| 1 | // Transport: turn a registered computer into an executor. |
| 2 | // - local: the Codewhale Computer Use app when it is running or registered |
| 3 | // (it owns the OS permissions), otherwise spawn directly |
| 4 | // - ssh: run the codewhale-cu remote agent over ssh (args travel as base64 JSON, |
| 5 | // so no tool argument can ever become remote shell syntax) |
| 6 | // - hdc: HarmonyOS device over `hdc` shell / file push-pull |
| 7 | import { run, runOk, runInputLease, ExecError, currentSignal } from "./exec.mjs"; |
| 8 | import { ensureApp, appSessionRequest } from "./app-socket.mjs"; |
| 9 | import { spawn } from "node:child_process"; |
| 10 | import crypto from "node:crypto"; |
| 11 | import fs from "node:fs"; |
| 12 | import path from "node:path"; |
| 13 | import os from "node:os"; |
| 14 | import url from "node:url"; |
| 15 | |
| 16 | const __dirname = path.dirname(url.fileURLToPath(import.meta.url)); |
| 17 | export const PLUGIN_ROOT = path.resolve(__dirname, ".."); |
| 18 | // A new MCP process always starts a fresh app/input/raster binding, even when |
| 19 | // the permission-owning desktop helper remains running across tasks. |
| 20 | export const SESSION_ID = crypto.randomUUID(); |
| 21 | let usedApp = false; |
| 22 | let appSessionClosed = false; |
| 23 | export function closeAppSession({ releaseOnly = false } = {}) { |
| 24 | if (!usedApp || appSessionClosed) return Promise.resolve(); |
| 25 | return appSessionRequest({ tool: releaseOnly ? "release_session_input" : "close_session", sessionId: SESSION_ID }, { timeoutMs: 2_500, signal: null }).then((reply) => { |
| 26 | if (!reply?.ok) throw Object.assign(new ExecError(reply?.error?.message ?? "Computer input cleanup failed"), { code: reply?.error?.code ?? "input_release_failed" }); |
| 27 | if (!releaseOnly) appSessionClosed = true; |
| 28 | }); |
| 29 | } |
| 30 | |
| 31 | export function b64(obj) { |
| 32 | return Buffer.from(JSON.stringify(obj), "utf8").toString("base64"); |
| 33 | } |
| 34 | |
| 35 | /** |
| 36 | * Validate a remote-side filesystem path we construct ourselves. |
| 37 | * Blocks shell metacharacters and traversal outside the agent dir. |
| 38 | */ |
| 39 | export function safeRemotePath(p) { |
| 40 | if (typeof p !== "string" || !/^[A-Za-z0-9.][A-Za-z0-9/._-]{0,511}$/.test(p) || p.includes("..")) { |
| 41 | throw new ExecError(`refusing unsafe remote path: ${JSON.stringify(p)}`); |
| 42 | } |
| 43 | return p; |
| 44 | } |
| 45 | |
| 46 | /** |
| 47 | * Local executor bound to a platform backend name. |
| 48 | * All backends receive this shape. |
| 49 | */ |
| 50 | export function localExec() { |
| 51 | return { |
| 52 | kind: "local", |
| 53 | run, |
| 54 | runOk, |
| 55 | runInputLease, |
| 56 | async readFile(p) { return fs.promises.readFile(p); }, |
| 57 | async writeFile(p, data) { return fs.promises.writeFile(p, data); }, |
| 58 | tmpFile(prefix) { |
| 59 | return path.join(fs.mkdtempSync(path.join(os.tmpdir(), prefix)), "out"); |
| 60 | }, |
| 61 | }; |
| 62 | } |
| 63 | |
| 64 | /** |
| 65 | * App executor: the local computer driven through the desktop app's socket. |
| 66 | * Same `remote()` contract as ssh, but files the app writes are on this disk. |
| 67 | */ |
| 68 | export function appExec(app, sessionId = SESSION_ID) { |
| 69 | return { |
| 70 | ...localExec(), |
| 71 | kind: "app", |
| 72 | app, |
| 73 | filesLocal: true, |
| 74 | remote(request, opts = {}) { |
