Files
PaperClipAI/packages/paperclip-runner/src/cli/cursor-cloud-app-server-proxy.ts
T
DottaandPaperclip 3910352c92 feat(runner): add experimental Cursor Cloud provider (#18068)
## Thinking Path

> - Paperclip is the open source app people use to manage AI agents for
work.
> - Runner gives each provider durable execution, tool, and completion
contracts.
> - Cursor Cloud currently has a direct adapter outside that native
contract.
> - Cloud work needs exact ownership across HTTP, SSE, Stop, and
recovery.
> - This PR adds a qualified Cloud API driver and verified proxy to
Runner.
> - Product setup and hosted placement remain separate dependent
changes.

## Linked Issues or Issue Description

Refs #5664, #11803, #17299, and #17471. Product integration is #18069.
Managed-only API placement is #18310. The related direct-adapter
model/transcript work is not replaced. No duplicate native Cursor Cloud
integration was found.

## What Changed

- Add native `cursor_cloud` / `cursor_cloud_api` with explicit
model/repository validation, bounded HTTP/SSE, durable per-turn
checkpoints and a qualified Node proxy launched through the Rust facade.
- Commit terminal custody before publication/release; coalesce
consumers, preserve ambiguous create/Stop/recovery fences, reject late
tools, and distinguish definitely-unsent oversized requests from
ambiguous responses.
- Support scoped cold attachment, current authority, reconnect and
verified artifact staging. Keep unsupported warm/external and Windows
paths closed.
- Preserve early source cursors and verify immutable legacy prefixes
before suffix replay, including supported compaction/masking. Optional
proof publication failure falls back to complete cold verification.
- Preserve canonical records at the durable native producer with
JSON-safe privacy and the existing 8MiB bound; drain admitted successor
flushes. Reconcile upstream semantic/pack/schema/lifecycle changes and
retain S3 and legacy CLI checks.

## Verification

Core head `37c442ec8dd4`: all 66 check contexts are terminal
successful/neutral/skipped, including required `ci / verify` and `ci /
e2e`; fresh exact-head Greptile is 5/5 with zero unresolved threads.
Dependent Product `81e828d20a77` and Top `dd2959689627` include the
later client handoff correction and are running their own fresh gates.

Core: 47/47 Rust, 4/4 compiled-native retirement, 19/19 settlement and
normal Runner build passed. Runtime Top `9af1a5de42a2`: build and
recursive typecheck passed. Earlier source-scoped evidence includes 398
focused tests, five relay tests and token/type gates. The 251-assertion,
7/8-suite run failed physical PostgreSQL teardown; the unchanged stable
WSP suite passed 128 through the documented wrapper.

The downstream fixture corrections passed 7/7 browser and 121/121
startup tests without timeout/oracle relaxation; a broken-unsubscribe
control fails. Hosted runtime `9af1a5de42a2` has a passing normal build
and recursive typecheck. Latest dependent heads additionally contain a
pure client fix for stale live/history overlap, with 48 focused tests,
UI typecheck and token gates passing. Its new client is not deployed in
the frozen hosted image; server/native/API/provider source is unchanged
by these four downstream files.

Fresh hosted Linux admission on `9af1a5de42a2` retired its owned
runner/proxy/socket with zero prompts. The saved configuration survived
reopening, and Test authenticated the bound key/catalog/selected model
without a prompt. Artifact with `claude-sonnet-4-5` completed,
downloaded the exact 27 bytes, matched 121 canonical identities against
complete logs and independently reached FINISHED. The active Stop retry
was clicked within its immutable boundary, produced one canonical
interrupted turn and acknowledged run-only cancellation, retained a
valid 35.076s quiet window (maximum gap 4.584s), and independently
remained CANCELLED/artifact-absent beyond the original 180-second
command deadline. All provider-observer credential callbacks, DB clients
and loaders closed; accounting is settled and explicitly unpriced. The
first fresh Stop retains its failed 10.373s sampling gap and every
sample.

On the upgraded9af server, the historical F226 artifact was reopened and
downloaded with its exact original hash after its provider agent had
been deleted/GET404 earlier today. Historical F226 full Artifact/Stop
cleanup, Question, corrected Plan and reconnect retain their original
identities. Deleting the three newly created disposable provider agents
and the new-case post-delete grade remain blocked on explicit user
approval after automatic review rejected expanding the two original
approved IDs. No new-case cleanup or complete fresh matrix pass is
claimed; this is separate from the directly observed task/Stop behavior
and retained exact cleanup proof.

