fix(cli): materialize the managed install before the onboarding service install (#12148)

## Thinking Path

> - Paperclip is the open source app people use to manage AI agents for
work
> - Interactive onboarding offers to install Paperclip as a background
service, defaulting to yes
> - The service definition targets the managed command shim, but an
ephemeral npx run never installs it, and the service step never checks
> - The result is a crash-looping service, a doctor hint about a
nonexistent port conflict, and a first run that ends with nothing
serving
> - This pull request materializes the managed install before
registering the service, or declines with the repair path
> - The benefit is that saying yes to the service prompt yields a
working service — or an honest explanation

## Linked Issues or Issue Description

**What happened?**

On a machine with no managed install, `npx paperclipai@2026.824.0
onboard` (interactive), accepting the background-service prompt,
produced: a LaunchAgent pointing at `~/.local/bin/paperclipai` (which
does not exist), launchd exit code 78 in a KeepAlive crash loop, doctor
reporting "inactive but the configured port is serving another Paperclip
process — stop the conflicting foreground process" (no such process
existed), and "Service health: fetch failed". Reproduced twice on a
clean field. `latest` has carried this path since v2026.817.0 shipped;
CI never sees it because `--yes` onboarding skips the service prompt.

**Expected behavior**

Accepting the service prompt installs a working service (materializing
the managed payload and shim first when needed), and doctor diagnoses a
missing service binary as exactly that.

**Steps to reproduce**

On macOS with no `~/.local/bin/paperclipai`: `npx paperclipai@latest
onboard`, accept the service prompt, then `launchctl print
gui/$UID/ing.paperclip.paperclipai` (exit code 78, spawn scheduled) and
`paperclipai doctor`.

**Paperclip version or commit**

`2026.824.0` (path present since #10045).

## What Changed

- `cli/src/onboard-service.ts`: after the user opts in, an
`ensureServiceShim` step checks the service shim path. Missing +
managed-store location → run `installCommand` pinned to the onboarding
version (payload, shim, PATH block), then proceed. Missing + custom
`PAPERCLIP_SHIM_PATH`, or install failure → decline with `paperclipai
install` / `paperclipai service install` guidance and install nothing.
- `cli/src/checks/service-health-check.ts`: the runtime check diagnoses
a missing service binary with the install repair hint (instead of the
port-conflict hint); an inactive service with a healthy responder gets a
`warn` attributing the foreign process instead of a plain "Healthy"
pass.
- Tests: new cases for shim materialization ordering,
decline-on-failure, missing-binary diagnosis, and foreign-responder
attribution; existing fixtures updated to inject the new dependencies.

## Verification

- `vitest run` on both touched suites: 15 pass.
- `tsc --noEmit` error count identical to the master baseline (16
pre-existing, all in `server/`, none in changed files).
- The live failure was reproduced on macOS before the fix (twice, clean
field) and the mechanism confirmed in source: `install()` writes the
definition and bootstraps launchd only; `install-store` was previously
reachable solely from the `install`/`update` commands.

## Risks

- Low: the new path runs only when the user opts into the service and
the shim is absent. The managed install resolves the pinned onboarding
version from the public registry; on failure the flow declines exactly
as it does on unsupported platforms. `--yes` quickstarts, Docker, and
managed installs are untouched.

## Model Used

Claude Fable 5 (Claude Code)

## Pre-submission 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
This commit is contained in:
Devin Foley authored and GitHub committed 2026-08-24 23:30:54 -07:00
1 parent fa40a1b8d5
commit faad235aa2
5 files changed
+363 -17

No files matched your search

+57
View File
@@ -1,4 +1,5 @@
import fs from "node:fs/promises";
import { constants as fsConstants } from "node:fs";
import os from "node:os";
import path from "node:path";
import { execFile } from "node:child_process";
@@ -33,6 +34,7 @@ export interface ServiceManager {
restart(): Promise<void>;
status(): Promise<ServiceStatus>;
logs(follow: boolean, lines: number): Promise<void>;
installedExecutablePath(): Promise<string | null>;
enableLinger?(): Promise<void>;
}
@@ -75,6 +77,45 @@ export function resolveServiceShimPath(homeDir = os.homedir()): string {
return process.env.PAPERCLIP_SHIM_PATH?.trim() || path.join(homeDir, ".local", "bin", "paperclipai");
}
// The installed definition, not the current environment, is the truth
// about what the service executes: PAPERCLIP_SHIM_PATH may have changed
// or been unset since the definition was written.
function unescapeSystemd(value: string): string {
return value.replace(/\\\\|\\"|\$\$|%%/g, (m) =>
m === "\\\\" ? "\\" : m === '\\"' ? '"' : m === "$$" ? "$" : "%",
);
}
function unescapeXml(value: string): string {
return value.replace(/&(amp|lt|gt|quot|apos);/g, (_, name: string) =>
name === "amp" ? "&" : name === "lt" ? "<" : name === "gt" ? ">" : name === "quot" ? '"' : "'",
);
}
export function extractExecutableFromSystemdUnit(content: string): string | null {
const match = content.match(/^ExecStart="((?:\\.|[^"\\])*)"/m);
return match ? unescapeSystemd(match[1]) : null;
}
export function extractExecutableFromLaunchdPlist(content: string): string | null {
const match = content.match(/<key>ProgramArguments<\/key>\s*<array>\s*<string>([^<]+)<\/string>/);
return match ? unescapeXml(match[1]) : null;
}
// The service definition executes this path directly: existence is not
// enough — a directory or a non-executable file would satisfy fs.access's
// default mode and still crash the supervisor at spawn.
export async function isExecutableFile(filePath: string): Promise<boolean> {
try {
const stats = await fs.stat(filePath);
if (!stats.isFile()) return false;
await fs.access(filePath, fsConstants.X_OK);
return true;
} catch {
return false;
}
}
export function systemdServiceName(instanceId: string): string {
return instanceId === "default" ? "paperclipai.service" : `paperclipai-${instanceId}.service`;
}
@@ -174,6 +215,14 @@ export class SystemdServiceManager implements ServiceManager {
return renderSystemdUnit({ instanceId: this.instanceId, shimPath: this.shimPath, homeDir: this.homeDir });
}
async installedExecutablePath(): Promise<string | null> {
try {
return extractExecutableFromSystemdUnit(await fs.readFile(this.definitionPath, "utf8"));
} catch {
return null;
}
}
private async ensureCurrent(): Promise<boolean> {
const changed = await writeIfChanged(this.definitionPath, this.renderDefinition());
if (changed) await this.runner("systemctl", ["--user", "daemon-reload"]);
@@ -242,6 +291,14 @@ export class LaunchdServiceManager implements ServiceManager {
renderDefinition(): string { return renderLaunchdPlist({ instanceId: this.instanceId, shimPath: this.shimPath, homeDir: this.homeDir, stdoutPath: this.stdoutPath, stderrPath: this.stderrPath }); }
async installedExecutablePath(): Promise<string | null> {
try {
return extractExecutableFromLaunchdPlist(await fs.readFile(this.definitionPath, "utf8"));
} catch {
return null;
}
}
async install(options: ServiceInstallOptions): Promise<{ changed: boolean }> {
await fs.mkdir(path.dirname(this.stdoutPath), { recursive: true });
const changed = await writeIfChanged(this.definitionPath, this.renderDefinition());