4 },
],
});
const connection = await access.connectGalleryApp(
company.id,
{
galleryKey: "browser-use-cloud",
...(personal ? { grantKind: "user" as const } : {}),
connectionMethodKey: "cloud-v4",
credentialValues: { "credentials.apiKey": "bu_fixture_secret" },
},
actor,
);
const [grant] = await db
.select()
.from(connectionGrants)
.where(eq(connectionGrants.connectionId, connection.connectionId));
const [agent] = await db
.insert(agents)
.values({
companyId: company.id,
name: "Browser agent",
role: "engineer",
adapterType: "process",
adapterConfig: {},
})
.returning();
await access.finishGalleryAppConnection(
company.id,
connection.connectionId,
{
enabledCatalogEntryIds: connection.catalog.map((t) => t.id),
askFirstCatalogEntryIds: [],
access: { agentIds: [agent.id] },
},
actor,
);
const [issue] = await db
.insert(issues)
.values({
companyId: company.id,
title: "Browser task",
assigneeAgentId: agent.id,
})
.returning();
const [run] = await db
.insert(heartbeatRuns)
.values({
companyId: company.id,
agentId: agent.id,
invocationSource: "on_demand",
status: "running",
responsibleUserId: actor.actorId,
contextSnapshot: { issueId: issue.id },
})
.returning();
const [profile] = await db
.insert(toolProfiles)
.values({
companyId: company.id,
name: "Browser tools",
profileKey: randomUUID(),
defaultAction: "allow",
})
.returning();
await db.insert(toolProfileBindings).values({
companyId: company.id,
profileId: profile.id,
targetType: "agent",
targetId: agent.id,
});
const gateway = createToolGatewayService(db, {
remoteHttpRequest: request,
toolActionSigningSecret: "browser-fixture-signing-secret",
});
registerAssignedMcpGateway(db, gateway);
const gatewaySession = await gateway.createSession({
companyId: company.id,
agentId: agent.id,
runId: run.id,
});
const service = browserUseService(db, request);
const binding = {
companyId: company.id,
agentId: agent.id,
issueId: issue.id,
runId: run.id,
};
const tick = async () => {
await db
.update(browserUseSessions)
.set({ nextPollAt: new Date(0) })
.where(eq(browserUseSessions.companyId, company.id));
await service.sweep(connection.connectionId);
};
return {
company,
connection,
access,
grant,
agent,
issue,
run,
service,
binding,
gateway,
gatewaySession,
request,
tick,
ready: () => {
browserReady = true;
},
complete: () => {
status = "completed";
},
invalidate: () => {
invalidKey = true;
},
failCreate: () => {
failCreate = true;
},
loseReply: () => { loseReply = true; },
reject: (code: number) => { rejectCode = code; },
recoveryPages: (count: number) => { recoveryPages = count; },
delayCreate: (gate: () => Promise<void>) => { createGate = gate; },
usage: (value: string) => { totalCostUsd = value; },
paginate: () => {
paginatedEvents = true;
},
};
}
it("wakes the app worker from request-service commits and restores durable poll and lease deadlines", async () => {
const f = await fixture();
const background = browserUseService(db, f.request);
const sweep = vi.spyOn(background, "sweep");
const onError = vi.fn();
const coordinator = createDeliveryWorkCoordinator({ owner: db, canRun: () => true, onError });
try {
await registerBrowserUseCleanup(coordinator, background, () => true).ready;
expect(coordinator.nextWakeAt()).toBeNull();
expect(await background.nextSweepAt()).toBeNull();
const started = await f.service.execute(f.binding, f.grant, randomUUID(), "browser_start", { task: "Read example.com" }) as { sessionId: string };
await vi.waitFor(() => expect(sweep.mock.calls.length).toBeGreaterThan(1));
await coordinator.stop();
expect(onError).not.toHaveBeenCalled();
const poll = new Date(Date.now() + 60_000);
const lease = new Date(Date.now() + 180_000);
await db.update(browserUseSessions).set({ nextPollAt: poll, leaseUntil: lease })
.where(eq(browserUseSessions.id, started.sessionId));
expect(await background.nextSweepAt()).toBe(lease.getTime());
const restarted = createDeliveryWorkCoordinator({ owner: db, canRun: () => true, onError });
try {
await registerBrowserUseCleanup(restarted, browserUseService(db, f.request), () => true).ready;
expect(restarted.nextWakeAt()).toBe(lease.getTime());
} finally { await restarted.stop(); }
await db.update(browserUseSessions).set({ leaseUntil: new Date(0) })
.where(eq(browserUseSessions.id, started.sessionId));
expect(await background.nextSweepAt()).toBe(poll.getTime());
await db.update(browserUseSessions).set({ leaseUntil: null })
.where(eq(browserUseSessions.id, started.sessionId));
expect(await background.nextSweepAt()).toBe(poll.getTime());
const [session] = await db.select().from(browserUseSessions).where(eq(browserUseSessions.id, started.sessionId));
await f.service.control(session, "end");
expect(await background.nextSweepAt()).toBeLessThan(poll.getTime());
await db.update(browserUseSessions).set({ status: "closed" }).where(eq(browserUseSessions.id, started.sessionId));
expect(await background.nextSweepAt()).toBeNull();
} finally {
await coordinator.stop();
await db.update(browserUseSessions).set({ status: "closed" }).where(eq(browserUseSessions.companyId, f.company.id));
}
});
it("keeps Cloud instructions out of universal skills and unassigned runtime overlays", async () => {
const skills = await listPaperclipSkillEntries(fileURLToPath(new URL("../", import.meta.url)), [fileURLToPath(new URL("../../../skills", import.meta.url))]);
expect(skills.some(skill => skill.runtimeName === "paperclip")).toBe(true);
expect(skills.some(skill => ["browser-use", "browser-use-cloud"].includes(skill.runtimeName))).toBe(false);
const base = { paperclipSkillSync: { desiredSkills: ["paperclipai/paperclip/browser-use", "paperclipai/paperclip/browser-use-cloud"] } };
const unassigned = await applyConnectorSkills(base, [
{ key: "paperclipai/paperclip/browser-use", runtimeName: "browser-use", source: "/retired-browser-skill" },
{ key: "paperclipai/paperclip/browser-use-cloud", runtimeName: "browser-use-cloud", source: "/unassigned-cloud-skill" },
{ key: "custom/browser-use", runtimeName: "browser-use", source: "/another-browser-skill" },
], []);
expect(unassigned.paperclipRuntimeSkills).toEqual([{ key: "custom/browser-use", runtimeName: "browser-use", source: "/another-browser-skill" }]);
expect(unassigned.paperclipConnectorSkillDigest).toBeNull();
expect(base.paperclipSkillSync.desiredSkills).toHaveLength(2);
});
it("delivers the Cloud skill only with authorized connection tools in a task run", async () => {
const f = await fixture();
const [unassignedAgent] = await db.insert(agents).values({
companyId: f.company.id,
name: "Agent without the Cloud connection",
role: "engineer",
adapterType: "process",
adapterConfig: {},
}).returning();
expect(await f.gateway.browserUseResources({ companyId: f.company.id, agentId: f.agent.id })).toEqual([]);
expect(await resolveConnectorAssignments(db, { ...f.binding, agentId: unassignedAgent.id })).toEqual([]);
expect(await resolveConnectorAssignments(db, { ...f.binding, companyId: randomUUID() })).toEqual([]);
const assignments = await resolveConnectorAssignments(db, f.binding);
expect(assignments).toMatchObject([{ key: "browser-use-cloud", skillKey: "paperclipai/paperclip/browser-use-cloud", resources: [{ connectionId: f.connection.connectionId }] }]);