| 75 | if (sessionId === SESSION_ID && appSessionClosed) throw Object.assign(new ExecError("Computer session was closed; start a new MCP session to use the local helper again"), { code: "app_session_closed" }); |
| 76 | usedApp = true; |
| 77 | return appSessionRequest({ ...request, sessionId }, { timeoutMs: opts.timeoutMs ?? 30_000 }); |
| 78 | }, |
| 79 | }; |
| 80 | } |
| 81 | |
| 82 | /** |
| 83 | * A persistent ssh agent channel: one `ssh host node agent.mjs --serve` |
| 84 | * process carrying base64-JSON request lines in and JSON receipt lines out. |
| 85 | * Unlike the one-shot agent it keeps its backend alive between calls, so an |
| 86 | * open_application binding survives into later raw-input calls and held |
| 87 | * input/recording can be owned by the session. Requests written before the |
| 88 | * channel dies may already have run remotely — their failures are marked |
| 89 | * requestDispatched so the server reports outcome_unknown instead of |
| 90 | * inviting a blind retry. |
| 91 | */ |
| 92 | export function ensureSshChannel(binding, argv) { |
| 93 | let ch = binding.sshChannel; |
| 94 | if (ch?.alive) return ch; |
| 95 | const next = { alive: false, restarted: !!ch, everReplied: false, pending: new Map(), seq: 1, buf: "", proc: null }; |
| 96 | binding.sshChannel = next; |
| 97 | const failAll = (err) => { |
| 98 | for (const [, p] of next.pending) { clearTimeout(p.timer); p.reject(err); } |
| 99 | next.pending.clear(); |
| 100 | }; |
| 101 | let proc; |
| 102 | try { |
| 103 | proc = spawn(argv[0], argv.slice(1), { stdio: ["pipe", "pipe", "pipe"], windowsHide: true }); |
| 104 | } catch (err) { |
| 105 | next.spawnError = err; |
| 106 | return next; |
| 107 | } |
| 108 | next.proc = proc; |
| 109 | next.alive = true; |
| 110 | proc.stdin.on("error", () => {}); |
| 111 | proc.stderr.on("data", () => {}); // drain; stderr is never parsed |
| 112 | proc.stdout.setEncoding("utf8"); |
| 113 | proc.stdout.on("data", (d) => { |
| 114 | next.buf += d; |
| 115 | let i; |
| 116 | while ((i = next.buf.indexOf("\n")) !== -1) { |
| 117 | const line = next.buf.slice(0, i).trim(); |
| 118 | next.buf = next.buf.slice(i + 1); |
| 119 | if (!line.startsWith("{")) continue; // MOTD/banner noise |
| 120 | let msg; |
| 121 | try { msg = JSON.parse(line); } catch { continue; } |
| 122 | next.everReplied = true; |
| 123 | const p = next.pending.get(msg.id); |
| 124 | if (!p) continue; // timed-out or unknown request: drop the late reply |
| 125 | next.pending.delete(msg.id); |
| 126 | clearTimeout(p.timer); |
| 127 | p.resolve(msg); |
| 128 | } |
| 129 | }); |
| 130 | const dead = (why) => { |
| 131 | if (!next.alive) return; |
| 132 | next.alive = false; |
| 133 | failAll(Object.assign(new ExecError(`ssh agent channel closed${why ? `: ${why}` : ""}`), { code: "remote_session_lost", requestDispatched: true })); |
| 134 | }; |
| 135 | proc.on("error", (err) => dead(String(err?.message ?? err))); |
| 136 | proc.on("close", (code, sig) => dead(code != null ? `exited ${code}` : `signal ${sig}`)); |
| 137 | return next; |
| 138 | } |
| 139 | |
| 140 | export function closeSshChannel(binding) { |
| 141 | const ch = binding.sshChannel; |
| 142 | if (!ch) return; |
| 143 | binding.sshChannel = null; |
| 144 | ch.alive = false; |
| 145 | try { ch.proc?.stdin.end(); } catch {} |
| 146 | try { ch.proc?.kill("SIGTERM"); } catch {} |
| 147 | for (const [, p] of ch.pending ?? []) { |
| 148 | clearTimeout(p.timer); |