Retained failures: stale fixture 132/133; E8 48/135 with 87 reset-hook
failures; Mac 17 SQL failures/skips; earlier startup-title, title-save,
custody/Stop and deletion failures with unknown causes and bounded
recoveries. `dea023` Artifact passed 17 checks but Stop failed its false
warning and 50 duplicate log identities. Pre-repair ACK failure and
original Core CI failure remain recorded; retry alone establishes no
cause or full local-suite pass. Provider dollar cost stays UNKNOWN.

## Risks

Experimental: Cursor independently acknowledges neither the effective
model nor instruction digest before the first prompt. The caller’s
explicit values are submitted and behavior checked live. Unpriced
provider dollar cost stays UNKNOWN.

Only qualified per-turn controllers are admitted. Uncertain work and
incomplete cleanup stay fenced. Existing damaged/duplicate archives
remain failed evidence. Prefix proof adds bounded object-store I/O; the
8MiB producer bound does not guarantee future providers’ output fits.

## Model Used

OpenAI Codex, GPT-6 family, assisted with reasoning, repository tools,
code execution, browser verification, and delegated code review. The
serving model’s exact API ID and context window are not exposed. No
value is inferred from the model catalog.

## Checklist

- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used and the limits of available model
metadata
- [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 linked public issues or described the issue in this PR
- [x] I have not referenced internal or instance-local Paperclip issues
or links
- [x] My branch describes the change and contains no internal ticket ID
- [x] I have run the scoped local tests reported above; pending gates
are stated separately
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation
- [x] I have considered and documented the risks
- [x] All required CI gates pass on the current head
- [x] Fresh Greptile is 5/5 with no unresolved findings on the current
head
- [x] I will address all reviewer comments before requesting merge