const config = await applyConnectorSkills({}, [], assignments);
const skill = config.paperclipRuntimeSkills[0];
expect(skill.runtimeName).toBe("browser-use-cloud");
const markdown = await readFile(path.join(skill.source, "SKILL.md"), "utf8");
expect(markdown).toContain("name: browser-use-cloud");
expect(markdown).toContain("browser_start");
expect(markdown).toContain(f.connection.connectionId);
expect(markdown).not.toContain("bu_fixture_secret");
for (const adapterType of ["paperclip_runner", "codex_local", "claude_local", "kimi_local"]) {
const delivery = await prepareConnectorSkillDelivery({ ...config, engine: "cli" }, adapterType);
expect(delivery.config.paperclipRuntimeSkills).toEqual([skill]);
}
const sharedHome = await prepareConnectorSkillDelivery(config, "cursor_local");
expect(sharedHome.config.paperclipRuntimeSkills).toEqual([]);
expect(sharedHome.instructions).toContain("name: browser-use-cloud");
expect(sharedHome.config.paperclipConnectorSkillDigest).toBe(config.paperclipConnectorSkillDigest);
await db.update(connectionGrants).set({ status: "revoked" }).where(eq(connectionGrants.id, f.grant.id));
expect(await resolveConnectorAssignments(db, f.binding)).toEqual([]);
const revoked = await applyConnectorSkills(config, config.paperclipRuntimeSkills, []);
expect(revoked.paperclipRuntimeSkills).toEqual([]);
expect(revoked.paperclipConnectorSkillDigest).toBeNull();
expect((await prepareConnectorSkillDelivery(revoked, "cursor_local")).instructions).toBe("");
});
it.each([false, true])("skips tool discovery without an active company Browser Use connection (foreign: %s)", async foreign => {
const f = await fixture();
await db.update(toolConnections).set({ enabled: false }).where(eq(toolConnections.id, f.connection.connectionId));
if (foreign) await fixture();
f.request.mockClear();
const discovery = vi.spyOn(db, "transaction");
try {
expect(await f.gateway.browserUseResources(f.binding)).toEqual([]);
expect(discovery).not.toHaveBeenCalled();
expect(f.request).not.toHaveBeenCalled();
} finally { discovery.mockRestore(); }
});
it("resolves only the Browser Use catalog even while full discovery is saturated", async () => {
const f = await fixture();
const [application] = await db.insert(toolApplications).values({
companyId: f.company.id, name: "Unrelated MCP", type: "mcp_http", status: "active",
}).returning();
const [connection] = await db.insert(toolConnections).values({
companyId: f.company.id, applicationId: application.id, name: "Unrelated MCP",
uid: randomUUID(), transport: "mcp_remote", status: "active", enabled: true, healthStatus: "ok",
}).returning();
await db.insert(toolCatalogEntries).values(Array.from({ length: 500 }, (_, i) => ({
companyId: f.company.id, applicationId: application.id, connectionId: connection.id,
entryKind: "tool" as const, name: `unrelated_${i}`, toolName: `unrelated_${i}`,
inputSchema: { type: "object" }, riskLevel: "read" as const, status: "active" as const,
versionHash: "fixture", schemaHash: "fixture",
})));
const { recordingDb } = await import("./helpers/tool-gateway-listing-fixture.js");
const { toolDiscoveryScheduler } = await import("../services/tool-discovery-scheduler.js");
const releases: Array<() => void> = [];
const blockers = [0, 1].map(() => toolDiscoveryScheduler.run(() => new Promise<void>(resolve => releases.push(resolve))));
await vi.waitFor(() => expect(releases).toHaveLength(2));
const recorded = recordingDb(db);
const gateway = createToolGatewayService(recorded.db);
const pending = gateway.browserUseResources(f.binding);
let timeout: ReturnType<typeof setTimeout> | undefined;
try {
const resources = await Promise.race([pending, new Promise<never>((_, reject) => {
timeout = setTimeout(() => reject(new Error("Assignment waited for unrelated discovery")), 5_000);
})]);
expect(resources).toEqual([{ id: f.connection.connectionId, connectionId: f.connection.connectionId, label: "Browser Use Cloud" }]);
const catalogReads = recorded.statements.flatMap((query, index) =>
query.includes('from "tool_catalog_entries"') && query.includes('inner join "tool_connections"') ? [recorded.statementParams[index]] : []);
expect(catalogReads).toHaveLength(1);
expect(catalogReads[0]).toContain(f.connection.connectionId);
expect(catalogReads[0]).not.toContain(connection.id);
expect(recorded.statements.some(query => query.includes('from "chat_endpoints"'))).toBe(false);
} finally {
clearTimeout(timeout);
releases.forEach(release => release());
await Promise.all(blockers);
await pending;
}
});
it.each(["policy", "catalog", "grant"])("rechecks Browser Use %s revocation without a cached assignment", async kind => {
const f = await fixture();
expect(await f.gateway.browserUseResources(f.binding)).toHaveLength(1);
if (kind === "policy") await db.insert(toolPolicies).values({
companyId: f.company.id, name: "Revoke Browser Use", policyType: "block",
selectors: { connectionId: f.connection.connectionId }, priority: 100,
});
if (kind === "catalog") await db.update(toolCatalogEntries).set({ quarantinedAt: new Date() }).where(eq(toolCatalogEntries.connectionId, f.connection.connectionId));
if (kind === "grant") await db.update(connectionGrants).set({ status: "revoked" }).where(eq(connectionGrants.id, f.grant.id));
expect(await f.gateway.browserUseResources(f.binding)).toEqual([]);
});
it("withholds Cloud instructions when connection tools are disabled", async () => {
const f = await fixture();
await db.update(toolConnections).set({ enabled: false }).where(eq(toolConnections.id, f.connection.connectionId));
expect(await resolveConnectorAssignments(db, f.binding)).toEqual([]);
});
it("reapplies the renumbered migration without losing existing browser work", async () => {
const f = await fixture();
await f.service.execute(f.binding, f.grant, randomUUID(), "browser_start", {
task: "Read example.com",
maxCostUsd: 0.5,
});
f.ready(); f.complete(); await f.tick();
const previous = await db.select().from(browserUseRuns)
.where(eq(browserUseRuns.companyId, f.company.id));
expect(previous).toHaveLength(1);
const [original] = await db.select().from(toolConnections).where(eq(toolConnections.id, f.connection.connectionId));
await db.update(toolConnections).set({
config: { ...original.config, sourceTemplateKey: "browser-use" },
transportConfig: { ...original.transportConfig, sourceTemplateKey: "browser-use" },
}).where(eq(toolConnections.id, original.id));
await db.update(toolApplications).set({
applicationKey: `app-gallery:browser-use:${original.id}`,
metadata: { sourceTemplateKey: "browser-use", galleryKey: "browser-use" },
}).where(eq(toolApplications.id, original.applicationId));
const legacyCost = { provider: "browser-use", biller: "browser-use", billingCode: `browser-use:${previous[0].id}` };
await db.update(costEvents).set(legacyCost).where(eq(costEvents.companyId, f.company.id));
await db.update(financeEvents).set(legacyCost).where(eq(financeEvents.companyId, f.company.id));
const migration = await readFile(new URL(
"../../../packages/db/src/migrations/0290_browser_use_cloud.sql", import.meta.url,
), "utf8");
for (let attempt = 0; attempt < 2; attempt++) {
for (const statement of migration.split("--> statement-breakpoint")) {