| 149 | p.reject(Object.assign(new ExecError("ssh agent channel closed"), { code: "remote_session_lost", requestDispatched: true })); |
| 150 | } |
| 151 | ch.pending?.clear(); |
| 152 | } |
| 153 | |
| 154 | export function channelRequest(ch, request, timeoutMs) { |
| 155 | return new Promise((resolve, reject) => { |
| 156 | if (!ch.alive) { |
| 157 | reject(Object.assign(new ExecError("ssh agent channel is closed"), { code: "remote_session_lost" })); |
| 158 | return; |
| 159 | } |
| 160 | const id = ch.seq++; |
| 161 | const timer = setTimeout(() => { |
| 162 | ch.pending.delete(id); |
| 163 | // The request was written; the remote may still be executing it. |
| 164 | reject(Object.assign(new ExecError(`ssh agent timed out after ${timeoutMs}ms`), { code: "remote_timeout", requestDispatched: true })); |
| 165 | }, timeoutMs); |
| 166 | ch.pending.set(id, { resolve, reject, timer }); |
| 167 | ch.proc.stdin.write(b64({ id, tool: request.tool, args: request.args ?? {} }) + "\n"); |
| 168 | }); |
| 169 | } |
| 170 | |
| 171 | /** |
| 172 | * Shared persistent-channel front for executors whose requests ride one |
| 173 | * long-lived `<argv> --serve` process (ssh agent, docker exec). Read-only and |
| 174 | * identity requests may run on a restarted channel; input tools may not — |
| 175 | * the fresh remote agent no longer holds this session's open_application |
| 176 | * binding. |
| 177 | */ |
| 178 | const SAFE_AFTER_RESTART = new Set([ |
| 179 | "platform", "probe", "list_displays", "switch_display", "list_apps", "list_windows", |
| 180 | "get_app_state", "resolve_element", "screenshot", "zoom", "cursor_position", |
| 181 | "read_clipboard", "recordingList", "recordingStatus", "open_application", "preview", |
| 182 | ]); |
| 183 | function attachPersistentChannel(ex, binding, serveArgv) { |
| 184 | if (!binding) return; |
| 185 | ex.persistent = (request, opts = {}) => { |
| 186 | const ch = ensureSshChannel(binding, serveArgv); |
| 187 | if (ch.spawnError) { |
| 188 | return Promise.reject(Object.assign(new ExecError(`remote agent channel failed to start: ${ch.spawnError.message}`), { code: "remote_session_lost" })); |
| 189 | } |
| 190 | if (ch.restarted) { |
| 191 | ch.restarted = false; |
| 192 | binding.needsObservation = true; |
| 193 | if (!SAFE_AFTER_RESTART.has(request.tool)) { |
| 194 | return Promise.reject(Object.assign(new ExecError("the remote agent session restarted — rebind with open_application and observe before acting"), { code: "remote_session_restarted" })); |
| 195 | } |
| 196 | } |
| 197 | return channelRequest(ch, request, opts.timeoutMs ?? 25_000); |
| 198 | }; |
| 199 | ex.closeChannel = () => closeSshChannel(binding); |
| 200 | } |
| 201 | |
| 202 | /** ssh executor: speaks to the remote agent installed by installRemoteAgent(). */ |
| 203 | export function sshExec(computer, binding) { |
| 204 | const userHost = computer.user ? `${computer.user}@${computer.host}` : computer.host; |
| 205 | const portArgs = computer.port ? ["-p", String(computer.port)] : []; |
| 206 | const remoteAgent = safeRemotePath(computer.agentPath ?? ".codewhale-cu/agent/agent.mjs"); |
| 207 | const base = ["-o", "BatchMode=yes", "-o", "ConnectTimeout=8", "-o", "StrictHostKeyChecking=accept-new", ...portArgs, userHost]; |
| 208 | const ex = { |
| 209 | kind: "ssh", |
| 210 | base, |
| 211 | userHost, |
| 212 | remoteAgent, |
| 213 | run(cmd, args = [], opts = {}) { |
| 214 | // Local side commands (e.g. ssh itself) run directly. |
| 215 | return run(cmd, args, opts); |