---------

Co-authored-by: Paperclip <noreply@paperclip.ing>
2026-11-03 08:38:57 -06:00

219 lines
15 KiB
JavaScript

#!/usr/bin/env node
import { constants } from "node:fs";
import { open as openFile } from "node:fs/promises";
import { dirname, resolve } from "node:path";
import { pathToFileURL } from "node:url";
import type { Readable, Writable } from "node:stream";
import type { HarnessSession, PersistedHarnessSession } from "../contracts/harness-driver.js";
import { parseNativeRuntimeContext } from "../contracts/runtime-context.js";
import { CursorCloudDriver, CURSOR_CLOUD_DRIVER_KIND, CURSOR_CLOUD_DRIVER_VERSION, type CursorCloudSessionHandle } from "../drivers/cursor-cloud/cursor-cloud-driver.js";
import { cursorCloudRecord, parseCursorCloudBinding, parseNativeCursorCloudConfig } from "../drivers/cursor-cloud/config.js";
import { enqueueSerialInput } from "./serial-input-queue.js";
type Rpc = { id?: string | number; method?: string; params?: unknown; result?: unknown; error?: unknown };
const MAX_FRAME_BYTES = 2_097_152;
const object = (value: unknown): Record<string, unknown> => value && typeof value === "object" && !Array.isArray(value) ? value as Record<string, unknown> : {};
const text = (value: unknown, fallback = "") => typeof value === "string" ? value : fallback;
async function privateJson(path: string, maxBytes: number): Promise<unknown> {
const handle = await openFile(path, constants.O_RDONLY | (constants.O_NOFOLLOW ?? 0));
try {
const stat = await handle.stat();
if (!stat.isFile() || stat.size > maxBytes || (stat.mode & 0o077) !== 0 || (process.getuid && stat.uid !== process.getuid())) throw new Error("Cursor Cloud configuration file must be owned, private and bounded");
const buffer = Buffer.alloc(stat.size + 1);
const { bytesRead } = await handle.read(buffer, 0, buffer.length, 0);
if (bytesRead !== stat.size) throw new Error("Cursor Cloud configuration changed during admission");
return JSON.parse(new TextDecoder("utf-8", { fatal: true }).decode(buffer.subarray(0, bytesRead)));
} finally { await handle.close(); }
}
export function startCursorCloudAppServerProxy(options: {
input?: Readable;
output?: Writable;
diagnostic?: Writable;
environment?: NodeJS.ProcessEnv;
} = {}): { done: Promise<void>; shutdown(): Promise<void> } {
const input = options.input ?? process.stdin;
const output = options.output ?? process.stdout;
const diagnostic = options.diagnostic ?? process.stderr;
const env = options.environment ?? process.env;
const pending = new Map<string, { resolve(value: unknown): void; reject(error: Error): void }>();
let sequence = 0;
let session: CursorCloudSessionHandle | null = null;
let cwd = "";
let model = "";
let closed = false;
let shuttingDown: Promise<void> | null = null;
let commandQueue = Promise.resolve();
let buffer = Buffer.alloc(0);
let bootstrapFailed = false;
let finish!: () => void;
const done = new Promise<void>(resolveDone => { finish = resolveDone; });
function send(value: unknown): void {
const encoded = `${JSON.stringify(value)}\n`;
if (Buffer.byteLength(encoded) > MAX_FRAME_BYTES || output.writableLength + Buffer.byteLength(encoded) > 4_194_304) throw new Error("Cursor Cloud output frame/consumer exceeded its bound");
output.write(encoded);
}
function requestController(params: Record<string, unknown>): Promise<unknown> {
if (closed || pending.size >= 128) return Promise.reject(new Error("Cursor Cloud controller is closed or exceeds its pending-call bound"));
const id = `cursor-cloud-${++sequence}`;
const promise = new Promise<unknown>((resolveValue, reject) => pending.set(id, { resolve: resolveValue, reject }));
send({ id, method: "item/tool/call", params });
return promise;
}
async function openSession(params: Record<string, unknown>, resume: boolean): Promise<unknown> {
if (session) throw new Error("Cursor Cloud thread is already open");
const configPath = env.PAPERCLIP_CURSOR_CLOUD_CONFIG_PATH?.trim();
if (!configPath) throw new Error("PAPERCLIP_CURSOR_CLOUD_CONFIG_PATH is required");
const raw = cursorCloudRecord(await privateJson(configPath, 65_536));
const { binding: rawBinding, ...publicConfig } = raw;
const binding = parseCursorCloudBinding(rawBinding);
const config = parseNativeCursorCloudConfig(publicConfig);
if (binding.normalizedSessionId !== env.PAPERCLIP_NORMALIZED_SESSION_ID) throw new Error("Cursor Cloud normalized session binding mismatch");
const runtimeContextPath = env.PAPERCLIP_NATIVE_RUNTIME_CONTEXT_PATH?.trim();
const runtimeContext = runtimeContextPath ? parseNativeRuntimeContext(await privateJson(runtimeContextPath, 1_048_576)) : null;
cwd = resolve(text(params.cwd, process.cwd())); model = text(params.model);
const runId = env.PAPERCLIP_RUN_ID;
if (!runId) throw new Error("PAPERCLIP_RUN_ID is required");
const relayUrl = env.PAPERCLIP_CURSOR_CLOUD_RELAY_URL;
const relayToken = env.PAPERCLIP_CURSOR_CLOUD_RELAY_TOKEN;