if (statement.trim()) await db.execute(sql.raw(statement));
}
}
expect(await db.select().from(browserUseRuns)
.where(eq(browserUseRuns.companyId, f.company.id))).toEqual(previous);
const [migrated] = await db.select().from(toolConnections).where(eq(toolConnections.id, original.id));
expect(migrated).toMatchObject({ config: { sourceTemplateKey: "browser-use-cloud" }, transportConfig: { sourceTemplateKey: "browser-use-cloud" }, credentialSecretRefs: original.credentialSecretRefs });
expect(await db.select().from(toolApplications).where(eq(toolApplications.id, original.applicationId))).toMatchObject([{ applicationKey: `app-gallery:browser-use-cloud:${original.id}`, metadata: { sourceTemplateKey: "browser-use-cloud", galleryKey: "browser-use-cloud" } }]);
expect(await resolveConnectorAssignments(db, f.binding)).toMatchObject([{ key: "browser-use-cloud" }]);
await f.tick();
expect(await db.select().from(costEvents).where(eq(costEvents.companyId, f.company.id))).toMatchObject([{ provider: "browser-use-cloud", biller: "browser-use-cloud", billingCode: `browser-use-cloud:${previous[0].id}`, costCents: 15 }]);
expect(await db.select().from(financeEvents).where(eq(financeEvents.companyId, f.company.id))).toMatchObject([{ provider: "browser-use-cloud", amountCents: 15 }]);
});
it("delivers Browser Use through the pinned native and CLI runtime gateway", async () => {
const f = await fixture();
await db.insert(toolConnectionInstalls).values({
companyId: f.company.id,
connectionId: f.connection.connectionId,
targetType: "agent",
targetId: f.agent.id,
});
const snapshot = await resolveNativeRuntimeMcpSnapshot({
db,
agent: f.agent,
runId: f.run.id,
});
expect(snapshot.bindingId).not.toBeNull();
const servers = await buildPaperclipRuntimeMcpServers({
db,
agent: f.agent,
runId: f.run.id,
expectedAssignmentDigest: snapshot.digest,
});
expect(servers).toHaveLength(1);
const gatewayPublicId = new URL(servers[0]!.url).pathname
.split("/")
.at(-1)!;
const tools = await f.gateway.listToolsForNamedGateway({
gatewayPublicId,
bearerToken: servers[0]!.token!,
});
expect(tools).toHaveLength(7);
expect(tools.some((tool) => tool.name.endsWith(":browser-start"))).toBe(
true,
);
});
it("verifies credentials without paid work, exposes REST tools and rejects invalid keys", async () => {
const f = await fixture();
expect(f.connection.catalog).toHaveLength(7);
expect(
f.request.mock.calls.every(([url]) => url.includes("/profiles")),
).toBe(true);
expect(JSON.stringify(f.connection)).not.toContain("bu_fixture_secret");
const tools = await f.gateway.listToolsForSession(f.gatewaySession.token);
expect(
tools.find((t) => t.upstreamToolName === "browser_start"),
).toMatchObject({ providerType: "provider_rest", risk: "destructive" });
const start = tools.find((t) => t.upstreamToolName === "browser_start")!;
const idempotencyKey = randomUUID();
const result = await f.gateway.executeTool({
sessionToken: f.gatewaySession.token,
tool: start.name,
parameters: { task: "Read example.com" },
idempotencyKey,
});
expect(result.status).toBe("completed");
expect(JSON.stringify(result)).not.toContain("bu_fixture_secret");
await f.gateway.executeTool({
sessionToken: f.gatewaySession.token,
tool: start.name,
parameters: { task: "Read example.com" },
idempotencyKey,
});
expect(
f.request.mock.calls.filter(
([url, init]) => url.endsWith("/runs") && init.method === "POST",
),
).toHaveLength(1);
f.invalidate();
await expect(
f.access.refreshCatalog(f.connection.connectionId, actor),
).rejects.toThrow();
});
it("observes delayed browser.ready, hides viewer credentials, accounts once and cleans up", async () => {
const f = await fixture();
const invocation = randomUUID();
const started = (await f.service.execute(
f.binding,
f.grant,
invocation,
"browser_start",
{ task: "Read example.com", maxCostUsd: 0.5 },
)) as { sessionId: string };
expect(started.sessionId).toBeTruthy();
const createCall = f.request.mock.calls.find(
([url, init]) => url.endsWith("/runs") && init.method === "POST",
);
expect(
JSON.parse(String(createCall?.[1].body)).browserSettings.allowResizing,
).toBe(true);
await f.tick();
expect(
await db
.select()
.from(browserUseBrowsers)
.where(eq(browserUseBrowsers.companyId, f.company.id)),
).toHaveLength(0);
f.ready();
await f.tick();
const list = await f.service.list(
f.company.id,
f.issue.id,
actor.actorId,
);
expect(list).toHaveLength(1);
expect(JSON.stringify(list)).not.toContain("canary");
expect(
await f.service.viewer(
f.company.id,
f.issue.id,
list[0].id,
actor.actorId,
),
).toEqual({
url: "https://live.browser-use.com/canary",
viewport: "fit",
viewportState: { preset: "fit" },
});
await expect(
f.service.viewer(f.company.id, randomUUID(), list[0].id, actor.actorId),
).rejects.toThrow("not found");
f.complete();
await f.tick();
await f.tick();
const [s] = await db
.select()
.from(browserUseSessions)
.where(eq(browserUseSessions.id, started.sessionId));
expect(s.status).toBe("idle");
expect(s.idleDeadline).toBeTruthy();
expect(
await db
.select()
.from(costEvents)
.where(eq(costEvents.companyId, f.company.id)),
).toHaveLength(1);
const ledger = await db
.select()
.from(financeEvents)
.where(eq(financeEvents.companyId, f.company.id));
const [cost] = await db
.select()
.from(costEvents)
.where(eq(costEvents.companyId, f.company.id));
expect(ledger).toHaveLength(1);
expect(ledger[0]).toMatchObject({
issueId: f.issue.id,
agentId: f.agent.id,
costEventId: cost.id,
amountCents: 15,
biller: "browser-use-cloud",
direction: "debit",
currency: "USD",
});
expect(
JSON.stringify(
await f.service.execute(
f.binding,
f.grant,
randomUUID(),
"browser_status",
{ sessionId: s.id },
),
),
).not.toContain("canary");
await f.service.control(s, "end");
await f.tick();
expect(
(
await db
.select()
.from(browserUseSessions)
.where(eq(browserUseSessions.id, s.id))
)[0].status,
).toBe("closed");
expect(
(
await db
.select()
.from(browserUseBrowsers)
.where(eq(browserUseBrowsers.sessionId, s.id))
)[0].status,
).toBe("stopped");
});
it("keeps a viewed idle browser alive but still honors expiry, stop and authorization", async () => {
const f = await fixture();
const started = (await f.service.execute(
f.binding,
f.grant,
randomUUID(),
"browser_start",
{ task: "Read" },
)) as { sessionId: string };
f.ready();
f.complete();
await f.tick();
const [browser] = await f.service.list(
f.company.id,
f.issue.id,
actor.actorId,
);
const read = async () =>
(
await db
.select()
.from(browserUseSessions)
.where(eq(browserUseSessions.id, started.sessionId))
)[0];
const original = new Date(Date.now() + 1_000);
await db
.update(browserUseSessions)
.set({ idleDeadline: original })
.where(eq(browserUseSessions.id, started.sessionId));
await expect(
f.service.presence(
f.company.id,
randomUUID(),
browser.id,
actor.actorId,
),
).rejects.toThrow();
expect((await read()).idleDeadline?.getTime()).toBe(original.getTime());
expect(
await f.service.presence(
f.company.id,
f.issue.id,
browser.id,
actor.actorId,
),
).toEqual({ accepted: true });
expect((await read()).idleDeadline!.getTime()).toBeGreaterThan(
Date.now() + 9 * 60_000,
);
await f.tick();
expect((await read()).status).toBe("idle");
// Provider lifetime wins over viewer presence.