| 216 | }, |
| 217 | async remote(request, opts = {}) { |
| 218 | const r = await run("ssh", [...base, "node", remoteAgent, b64({ args: request.args ?? {}, tool: request.tool, nonce: crypto.randomBytes(6).toString("hex") })], { |
| 219 | timeoutMs: opts.timeoutMs ?? 25_000, |
| 220 | }); |
| 221 | if (r.aborted) throw Object.assign(new ExecError("computer request cancelled", r), { code: "cancelled" }); |
| 222 | if (r.timedOut) throw new ExecError(`ssh ${userHost}: timed out`, r); |
| 223 | if (r.code !== 0) throw new ExecError(`ssh ${userHost} exited ${r.code}: ${r.stderr.trim().slice(0, 400)}`, r); |
| 224 | // The agent prints exactly one JSON line; anything before it is MOTD noise. |
| 225 | const line = r.stdout.trim().split("\n").filter((l) => l.startsWith("{")).pop(); |
| 226 | const reply = line ? JSON.parse(line) : null; |
| 227 | if (!reply) throw new ExecError(`ssh ${userHost}: agent returned no JSON receipt`, r); |
| 228 | return reply; |
| 229 | }, |
| 230 | }; |
| 231 | attachPersistentChannel(ex, binding, ["ssh", ...base, "node", remoteAgent, "--serve"]); |
| 232 | return ex; |
| 233 | } |
| 234 | |
| 235 | /** |
| 236 | * docker executor: a spawned task-owned desktop container. Same agent contract |
| 237 | * as ssh, but the channel is `docker exec` — no sshd, no keys, the container |
| 238 | * boundary itself is the isolation. Every call goes through |
| 239 | * docker/agent-exec.sh, which joins the desktop session env (display + bus) |
| 240 | * the container entrypoint recorded before serving requests. |
| 241 | */ |
| 242 | export function dockerExec(computer, binding) { |
| 243 | const container = safeRemotePath(computer.container); |
| 244 | const remoteAgent = "/app/docker/agent-exec.sh"; |
| 245 | const ex = { |
| 246 | kind: "docker", |
| 247 | container, |
| 248 | remoteAgent, |
| 249 | run(cmd, args = [], opts = {}) { |
| 250 | // Local side commands (docker itself) run directly. |
| 251 | return run(cmd, args, opts); |
| 252 | }, |
| 253 | async remote(request, opts = {}) { |
| 254 | const r = await run("docker", ["exec", container, "/bin/sh", remoteAgent, b64({ args: request.args ?? {}, tool: request.tool, nonce: crypto.randomBytes(6).toString("hex") })], { |
| 255 | timeoutMs: opts.timeoutMs ?? 25_000, |
| 256 | }); |
| 257 | if (r.aborted) throw Object.assign(new ExecError("computer request cancelled", r), { code: "cancelled" }); |
| 258 | if (r.timedOut) throw new ExecError(`docker exec ${container}: timed out`, r); |
| 259 | if (r.code !== 0) throw new ExecError(`docker exec ${container} exited ${r.code}: ${r.stderr.trim().slice(0, 400)}`, r); |
| 260 | const line = r.stdout.trim().split("\n").filter((l) => l.startsWith("{")).pop(); |
| 261 | const reply = line ? JSON.parse(line) : null; |
| 262 | if (!reply) throw new ExecError(`docker exec ${container}: agent returned no JSON receipt`, r); |
| 263 | return reply; |
| 264 | }, |
| 265 | }; |
| 266 | attachPersistentChannel(ex, binding, ["docker", "exec", "-i", container, "/bin/sh", remoteAgent, "--serve"]); |
| 267 | return ex; |
| 268 | } |
| 269 | |
| 270 | /** Push the self-contained remote agent + src tree to an ssh computer. */ |
| 271 | export async function installRemoteAgent(computer) { |
| 272 | const ex = sshExec(computer); |
| 273 | const srcDir = path.join(PLUGIN_ROOT, "src"); |
| 274 | const rels = ["agent.mjs"]; |
| 275 | for (const dir of ["", "backends"]) { |
| 276 | const full = path.join(srcDir, dir); |