const socketPath = env.PAPERCLIP_CURSOR_CLOUD_RELAY_SOCKET_PATH;
const relayExpiresAt = Number(env.PAPERCLIP_CURSOR_CLOUD_RELAY_EXPIRES_AT);
if (!relayUrl || !relayToken || !socketPath || !Number.isSafeInteger(relayExpiresAt) || relayExpiresAt <= 0) throw new Error("Cursor Cloud semantic relay binding and expiry are required");
const driver = new CursorCloudDriver({ model, config, binding, runtimeDirectory: dirname(resolve(configPath)), artifactRoot: env.PAPERCLIP_CURSOR_CLOUD_ARTIFACT_ROOT, relay: { url: relayUrl, token: relayToken, socketPath, expiresAt: relayExpiresAt },
instructions: text(params.baseInstructions, "Complete only the supplied Paperclip task. Use Paperclip tools and finish with paperclip_finish or paperclip_block."), runtimeContext,
dynamicTools: Array.isArray(params.dynamicTools) ? params.dynamicTools.map(value => cursorCloudRecord(value)) : [],
dynamicToolHandler: async call => requestController({ ...call }), environment: env,
runnerInstanceId: env.PAPERCLIP_RUNNER_INSTANCE_ID ?? "cursor-cloud-sidecar" });
if (resume) {
const threadId = text(params.threadId);
const snapshot: PersistedHarnessSession = { driverKind: CURSOR_CLOUD_DRIVER_KIND, driverSessionId: threadId, normalizedSessionId: binding.normalizedSessionId, runId, workingDirectory: cwd };
const restored = await driver.recoverSession(snapshot);
if (!restored.recovered || !restored.session) throw new Error(restored.reason ?? "Cursor Cloud recovery failed");
session = restored.session;
} else session = await driver.openSession({ runId, normalizedSessionId: binding.normalizedSessionId, workingDirectory: cwd });
void pump(session).catch(() => shutdown());
return { thread: { id: session.ids().driverSessionId, sessionId: session.ids().driverSessionId, cwd }, submittedModel: model, effectiveModelAcknowledged: false, instructionsAcknowledged: false };
}
async function pump(opened: HarnessSession): Promise<void> {
for await (const event of opened.events()) {
const payload = event.payload;
const threadId = opened.ids().driverSessionId;
const turnId = event.turnId;
if (event.eventType === "turn.started") send({ method: "turn/started", params: { threadId, turnId, turn: { id: turnId, status: "inProgress" } } });
else if (["item.started", "item.delta", "item.completed"].includes(event.eventType)) {
const assistant = payload.kind === "agentMessage";
const nativeItem = object(payload.item);
const item = payload.kind === "tool" ? { ...nativeItem, type: "builtinToolCall", tool: nativeItem.name,
...(nativeItem.result === undefined ? {} : { output: typeof nativeItem.result === "string" ? nativeItem.result : JSON.stringify(nativeItem.result) }) }
: payload.item ?? { id: event.itemId, type: assistant ? "agentMessage" : payload.kind, text: payload.text };
send({ method: event.eventType === "item.delta" && assistant ? "item/agentMessage/delta" : event.eventType.replace(".", "/"), params: {
threadId, turnId, itemId: event.itemId, ...payload, item,
...(event.eventType === "item.delta" && assistant ? { delta: payload.text } : {}),
} });
} else if (["turn.completed", "turn.failed", "turn.interrupted"].includes(event.eventType)) send({ method: event.eventType.replace(".", "/"), params: { threadId, turnId, ...(payload.cursorCloudTerminalDelivery ? { cursorCloudTerminalDelivery: payload.cursorCloudTerminalDelivery } : {}), turn: { id: turnId, status: event.eventType.slice(5), ...(payload.error ? { error: payload.error } : {}) } } });
else if (event.eventType === "harness.diagnostic") send({ method: "warning", params: { threadId, turnId, code: payload.code, message: payload.message ?? payload.code } });
}
}
async function stopReceipt(params: Record<string, unknown>): Promise<Record<string, unknown>> {
if (!session) return { remoteStopped: true, remoteRunId: null };
await session.interrupt?.({});
const state = await session.read?.() ?? {};
const expectedTurnId = typeof params.turnId === "string" ? params.turnId : null;
if (expectedTurnId !== null && expectedTurnId !== state.turnId) throw new Error("Cursor Cloud shutdown turn identity changed");
const stoppedTurnId = state.pendingTerminalDelivery ? state.turnId : expectedTurnId;
if (state.runId && state.phase !== "terminal") throw new Error("Cursor Cloud remote Stop is unconfirmed");
if (!state.runId && state.phase !== "idle") throw new Error("Cursor Cloud pending remote create is unconfirmed");
return { remoteStopped: true, threadId: session.ids().driverSessionId, turnId: stoppedTurnId ?? null, remoteAgentId: state.agentId ?? null, remoteRunId: state.runId ?? null, ...(state.pendingTerminalDelivery ? { cursorCloudTerminalDelivery: state.pendingTerminalDelivery } : {}), ...(state.providerStatus ? { status: state.providerStatus } : {}) };
}
async function handle(message: Rpc): Promise<void> {
if (message.method === undefined && message.id !== undefined) {
const waiter = pending.get(String(message.id)); if (!waiter) return;
pending.delete(String(message.id));