const expiry = new Date(Date.now() + 3 * 60_000);
await db
.update(browserUseBrowsers)
.set({ expiresAt: expiry })
.where(eq(browserUseBrowsers.id, browser.id));
await db
.update(browserUseSessions)
.set({ idleDeadline: original })
.where(eq(browserUseSessions.id, started.sessionId));
await f.service.presence(
f.company.id,
f.issue.id,
browser.id,
actor.actorId,
);
expect((await read()).idleDeadline?.getTime()).toBe(expiry.getTime());
await f.service.control(await read(), "end");
await expect(
f.service.presence(f.company.id, f.issue.id, browser.id, actor.actorId),
).rejects.toThrow();
await f.tick();
expect((await read()).status).toBe("closed");
});
it("executes an approved start once after the original turn ends and attaches the resumed run", async () => {
const f = await fixture();
await db.insert(toolPolicies).values({
companyId: f.company.id,
name: "Review browser writes",
policyType: "require_approval",
selectors: { connectionId: f.connection.connectionId },
priority: 10,
});
const start = (
await f.gateway.listToolsForSession(f.gatewaySession.token)
).find((t) => t.upstreamToolName === "browser_start")!;
await expect(
f.gateway.executeTool({
sessionToken: f.gatewaySession.token,
tool: start.name,
parameters: { task: "Read example.com" },
idempotencyKey: randomUUID(),
}),
).rejects.toMatchObject({ reasonCode: "approval_required" });
const [pending] = await db
.select()
.from(toolActionRequests)
.where(eq(toolActionRequests.companyId, f.company.id));
await db
.update(heartbeatRuns)
.set({ status: "succeeded" })
.where(eq(heartbeatRuns.id, f.run.id));
if (pending.approvalId)
await db
.update(approvals)
.set({
status: "approved",
decidedByUserId: actor.actorId,
decidedAt: new Date(),
})
.where(eq(approvals.id, pending.approvalId));
await f.gateway.approveActionRequest({
companyId: f.company.id,
actionRequestId: pending.id,
actor: { userId: actor.actorId },
});
const [record] = await db
.select()
.from(browserUseRuns)
.where(eq(browserUseRuns.companyId, f.company.id));
expect(record?.providerRunId).toBeTruthy();
expect(record.detachedUntil).toBeTruthy();
f.ready();
await f.tick();
expect(
(
await db
.select()
.from(browserUseSessions)
.where(eq(browserUseSessions.id, record.sessionId))
)[0].status,
).toBe("running");
const [resumed] = await db
.insert(heartbeatRuns)
.values({
companyId: f.company.id,
agentId: f.agent.id,
invocationSource: "on_demand",
status: "running",
responsibleUserId: actor.actorId,
contextSnapshot: { issueId: f.issue.id },
})
.returning();
await f.service.execute(
{ ...f.binding, runId: resumed.id },
f.grant,
randomUUID(),
"browser_status",
{ sessionId: record.sessionId },
);
const [attached] = await db
.select()
.from(browserUseRuns)
.where(eq(browserUseRuns.id, record.id));
expect(attached.detachedUntil).toBeNull();
expect(attached.heartbeatRunId).toBe(resumed.id);
await f.gateway.approveActionRequest({
companyId: f.company.id,
actionRequestId: pending.id,
actor: { userId: actor.actorId },
});
expect(
f.request.mock.calls.filter(
([url, init]) => url.endsWith("/runs") && init.method === "POST",
),
).toHaveLength(1);
});
it("enforces task, agent, grant and profile ownership and never replays a create", async () => {
const f = await fixture();
const invocation = randomUUID();
await expect(
f.service.execute(f.binding, f.grant, invocation, "browser_start", {
task: "Read",
profileId: randomUUID(),
}),
).rejects.toThrow("profile");
const start = (await f.service.execute(
f.binding,
f.grant,
invocation,
"browser_start",
{ task: "Read" },
)) as { sessionId: string };
await f.service.execute(f.binding, f.grant, invocation, "browser_start", {
task: "Read",
});
expect(
f.request.mock.calls.filter(
([url, init]) => url.endsWith("/runs") && init.method === "POST",
),
).toHaveLength(1);
await expect(
f.service.execute(
{ ...f.binding, issueId: randomUUID() },
f.grant,
randomUUID(),
"browser_status",
{ sessionId: start.sessionId },
),
).rejects.toThrow();
await expect(
f.service.execute(
f.binding,
{ ...f.grant, id: randomUUID() },
randomUUID(),
"browser_status",
{ sessionId: start.sessionId },
),
).rejects.toThrow();
await expect(
f.service.execute(
f.binding,
f.grant,
randomUUID(),
"browser_continue",
{ sessionId: start.sessionId, task: "Again" },
),
).rejects.toThrow("busy");
f.ready();
await f.tick();
await db
.update(connectionGrants)
.set({ status: "revoked" })
.where(eq(connectionGrants.id, f.grant.id));
expect(
await f.service.list(f.company.id, f.issue.id, actor.actorId),
).toEqual([]);
await f.tick();
expect(
(
await db
.select()
.from(browserUseSessions)
.where(eq(browserUseSessions.id, start.sessionId))
)[0].status,
).toBe("closed");
});
it("does not replay an uncertain paid create", async () => {
const f = await fixture();
f.failCreate();
const invocationId = randomUUID();
await expect(
f.service.execute(f.binding, f.grant, invocationId, "browser_start", {
task: "Read",
}),
).rejects.toThrow("could not be confirmed");
const result = await f.service.execute(
f.binding,
f.grant,
invocationId,
"browser_start",
{ task: "Read" },
);
expect(result).toMatchObject({ status: "stopping", runStatus: "unknown" });
expect(
f.request.mock.calls.filter(
([url, init]) => url.endsWith("/runs") && init.method === "POST",
),
).toHaveLength(1);
});
it.each([409, 429])("keeps a browser visible and closable after a rejected continuation (%s)", async (code) => {
const f = await fixture();
const start = await f.service.execute(f.binding, f.grant, randomUUID(), "browser_start", { task: "Read" }) as { sessionId: string };
f.ready(); f.complete(); await f.tick();
f.reject(code);
await expect(f.service.execute(f.binding, f.grant, randomUUID(), "browser_continue", { task: "Read again", sessionId: start.sessionId })).rejects.toThrow();
expect(await f.service.list(f.company.id, f.issue.id, actor.actorId)).toMatchObject([{ status: "idle" }]);
const [session] = await db.select().from(browserUseSessions).where(eq(browserUseSessions.id, start.sessionId));
await f.service.control(session, "end", actor.actorId);
// The 429 response delays the whole credential's provider requests.
if (code === 429) await new Promise(resolve => setTimeout(resolve, 1050));
await f.tick();
expect(await f.service.list(f.company.id, f.issue.id, actor.actorId)).toMatchObject([{ status: "closed" }]);
});
it("rolls back an interrupted rejection and closes the existing browser after restart", async () => {
const f = await fixture();
const start = await f.service.execute(f.binding, f.grant, randomUUID(), "browser_start", { task: "Read" }) as { sessionId: string };
f.ready(); f.complete(); await f.tick();
f.reject(409);
const invocationId = randomUUID();
const transaction = db.transaction.bind(db);
const interrupted = vi.spyOn(db, "transaction").mockImplementation((callback, config) =>
transaction(async tx => {
const result = await callback(tx);
const [run] = await tx.select().from(browserUseRuns).where(eq(browserUseRuns.invocationId, invocationId));
if (run?.status === "failed") throw new Error("process interrupted before commit");
return result;
}, config));
try {
await expect(f.service.execute(f.binding, f.grant, invocationId, "browser_continue", { task: "Read again", sessionId: start.sessionId })).rejects.toThrow("process interrupted");
} finally {
interrupted.mockRestore();
}
const [pending] = await db.select().from(browserUseRuns).where(eq(browserUseRuns.invocationId, invocationId));
expect(pending).toMatchObject({ status: "creating", eventsDrained: 0, providerRunId: null });
await db.update(browserUseRuns).set({ createdAt: new Date(Date.now() - 61_000) }).where(eq(browserUseRuns.id, pending.id));
await db.update(browserUseSessions).set({ nextPollAt: new Date(0) }).where(eq(browserUseSessions.id, start.sessionId));
await browserUseService(db, f.request).sweep();
expect(await db.select().from(browserUseBrowsers).where(eq(browserUseBrowsers.sessionId, start.sessionId))).toMatchObject([{ status: "stopped" }]);
expect(f.request.mock.calls.filter(([url, init]) => url.endsWith("/runs") && init.method === "POST")).toHaveLength(2);
});
it("recovers an accepted start after a lost response and restart, then stops and accounts without replay", async () => {
const f = await fixture();
f.loseReply(); f.recoveryPages(6);
await expect(f.service.execute(f.binding, f.grant, randomUUID(), "browser_start", { task: "Read" })).rejects.toThrow("could not be confirmed");
await f.tick();
const [pending] = await db.select().from(browserUseRuns).where(eq(browserUseRuns.companyId, f.company.id));
expect(pending.recoveryCursor).toBe("5");
await db.update(browserUseSessions).set({ nextPollAt: new Date(0) }).where(eq(browserUseSessions.companyId, f.company.id));
await browserUseService(db, f.request).sweep();
expect(await f.service.list(f.company.id, f.issue.id, actor.actorId)).toMatchObject([{ status: "closed" }]);