| 277 | for (const f of fs.readdirSync(full)) { |
| 278 | if (f.endsWith(".mjs") || f.endsWith(".m") || f.endsWith(".h")) rels.push(`src/${dir ? dir + "/" : ""}${f}`); |
| 279 | } |
| 280 | } |
| 281 | const marker = ".codewhale-cu/agent"; |
| 282 | let r = await run("ssh", [...ex.base, "mkdir", "-p", `${marker}/src/backends`], { timeoutMs: 15_000 }); |
| 283 | if (r.code !== 0) throw new ExecError(`ssh ${ex.userHost}: mkdir failed: ${r.stderr.trim().slice(0, 300)}`, r); |
| 284 | for (const rel of rels) { |
| 285 | const localPath = rel === "agent.mjs" ? path.join(PLUGIN_ROOT, "agent.mjs") : path.join(srcDir, rel.slice(4)); |
| 286 | const dest = safeRemotePath(`${marker}/${rel}`); |
| 287 | r = await run("scp", [...(computer.port ? ["-P", String(computer.port)] : []), localPath, `${ex.userHost}:${dest}`], { timeoutMs: 30_000 }); |
| 288 | if (r.code !== 0) throw new ExecError(`scp ${rel} failed: ${r.stderr.trim().slice(0, 300)}`, r); |
| 289 | } |
| 290 | // Probe remote platform via the agent itself. |
| 291 | const reply = await ex.remote({ tool: "platform" }); |
| 292 | return { installed: rels.length, remotePlatform: reply.platform, agentPath: `${marker}/agent.mjs` }; |
| 293 | } |
| 294 | |
| 295 | /** hdc (HarmonyOS) executor. Commands run on-device; files pull to local tmp. */ |
| 296 | export function hdcExec(computer) { |
| 297 | const targetArgs = computer.target ? ["-t", computer.target] : []; |
| 298 | const shell = (args, opts = {}) => run("hdc", [...targetArgs, "shell", ...args], opts); |
| 299 | return { |
| 300 | kind: "hdc", |
| 301 | targetArgs, |
| 302 | run, |
| 303 | runOk, |
| 304 | shell, |
| 305 | async pullFile(remotePath, localPath, opts = {}) { |
| 306 | // HDC device captures use absolute paths; SSH agent paths are relative. |
| 307 | // Validate the remaining path with the same traversal/metacharacter guard. |
| 308 | safeRemotePath(typeof remotePath === "string" ? remotePath.replace(/^\//, "") : remotePath); |
| 309 | const r = await run("hdc", [...targetArgs, "file", "recv", remotePath, localPath], opts); |
| 310 | if (r.code !== 0) throw new ExecError(`hdc file recv failed: ${r.stderr.trim().slice(0, 300)}`, r); |
| 311 | return localPath; |
| 312 | }, |
| 313 | async readFile(remotePath, opts = {}) { |
| 314 | // Containment: pull into a private mkdtemp dir and remove exactly that |
| 315 | // dir. Never rm() the parent of a file placed directly in os.tmpdir() — |
| 316 | // that recursively deletes the entire user temp directory. |
| 317 | const dir = await fs.promises.mkdtemp(path.join(os.tmpdir(), "cu-hdc-")); |
| 318 | try { |
| 319 | const tmp = path.join(dir, "out"); |
| 320 | await this.pullFile(remotePath, tmp, opts); |
| 321 | return await fs.promises.readFile(tmp); |
| 322 | } finally { |
| 323 | // Cleanup must not replace downloaded bytes or the original I/O error. |
| 324 | await fs.promises.rm(dir, { recursive: true, force: true }).catch(() => {}); |
| 325 | } |
| 326 | }, |
| 327 | }; |
| 328 | } |
| 329 | |
| 330 | export async function executorFor(computer, binding) { |
| 331 | if (computer.transport === "local") { |
| 332 | // Test hook: exercise the out-of-process wire path (desktop app / ssh |
| 333 | // agent) in-process, so wire argument preparation is covered by tests. |
| 334 | if (process.env.CODEWHALE_CU_TEST_REMOTE === "1") { |
| 335 | const { handle } = await import("./app-handler.mjs"); |