if (message.error !== undefined) waiter.reject(new Error("Cursor Cloud controller rejected its tool request")); else waiter.resolve(message.result);
return;
}
if (!message.method || message.id === undefined) return;
const params = object(message.params);
let result: unknown;
switch (message.method) {
case "initialize": result = { user: { sessionId: "cursor-cloud" }, serverInfo: { name: "cursor-cloud", version: CURSOR_CLOUD_DRIVER_VERSION } }; break;
case "thread/start": result = await openSession(params, false); break;
case "thread/resume": result = await openSession(params, true); break;
case "collaborationMode/list": result = { data: [{ name: "Default", mode: "default", model, reasoningEffort: null }] }; break;
case "turn/start": {
if (!session) throw new Error("Cursor Cloud thread is not open");
if (object(params.collaborationMode).mode === "plan" || params.reasoningMode !== undefined) throw new Error("Cursor Cloud native plan/reasoning override is unsupported");
const items = Array.isArray(params.input) ? params.input.map(object) : [];
if (items.some(item => item.type !== "text" && item.type !== "inputText")) throw new Error("Cursor Cloud supports text turn input only");
const turn = await session.startTurn({ message: { role: "user", text: items.map(item => text(item.text)).join("\n") } });
result = { turn: { id: turn.turnId, status: "inProgress" } }; break;
}
case "turn/interrupt":
if (!session?.interrupt) throw new Error("Cursor Cloud thread is not open");
await session.interrupt({ ...(typeof params.turnId === "string" ? { turnId: params.turnId } : {}) }); result = true; break;
case "thread/read": {
if (!session) throw new Error("Cursor Cloud thread is not open");
await session.snapshot();
const state = await session.read();
const activeTurnId = ["creating", "running", "stopping"].includes(String(state.phase)) && typeof state.turnId === "string" ? state.turnId : null;
result = { cursorCloudTerminalDelivery: state.pendingTerminalDelivery, thread: { id: session.ids().driverSessionId, cwd, turns: activeTurnId ? [{ id: activeTurnId, status: "inProgress" }] : [] }, transcript: state }; break;
}
case "paperclip/ackTerminal":
if (!session || typeof params.turnId !== "string" || typeof params.receiptId !== "string") throw new Error("Cursor Cloud terminal receipt is missing");
await session.acknowledgeTerminalDelivery({ turnId: params.turnId, receiptId: params.receiptId });
result = { turnId: params.turnId, receiptId: params.receiptId, acknowledged: true }; break;
case "paperclip/shutdown": result = await stopReceipt(params); break;
default: throw new Error("Unsupported Cursor Cloud proxy method");
}
send({ id: message.id, result });
}
async function shutdown(): Promise<void> {
shuttingDown ??= (async () => {
closed = true;
for (const waiter of pending.values()) waiter.reject(new Error("Cursor Cloud controller closed")); pending.clear();
try { await session?.close({ reason: "proxy_shutdown", force: true }); }
catch { diagnostic.write("[cursor-cloud] Remote termination is unconfirmed; checkpoint retained.\n"); process.exitCode = 1; }
finish();
})();
return shuttingDown;
}
function receive(message: Rpc): void {
if (closed) return;
if (message.method === undefined && message.id !== undefined) { void handle(message); return; }
commandQueue = enqueueSerialInput(commandQueue, async () => {
if (bootstrapFailed && message.method !== "paperclip/shutdown") throw new Error("Cursor Cloud bootstrap previously failed");
await handle(message);
}, error => {
if (["initialize", "thread/start", "thread/resume"].includes(message.method ?? "")) bootstrapFailed = true;
const reason = error instanceof Error ? error.message : "Cursor Cloud request failed";
send({ id: message.id, error: { code: -32000, message: reason.slice(0, 2_000) } });
});
}
input.on("data", (chunk: Buffer | string) => {
if (closed) return;
buffer = Buffer.concat([buffer, Buffer.isBuffer(chunk) ? chunk : Buffer.from(chunk)]);
while (true) {
const index = buffer.indexOf(10);
if (index < 0) break;
const line = buffer.subarray(0, index); buffer = buffer.subarray(index + 1);
if (line.length > MAX_FRAME_BYTES) { void shutdown(); return; }
if (!line.toString().trim()) continue;
try { const message = JSON.parse(line.toString()) as Rpc; receive(message); }
catch { send({ id: null, error: { code: -32700, message: "Invalid Cursor Cloud JSON-RPC input" } }); }
}
if (buffer.length > MAX_FRAME_BYTES) void shutdown();
});
input.on("end", () => { void commandQueue.then(shutdown); });
input.on("error", () => { void shutdown(); });
return { done, shutdown };
}
if (process.argv[1] && import.meta.url === pathToFileURL(resolve(process.argv[1])).href) {
const proxy = startCursorCloudAppServerProxy();
process.once("SIGTERM", () => { void proxy.shutdown(); });
process.once("SIGINT", () => { void proxy.shutdown(); });
void proxy.done.then(() => process.exit(process.exitCode ?? 0));
}