expect(await db.select().from(financeEvents).where(eq(financeEvents.companyId, f.company.id))).toMatchObject([{ amountCents: 15 }]);
expect(f.request.mock.calls.filter(([url, init]) => url.endsWith("/runs") && init.method === "POST")).toHaveLength(1);
});
it("recovers a continuation's lost reply within its existing provider session", async () => {
const f = await fixture();
const start = await f.service.execute(f.binding, f.grant, randomUUID(), "browser_start", { task: "Read" }) as { sessionId: string };
f.ready(); f.complete(); await f.tick();
f.loseReply();
await expect(f.service.execute(f.binding, f.grant, randomUUID(), "browser_continue", { task: "Read again", sessionId: start.sessionId })).rejects.toThrow("could not be confirmed");
await f.tick();
expect(await f.service.list(f.company.id, f.issue.id, actor.actorId)).toMatchObject([{ status: "closed" }]);
expect(f.request.mock.calls.some(([url]) => new URL(url).searchParams.has("sessionId"))).toBe(true);
expect(f.request.mock.calls.filter(([url, init]) => url.endsWith("/runs") && init.method === "POST")).toHaveLength(2);
});
it("recovers a process crash before provider identifiers were saved", async () => {
const f = await fixture();
const start = await f.service.execute(f.binding, f.grant, randomUUID(), "browser_start", { task: "Read" }) as { sessionId: string };
f.ready();
await db.update(browserUseRuns).set({ providerRunId: null, status: "creating", createdAt: new Date(Date.now() - 61_000) }).where(eq(browserUseRuns.sessionId, start.sessionId));
await db.update(browserUseSessions).set({ providerSessionId: null, status: "starting" }).where(eq(browserUseSessions.id, start.sessionId));
await f.tick();
expect(await f.service.list(f.company.id, f.issue.id, actor.actorId)).toMatchObject([{ status: "closed" }]);
expect(f.request.mock.calls.filter(([url, init]) => url.endsWith("/runs") && init.method === "POST")).toHaveLength(1);
});
it("drains bounded event pages after restart before closing and accounting", async () => {
const f = await fixture();
const start = (await f.service.execute(
f.binding,
f.grant,
randomUUID(),
"browser_start",
{ task: "Read" },
)) as { sessionId: string };
f.ready();
await f.tick();
f.paginate();
f.complete();
const [session] = await db
.select()
.from(browserUseSessions)
.where(eq(browserUseSessions.id, start.sessionId));
await f.service.control(session, "end");
await f.tick();
expect(
(
await db
.select()
.from(browserUseSessions)
.where(eq(browserUseSessions.id, start.sessionId))
)[0].status,
).toBe("stopping");
expect(
await db
.select()
.from(costEvents)
.where(eq(costEvents.companyId, f.company.id)),
).toHaveLength(0);
await db
.update(browserUseSessions)
.set({ nextPollAt: new Date(0) })
.where(eq(browserUseSessions.id, start.sessionId));
await browserUseService(db, f.request).sweep();
expect(
(
await db
.select()
.from(browserUseSessions)
.where(eq(browserUseSessions.id, start.sessionId))
)[0].status,
).toBe("closed");
expect(
(
await db
.select()
.from(browserUseRuns)
.where(eq(browserUseRuns.sessionId, start.sessionId))
)[0],
).toMatchObject({ eventCursor: 9, eventsDrained: 1, accountedCents: 15 });
});
it.each(["task", "saved chat"])("protects %s browser routes from agent, other-company and read-only access", async (kind) => {
const f = await fixture();
if (kind === "saved chat") {
await db.update(issues).set({
conversationAgentId: f.agent.id,
conversationUserId: actor.actorId,
conversationState: "active",
}).where(eq(issues.id, f.issue.id));
}
await f.service.execute(
f.binding,
f.grant,
randomUUID(),
"browser_start",
{ task: "Read" },
);
f.ready();
await f.tick();
const [browser] = await f.service.list(
f.company.id,
f.issue.id,
actor.actorId,
);
let mode = "board";
const app = express();
app.use(express.json());
app.use((req, _res, next) => {
req.actor =
mode === "agent"
? {
type: "agent",
companyId: f.company.id,
agentId: f.agent.id,
source: "agent_key",
}
: {
type: "board",
userId: actor.actorId,
source: "session",
companyIds: mode === "other-company" ? [] : [f.company.id],
};
next();
});
app.use(browserUseRoutes(db, f.service));
app.use(errorHandler);
const listPath = `/issues/${f.issue.id}/browsers`;
const listed = await http(app).get(listPath);
expect(listed.status).toBe(200);
expect(listed.body).toMatchObject([{ id: browser.id, issueId: f.issue.id }]);
const missing = await http(app).get(`/issues/${randomUUID()}/browsers`);
expect(missing.status).toBe(404);
expect(missing.body).toEqual({ error: "Task not found" });
const path = `/issues/${f.issue.id}/browsers/${browser.id}/viewer`;
const ok = await http(app).get(path);
expect(ok.status).toBe(200);
expect(ok.headers["cache-control"]).toBe("no-store");
expect(ok.headers["referrer-policy"]).toBe("no-referrer");
const controlPath = path.replace("/viewer", "/control");
const presencePath = path.replace("/viewer", "/presence");
expect((await http(app).post(presencePath)).status).toBe(200);
const resizePath = path.replace("/viewer", "/viewport");
const resizeSpy = vi
.spyOn(browserUseViewports, "resize")
.mockResolvedValue({ preset: "phone" });
expect(
(await http(app).post(resizePath).send({ preset: "phone" })).body,
).toEqual({ preset: "phone" });
expect(resizeSpy).toHaveBeenCalledTimes(1);
const viewerId = randomUUID();
resizeSpy.mockResolvedValueOnce({
preset: "fit",
width: 531,
height: 712,
applied: true,
});
expect(
(
await http(app)
.post(resizePath)
.send({ preset: "fit", width: 531, height: 712, viewerId })
).body,
).toMatchObject({ preset: "fit", width: 531 });
expect(resizeSpy).toHaveBeenLastCalledWith(
expect.any(String),
expect.any(String),
{ preset: "fit", width: 531, height: 712, viewerId },
expect.any(String),
`${actor.actorId}:${viewerId}`,
);
expect(
(await http(app).post(`${resizePath}/release`).send({ viewerId }))
.status,
).toBe(200);
expect(
(await http(app).post(resizePath).send({ preset: "custom", width: 1 }))
.status,
).toBe(400);
expect((await http(app).post(controlPath).send({ action: "cancel" })).status).toBe(200);
mode = "agent";
expect((await http(app).get(listPath)).status).toBe(403);
expect((await http(app).post(controlPath).send({ action: "keep_open" })).status).toBe(403);
expect((await http(app).post(presencePath)).status).toBe(403);
expect(
(await http(app).post(`${resizePath}/release`).send({ viewerId }))
.status,
).toBe(403);
expect(
(await http(app).post(resizePath).send({ preset: "phone" })).status,
).toBe(403);
expect((await http(app).get(path)).status).toBe(403);
mode = "other-company";
const inaccessible = await http(app).get(listPath);
expect(inaccessible.status).toBe(missing.status);
expect(inaccessible.body).toEqual(missing.body);
expect((await http(app).post(controlPath).send({ action: "keep_open" })).status).toBe(404);
expect((await http(app).post(presencePath)).status).toBe(404);
expect(
(await http(app).post(`${resizePath}/release`).send({ viewerId }))
.status,
).toBe(404);
expect(
(await http(app).post(resizePath).send({ preset: "phone" })).status,
).toBe(404);
expect((await http(app).get(path)).status).toBe(404);
mode = "board";
await db
.update(companyMemberships)
.set({ membershipRole: "viewer" })
.where(eq(companyMemberships.companyId, f.company.id));
expect((await http(app).get(path)).status).toBe(403);
expect((await http(app).get(listPath)).body).toEqual([]);
expect((await http(app).post(controlPath).send({ action: "keep_open" })).status).toBe(403);
expect((await http(app).post(presencePath)).status).toBe(403);
expect(
(await http(app).post(resizePath).send({ preset: "phone" })).status,
).toBe(403);
expect(resizeSpy).toHaveBeenCalledTimes(2);
resizeSpy.mockRestore();
});
it("rechecks current gateway access and closes browsers when it is removed", async () => {
const f = await fixture();
const start = (
await f.gateway.listToolsForSession(f.gatewaySession.token)
).find((t) => t.upstreamToolName === "browser_start")!;
await f.gateway.executeTool({
sessionToken: f.gatewaySession.token,
tool: start.name,
parameters: { task: "Read" },
idempotencyKey: randomUUID(),
});
f.ready();
await f.tick();
const [run] = await db
.select()
.from(browserUseRuns)
.where(eq(browserUseRuns.companyId, f.company.id));
const scope = {
...f.binding,
runId: run.heartbeatRunId,
connectionId: f.grant.connectionId,
grantId: f.grant.id,
invocationId: run.invocationId,
};
expect(await f.gateway.browserUseSessionAuthorized(scope)).toBe(true);
await db.insert(toolPolicies).values({
companyId: f.company.id,
name: "Disable browser access",
policyType: "block",
selectors: { connectionId: f.connection.connectionId },
priority: 100,
});
expect(await f.gateway.browserUseSessionAuthorized(scope)).toBe(false);
await db
.update(browserUseSessions)
.set({ nextPollAt: new Date(0) })
.where(eq(browserUseSessions.companyId, f.company.id));
await browserUseService(db, f.request, {}, () =>