| 336 | return { ...appExec({ id: "test", name: "test app" }), remote: (request) => handle(request, { sessionId: SESSION_ID, signal: currentSignal() }) }; |
| 337 | } |
| 338 | const status = await ensureApp(); |
| 339 | if (status.via === "app") { |
| 340 | if (status.app.sessionProtocol !== 2) throw Object.assign(new ExecError("The installed Computer Use helper needs an update for isolated sessions and disconnect cleanup. Rebuild/reinstall it, then retry."), { code: "app_upgrade_required" }); |
| 341 | if (process.platform === "darwin" && status.app.backgroundProtocol !== 1) throw Object.assign(new ExecError("The installed Computer Use app predates background scrolling, scoped observations and foreground preemption. Update and restart the helper before using it."), { code: "app_upgrade_required" }); |
| 342 | return appExec(status.app); |
| 343 | } |
| 344 | return { ...localExec(), appReason: status.reason }; |
| 345 | } |
| 346 | if (computer.transport === "ssh") return sshExec(computer, binding); |
| 347 | if (computer.transport === "docker") return dockerExec(computer, binding); |
| 348 | if (computer.transport === "hdc") return hdcExec(computer); |
| 349 | throw new ExecError(`unknown transport ${computer.transport}`); |
| 350 | } |
| 351 | |
| 352 | /** |
| 353 | * Map a computer to its backend module. Local platform is fixed; ssh |
| 354 | * computers may carry platformHint (probed at registration). |
| 355 | */ |
| 356 | function effectivePlatform(computer) { |
| 357 | let platform = computer.platform ?? computer.platformHint; |
| 358 | if (!platform) { |
| 359 | if (computer.transport === "local") platform = process.platform; |
| 360 | else if (computer.transport === "hdc") platform = "harmonyos"; |
| 361 | else if (computer.transport === "docker") platform = "linux"; // spawned containers are always the Linux desktop image |
| 362 | else platform = "linux"; // conservative default for ssh; registration probes it |
| 363 | } |
| 364 | return platform; |
| 365 | } |
| 366 | |
| 367 | /** Identity of the effective route, excluding catalog presentation metadata. */ |
| 368 | export function routeFingerprint(computer) { |
| 369 | const route = [computer.transport, effectivePlatform(computer)]; |
| 370 | if (computer.transport === "docker") route.push(computer.container || null); |
| 371 | if (computer.transport === "hdc") route.push(computer.target || null); |
| 372 | if (computer.transport === "ssh") route.push(computer.host, computer.user || null, |
| 373 | computer.port || null, computer.agentPath ?? ".codewhale-cu/agent/agent.mjs"); |
| 374 | return JSON.stringify(route); |
| 375 | } |
| 376 | |
| 377 | export async function backendFor(computer) { |
| 378 | const platform = effectivePlatform(computer); |
| 379 | // Test hook: inject a fake local backend by absolute path to an .mjs |
| 380 | // module exporting `create` (used by tests/, never set in production). |
| 381 | const testBackend = computer.transport === "local" && process.env.CODEWHALE_CU_TEST_BACKEND; |
| 382 | const mod = await import(testBackend ? url.pathToFileURL(testBackend).href : `./backends/${platform}.mjs`); |
| 383 | // Backends always get a direct executor; routing through the app happens |
| 384 | // one level up (the server dispatches to `executor.remote` when present). |
| 385 | const exec = computer.transport === "local" ? localExec() : await executorFor(computer); |
| 386 | return { backend: mod.create({ exec, computer, platform }), platform }; |
| 387 | } |
| 388 |