f.gateway.browserUseSessionAuthorized(scope),
).sweep();
expect(
(
await db
.select()
.from(browserUseSessions)
.where(eq(browserUseSessions.id, run.sessionId))
)[0].status,
).toBe("closed");
});
it("resolves only the selected personal credential and denies another member's viewer", async () => {
const f = await fixture(true);
expect(f.grant.kind).toBe("user");
const tool = (
await f.gateway.listToolsForSession(f.gatewaySession.token)
).find((t) => t.upstreamToolName === "browser_start")!;
await f.gateway.executeTool({
sessionToken: f.gatewaySession.token,
tool: tool.name,
parameters: { task: "Read" },
idempotencyKey: randomUUID(),
});
f.ready();
await f.tick();
const [browser] = await f.service.list(
f.company.id,
f.issue.id,
actor.actorId,
);
expect(
await f.service.viewer(
f.company.id,
f.issue.id,
browser.id,
actor.actorId,
),
).toHaveProperty("url");
await db.insert(companyMemberships).values({
companyId: f.company.id,
principalType: "user",
principalId: "other-reviewer",
membershipRole: "admin",
status: "active",
});
await expect(
f.service.viewer(
f.company.id,
f.issue.id,
browser.id,
"other-reviewer",
),
).rejects.toThrow("another credential");
await f.access.archiveConnection(
f.connection.connectionId,
f.company.id,
actor,
);
expect(
(
await db
.select()
.from(browserUseSessions)
.where(eq(browserUseSessions.companyId, f.company.id))
)[0].status,
).toBe("closed");
});
it("continues an idle conversation and bounds Keep open with idle cleanup", async () => {
const f = await fixture();
const start = (await f.service.execute(
f.binding,
f.grant,
randomUUID(),
"browser_start",
{ task: "Read" },
)) as { sessionId: string };
f.ready();
f.complete();
await f.tick();
const [idle] = await db
.select()
.from(browserUseSessions)
.where(eq(browserUseSessions.id, start.sessionId));
await f.service.control(idle, "keep_open");
await f.service.execute(
f.binding,
f.grant,
randomUUID(),
"browser_continue",
{ sessionId: start.sessionId, task: "Follow up" },
);
expect(
await db
.select()
.from(browserUseRuns)
.where(eq(browserUseRuns.sessionId, start.sessionId)),
).toHaveLength(2);
const calls = f.request.mock.calls.filter(
([url, init]) => url.endsWith("/runs") && init.method === "POST",
);
expect(JSON.parse(String(calls[1][1].body)).sessionId).toBe(
idle.providerSessionId,
);
f.complete();
await f.tick();
await db
.update(browserUseSessions)
.set({ idleDeadline: new Date(0) })
.where(eq(browserUseSessions.id, start.sessionId));
await f.tick();
expect(
(
await db
.select()
.from(browserUseSessions)
.where(eq(browserUseSessions.id, start.sessionId))
)[0].status,
).toBe("closed");
});
it("serializes browser holds with parent reservations while provider work is delayed", async () => {
const f = await fixture();
const policy = await budgetService(db).upsertPolicy(f.company.id,
{ scopeType: "company", scopeId: f.company.id, amount: 100, hardStopEnabled: true }, actor.actorId);
await db.insert(budgetReservations).values({ companyId: f.company.id, runId: f.run.id, agentId: f.agent.id, amountCents: "40" });
let release!: () => void, started!: () => void;
const providerEntered = new Promise<void>(resolve => { started = resolve; });
const providerRelease = new Promise<void>(resolve => { release = resolve; });
f.delayCreate(async () => { started(); await providerRelease; });
const first = f.service.execute(f.binding, f.grant, randomUUID(), "browser_start", { task: "Read", maxCostUsd: 0.8 });
// Attach the rejection handler before waiting for the controlled POST.
await Promise.race([providerEntered, first.then(() => undefined)]);
try {
await expect(f.service.execute(f.binding, f.grant, randomUUID(), "browser_start", { task: "Concurrent" })).rejects.toThrow("reserved");
const [row] = await db.select().from(budgetPolicies).where(eq(budgetPolicies.id, policy.policyId));
expect(await computeObservedSpend(db, row)).toMatchObject({ ledgerCents: "0.0000000", effectiveHeldCents: "100.0000000" });
expect(f.request.mock.calls.filter(([url, init]) => url.endsWith("/runs") && init.method === "POST")).toHaveLength(1);
const call = f.request.mock.calls.find(([url, init]) => url.endsWith("/runs") && init.method === "POST")!;
expect(JSON.parse(String(call[1].body)).maxCostUsd).toBe(0.6);
// Accounting lock is not held while the hosted browser is waiting.
expect(await db.select({ id: companies.id }).from(companies).where(eq(companies.id, f.company.id))).toHaveLength(1);
} finally { release(); await first; }
});
it.each(["reduction", "new_policy", "config", "charge", "cancel", "session_end", "policy_deny"])("rechecks %s at the actual provider handoff", async race => {
const f = await fixture();
if (race === "reduction" || race === "charge") await budgetService(db).upsertPolicy(f.company.id,
{ scopeType: "company", scopeId: f.company.id, amount: 100, hardStopEnabled: true }, actor.actorId);
const invocationId = randomUUID();
const transaction = db.transaction.bind(db);
let injected = false;
const barrier = vi.spyOn(db, "transaction").mockImplementation(async (callback, config) => {
const result = await transaction(callback, config);
const [run] = await db.select().from(browserUseRuns).where(eq(browserUseRuns.invocationId, invocationId));
if (!injected && run?.status === "creating" && !(run.result as { paperclipBudgetHold?: { dispatchCommittedAt?: string } })?.paperclipBudgetHold?.dispatchCommittedAt) {
injected = true;
if (race === "reduction" || race === "new_policy") await budgetService(db).upsertPolicy(f.company.id,
{ scopeType: "company", scopeId: f.company.id, amount: race === "new_policy" ? 100 : 10, hardStopEnabled: true }, actor.actorId);
else if (race === "config") await f.service.saveSettings(f.company.id, f.grant.id, actor.actorId, { allowedProfileIds: [], maxCostUsd: 0.1 });
else if (race === "charge") {
const { costService } = await import("../services/costs.js");
await costService(db).createEvent(f.company.id, { agentId: f.agent.id, provider: "fixture", model: "fixture", costCents: 20, occurredAt: new Date() });
} else if (race === "policy_deny") await db.insert(toolPolicies).values({ companyId: f.company.id,
name: "Deny before browser handoff", policyType: "block", selectors: { connectionId: f.connection.connectionId }, priority: 0 });
else if (race === "session_end") await db.update(browserUseSessions).set({ status: "stopping", stopRequested: "end" }).where(eq(browserUseSessions.id, run.sessionId));
else await db.update(heartbeatRuns).set({ resultJson: { startupCancellation: { requestedAt: new Date().toISOString() } } }).where(eq(heartbeatRuns.id, f.run.id));
}
return result;
});
try {
await expect(f.service.execute(f.binding, f.grant, invocationId, "browser_start", { task: "Must not start" })).rejects.toThrow();
} finally { barrier.mockRestore(); }
expect(injected).toBe(true);
expect(f.request.mock.calls.filter(([url, init]) => url.endsWith("/runs") && init.method === "POST")).toHaveLength(0);
const [run] = await db.select().from(browserUseRuns).where(eq(browserUseRuns.invocationId, invocationId));
expect(run).toMatchObject({ status: "failed", eventsDrained: 1, result: { paperclipBudgetHold: { state: "released" } } });
});
it.each(["default_allow", "tool_name"])("retains valid %s profile authorization at browser admission", async selector => {
const f = await fixture();
await db.update(toolProfiles).set({ status: "inactive" }).where(eq(toolProfiles.companyId, f.company.id));
const [profile] = await db.insert(toolProfiles).values({ companyId: f.company.id, name: "Scoped browser access",
profileKey: randomUUID(), defaultAction: selector === "default_allow" ? "allow" : "deny" }).returning();
await db.insert(toolProfileBindings).values({ companyId: f.company.id, profileId: profile.id, targetType: "agent", targetId: f.agent.id });
if (selector === "tool_name") await db.insert(toolProfileEntries).values({ companyId: f.company.id, profileId: profile.id,
selectorType: "tool_name", toolName: "browser_start", effect: "include" });
await expect(f.service.execute(f.binding, f.grant, randomUUID(), "browser_start", { task: "Allowed by profile" })).resolves.toMatchObject({ status: "running" });
expect(f.request.mock.calls.filter(([url, init]) => url.endsWith("/runs") && init.method === "POST")).toHaveLength(1);
});
it("checks the gateway invocation's namespaced tool identity for explicit denial", async () => {
const f = await fixture();
const invocationId = randomUUID(), toolName = "app.browser-fixture:browser-start";
const [entry] = await db.select().from(toolCatalogEntries).where(and(eq(toolCatalogEntries.connectionId, f.connection.connectionId), eq(toolCatalogEntries.toolName, "browser_start")));
await db.insert(toolInvocations).values({ id: invocationId, companyId: f.company.id, agentId: f.agent.id,
runId: f.run.id, issueId: f.issue.id, actorType: "agent", actorId: f.agent.id,
connectionId: f.connection.connectionId, applicationId: entry.applicationId, catalogEntryId: entry.id,
toolName, upstreamToolName: "browser_start", policyDecision: "allow", status: "executing" });
await db.insert(toolPolicies).values({ companyId: f.company.id, name: "Block namespaced invocation",
policyType: "block", selectors: { toolName }, priority: 0 });
await expect(f.service.execute(f.binding, f.grant, invocationId, "browser_start", { task: "Denied" })).rejects.toThrow("blocked");
expect(f.request.mock.calls.filter(([url, init]) => url.endsWith("/runs") && init.method === "POST")).toHaveLength(0);
});
it("releases a crashed reservation that never reached provider dispatch authority", async () => {
const f = await fixture();
const policy = await budgetService(db).upsertPolicy(f.company.id,
{ scopeType: "agent", scopeId: f.agent.id, amount: 100, hardStopEnabled: true }, actor.actorId);
const [session] = await db.insert(browserUseSessions).values({ companyId: f.company.id, issueId: f.issue.id,
agentId: f.agent.id, connectionId: f.connection.connectionId, grantId: f.grant.id, status: "starting" }).returning();
const invocationId = randomUUID();
const [run] = await db.insert(browserUseRuns).values({ companyId: f.company.id, sessionId: session.id,
heartbeatRunId: f.run.id, invocationId, createdAt: new Date(Date.now() - 61_000),
result: { paperclipBudgetHold: { invocationId, amountCents: "50.0000000", state: "held", projectId: null } } }).returning();
f.request.mockClear();
await f.tick();
const [policyRow] = await db.select().from(budgetPolicies).where(eq(budgetPolicies.id, policy.policyId));
expect((await computeObservedSpend(db, policyRow)).effectiveHeldCents).toBe("0.0000000");
expect((await db.select().from(browserUseRuns).where(eq(browserUseRuns.id, run.id)))[0]).toMatchObject({ status: "failed", eventsDrained: 1, result: { paperclipBudgetHold: { state: "released" } } });
expect((await db.select().from(browserUseSessions).where(eq(browserUseSessions.id, session.id)))[0].status).toBe("failed");
expect(f.request.mock.calls.some(([url]) => new URL(url).pathname === "/api/v4/runs")).toBe(false);
});
it("releases a definitively rejected browser hold without inventing a charge", async () => {
const f = await fixture();
const policy = await budgetService(db).upsertPolicy(f.company.id,
{ scopeType: "agent", scopeId: f.agent.id, amount: 100, hardStopEnabled: true }, actor.actorId);
f.reject(409);
await expect(f.service.execute(f.binding, f.grant, randomUUID(), "browser_start", { task: "Rejected", maxCostUsd: 0.5 })).rejects.toThrow("busy");
const [row] = await db.select().from(budgetPolicies).where(eq(budgetPolicies.id, policy.policyId));
expect(await computeObservedSpend(db, row)).toMatchObject({ ledgerCents: "0.0000000", effectiveHeldCents: "0.0000000" });
expect(await db.select().from(costEvents).where(eq(costEvents.companyId, f.company.id))).toHaveLength(0);
await db.update(browserUseRuns).set({ result: null }).where(eq(browserUseRuns.companyId, f.company.id));
expect((await computeObservedSpend(db, row)).unknownBrowserRunCount).toBe(0);
});
it.each(["company", "agent", "project"] as const)("treats a running uncapped browser as unknown when a %s budget is introduced", async scopeType => {
for (const legacy of [false, true]) {
const f = await fixture();
const [project] = await db.insert(projects).values({ companyId: f.company.id, name: "Browser project" }).returning();
const [otherProject] = await db.insert(projects).values({ companyId: f.company.id, name: "Unrelated project" }).returning();
await db.update(issues).set({ projectId: project.id }).where(eq(issues.id, f.issue.id));
const invocation = randomUUID();
await expect(f.service.execute(f.binding, f.grant, invocation, "browser_start", { task: "Uncapped running work" })).resolves.toMatchObject({ status: "running" });
const [run] = await db.select().from(browserUseRuns).where(eq(browserUseRuns.invocationId, invocation));
expect(run.result).toMatchObject({ paperclipBudgetHold: { amountCents: "0.0000000", uncapped: true } });
expect(JSON.parse(String(f.request.mock.calls.find(([url, init]) => url.endsWith("/runs") && init.method === "POST")![1].body))).not.toHaveProperty("maxCostUsd");
if (legacy) await db.update(browserUseRuns).set({ result: null }).where(eq(browserUseRuns.id, run.id));
const scopeId = scopeType === "company" ? f.company.id : scopeType === "agent" ? f.agent.id : project.id;
const policy = await budgetService(db).upsertPolicy(f.company.id,
{ scopeType, scopeId, amount: 100, hardStopEnabled: true }, actor.actorId);
const [row] = await db.select().from(budgetPolicies).where(eq(budgetPolicies.id, policy.policyId));
expect(await computeObservedSpend(db, row)).toMatchObject({ ledgerCents: "0.0000000", effectiveHeldCents: "0.0000000", unknownBrowserRunCount: 1 });
expect((await computeObservedSpend(db, { ...row, scopeType: "project", scopeId: otherProject.id })).unknownBrowserRunCount).toBe(0);
expect(await budgetService(db).getInvocationBlock(f.company.id, f.agent.id, { projectId: project.id })).not.toBeNull();
f.request.mockClear();
await expect(f.service.execute(f.binding, f.grant, randomUUID(), "browser_start", { task: "Must wait for priced work" })).rejects.toThrow();
expect(f.request.mock.calls.filter(([url, init]) => url.endsWith("/runs") && init.method === "POST")).toHaveLength(0);
await budgetService(db).upsertPolicy(f.company.id,
{ scopeType, scopeId, amount: 100, hardStopEnabled: true, unpricedUsagePolicy: "allow" }, actor.actorId);
expect(await budgetService(db).getInvocationBlock(f.company.id, f.agent.id, { projectId: project.id })).toBeNull();
f.complete(); await f.tick();
expect(await computeObservedSpend(db, row)).toMatchObject({ ledgerCents: "15.0000000", effectiveHeldCents: "0.0000000", unknownBrowserRunCount: 0 });
}
});
it.each([
{ label: "explicit capped zero", uncapped: false, state: "held", dispatched: true, status: "running", expected: 0 },
{ label: "legacy zero cap without a flag", uncapped: undefined, state: "held", dispatched: true, status: "running", expected: 1 },
{ label: "uncapped before dispatch", uncapped: true, state: "held", dispatched: false, status: "creating", expected: 0 },
{ label: "failed before dispatch", uncapped: true, state: "released", dispatched: false, status: "failed", expected: 0 },
])("distinguishes $label from unfinished uncapped spending", async item => {
const f = await fixture();
const [session] = await db.insert(browserUseSessions).values({ companyId: f.company.id, issueId: f.issue.id,
agentId: f.agent.id, connectionId: f.connection.connectionId, grantId: f.grant.id, status: "running" }).returning();
const invocationId = randomUUID();
await db.insert(browserUseRuns).values({ companyId: f.company.id, sessionId: session.id, heartbeatRunId: f.run.id,
invocationId, status: item.status, providerRunId: item.dispatched ? randomUUID() : null, eventsDrained: item.status === "failed" ? 1 : 0,
result: { paperclipBudgetHold: { invocationId, amountCents: "0.0000000", projectId: null, state: item.state,
...(item.uncapped === undefined ? {} : { uncapped: item.uncapped }),
...(item.dispatched ? { dispatchCommittedAt: new Date().toISOString() } : {}) } } });
const policy = await budgetService(db).upsertPolicy(f.company.id,
{ scopeType: "company", scopeId: f.company.id, amount: 100, hardStopEnabled: true }, actor.actorId);
const [row] = await db.select().from(budgetPolicies).where(eq(budgetPolicies.id, policy.policyId));
expect(await computeObservedSpend(db, row)).toMatchObject({ ledgerCents: "0.0000000", effectiveHeldCents: "0.0000000", unknownBrowserRunCount: item.expected });
const block = await budgetService(db).getInvocationBlock(f.company.id, f.agent.id);
expect(Boolean(block)).toBe(item.expected > 0);
});
it("blocks new budgeted work after an uncapped browser POST becomes uncertain", async () => {
const f = await fixture();
f.loseReply();
const uncertain = randomUUID();
await expect(f.service.execute(f.binding, f.grant, uncertain, "browser_start", { task: "Uncapped" })).rejects.toThrow("could not be confirmed");
const policy = await budgetService(db).upsertPolicy(f.company.id,
{ scopeType: "company", scopeId: f.company.id, amount: 100, hardStopEnabled: true }, actor.actorId);
const [row] = await db.select().from(budgetPolicies).where(eq(budgetPolicies.id, policy.policyId));
expect(await computeObservedSpend(db, row)).toMatchObject({ ledgerCents: "0.0000000", effectiveHeldCents: "0.0000000", unknownBrowserRunCount: 1 });
f.request.mockClear();
await expect(f.service.execute(f.binding, f.grant, randomUUID(), "browser_start", { task: "Must wait" })).rejects.toThrow();
expect(f.request.mock.calls.filter(([url, init]) => url.endsWith("/runs") && init.method === "POST")).toHaveLength(0);
// The existing unpriced-usage policy remains the explicit operator override.
await budgetService(db).upsertPolicy(f.company.id,
{ scopeType: "company", scopeId: f.company.id, amount: 100, hardStopEnabled: true, unpricedUsagePolicy: "allow" }, actor.actorId);
expect(await budgetService(db).getInvocationBlock(f.company.id, f.agent.id)).toBeNull();
f.usage("0.15");
await f.tick();
expect(await computeObservedSpend(db, row)).toMatchObject({ ledgerCents: "15.0000000", unknownBrowserRunCount: 0 });
});
it("retains unknown browser holds and settles exact terminal usage without exposing internal metadata", async () => {
const f = await fixture();
const policy = await budgetService(db).upsertPolicy(f.company.id,
{ scopeType: "company", scopeId: f.company.id, amount: 100, hardStopEnabled: true }, actor.actorId);
f.loseReply();
const invocation = randomUUID();
await expect(f.service.execute(f.binding, f.grant, invocation, "browser_start", { task: "Read", maxCostUsd: 0.5 })).rejects.toThrow("could not be confirmed");
const [row] = await db.select().from(budgetPolicies).where(eq(budgetPolicies.id, policy.policyId));
expect((await computeObservedSpend(db, row)).effectiveHeldCents).toBe("50.0000000");
f.usage("0.150000001");
await f.tick();
expect(await computeObservedSpend(db, row)).toMatchObject({ ledgerCents: "15.0000001", effectiveHeldCents: "0.0000000" });
const [run] = await db.select().from(browserUseRuns).where(eq(browserUseRuns.invocationId, invocation));
expect(run.result).toMatchObject({ paperclipBudgetHold: { state: "settled", accountedCentsExact: "15.0000001" } });
const replay = await f.service.execute(f.binding, f.grant, invocation, "browser_start", { task: "Read" });
expect(JSON.stringify(replay)).not.toContain("paperclipBudgetHold");
expect(await db.select().from(costEvents).where(eq(costEvents.companyId, f.company.id))).toHaveLength(1);
});
it("retains exact cumulative accounting when an older browser row has no hold projection", async () => {
const f = await fixture();
const invocation = randomUUID();
await f.service.execute(f.binding, f.grant, invocation, "browser_start", { task: "Legacy result" });
const [run] = await db.select().from(browserUseRuns).where(eq(browserUseRuns.invocationId, invocation));
await db.update(browserUseRuns).set({ result: null }).where(eq(browserUseRuns.id, run.id));
f.usage("0.150000001"); f.complete(); await f.tick();
await db.update(browserUseRuns).set({ eventsDrained: 0 }).where(eq(browserUseRuns.id, run.id));
f.usage("0.150000002"); await f.tick();
const [total] = await db.select({ cents: sql<string>`sum(${costEvents.costCents})::text` }).from(costEvents).where(eq(costEvents.companyId, f.company.id));
expect(total.cents).toBe("15.0000002");
expect((await db.select().from(browserUseRuns).where(eq(browserUseRuns.id, run.id)))[0].result)
.toMatchObject({ paperclipBudgetHold: { state: "settled", accountedCentsExact: "15.0000002" } });
});
it("caps paid work at the remaining hard budget and enforces reported spend", async () => {
const f = await fixture();
await budgetService(db).upsertPolicy(
f.company.id,
{
scopeType: "agent",
scopeId: f.agent.id,
amount: 15,
hardStopEnabled: true,
},
actor.actorId,
);
await f.service.saveSettings(f.company.id, f.grant.id, actor.actorId, {
allowedProfileIds: [],
maxCostUsd: 0.5,
});
await f.service.execute(
f.binding,
f.grant,
randomUUID(),
"browser_start",
{ task: "Read", maxCostUsd: 1 },
);
const call = f.request.mock.calls.find(
([url, init]) => url.endsWith("/runs") && init.method === "POST",
)!;
expect(JSON.parse(String(call[1].body))).toMatchObject({
maxCostUsd: 0.15,
browserSettings: { record: false },
});
f.ready();
f.complete();
await f.tick();
await f.tick();
expect(
(await db.select().from(agents).where(eq(agents.id, f.agent.id)))[0],
).toMatchObject({ status: "paused", spentMonthlyCents: 15 });
expect(
(
await db
.select()
.from(browserUseSessions)
.where(eq(browserUseSessions.companyId, f.company.id))
)[0].status,
).toBe("closed");
await expect(
f.service.execute(f.binding, f.grant, randomUUID(), "browser_start", {
task: "Read again",
}),
).rejects.toThrow();
});
it("preserves the idle grace period when the task and parent finish before the last poll", async () => {
const f = await fixture();
await f.service.execute(
f.binding,
f.grant,
randomUUID(),
"browser_start",
{ task: "Read" },
);
f.ready();
f.complete();
await db
.update(heartbeatRuns)
.set({ status: "succeeded" })
.where(eq(heartbeatRuns.id, f.run.id));
await db
.update(issues)
.set({ status: "done" })
.where(eq(issues.id, f.issue.id));
await f.tick();
expect(
(
await db
.select()
.from(browserUseSessions)
.where(eq(browserUseSessions.companyId, f.company.id))
)[0].status,
).toBe("idle");
expect(
f.request.mock.calls.some(([url]) => url.endsWith("/cancel")),
).toBe(false);
});
it("confirms hosted shutdown before removing the connection credential", async () => {
const f = await fixture();
await f.service.execute(
f.binding,
f.grant,
randomUUID(),
"browser_start",
{ task: "Read" },
);
f.ready();
await f.tick();
await f.access.archiveConnection(
f.connection.connectionId,
f.company.id,
actor,
);
const [session] = await db
.select()
.from(browserUseSessions)
.where(eq(browserUseSessions.companyId, f.company.id));
expect(session.status).toBe("closed");
const [grant] = await db
.select()
.from(connectionGrants)
.where(eq(connectionGrants.id, f.grant.id));
expect(grant.status).toBe("revoked");
expect(grant.credentialSecretRefs).toEqual([]);
});
},
);