Merge pull request #12 from michaelgbcole/updates

Add ditto unpack CLI to turn clone result JSON into a project tree

CHANGELOG.md

@@ -6,6 +6,38 @@ so minor/patch semantics are not yet guaranteed.
 
 ## [Unreleased]
 
+### Changed — Local CLI success output and orientation
+
+- The `clone-static` CLI now prints a copy-paste-safe success summary to stderr: a
+  single **quoted** `cd "<app>" && npm install && npm run dev` line (survives terminal
+  wrapping — no more broken `cd` from a wrapped timestamped path) plus the key
+  `AGENTS.md` safe-to-edit pointers (`src/app/content.ts`, `src/app/components/`). The
+  machine-readable `{ "event": "done", ... }` JSON line on stdout is unchanged (now
+  also carrying `stableApp`).
+- Added `--serve` (run `npm install` + `npm run dev` in the generated app) and
+  `--open` (also launch the browser at the dev URL) so "see it locally" is one flag.
+- In the default runs layout, each clone now refreshes a `runs/<site>/latest` symlink
+  pointing at the newest run, giving a stable, timestamp-free path for `cd` and scripts.
+
+### Added — Docs hub and terminology/secret-hygiene clarifications
+
+- Added [`docs/README.md`](docs/README.md) as a central documentation index.
+- Clarified prominently that ditto.site "cloning" means generating a codebase from a
+  live URL, **not** `git clone` (no source repo required), in the README and docs hub.
+- Added an explicit "API keys are secrets — use `$DITTO_API_KEY`, never inline/commit,
+  rotate anytime" note beside the key/auth examples in the README and `docs/SERVICE.md`.
+
+### Added — `ditto` unpack CLI
+
+- **`packages/cli`** — a zero-dependency `ditto` command-line helper. `npm run
+  unpack -- <clone.json|-> <out-dir>` turns the `files` map returned by
+  `POST /v1/clones` (or `GET /v1/clones/:id/result`) into a real project tree on
+  disk: text files written inline, binary assets materialized from inline base64
+  or fetched from their reference URL (`$DITTO_API_URL` / `$DITTO_API_KEY`), with
+  path-traversal guards and pre-write `sha256` integrity checks. Reads from stdin
+  so a `curl` response can be piped straight in. Documented next to the REST
+  examples in the README and `docs/SERVICE.md`.
+
 ### Added — Open-source readiness
 
 - Added support, release, responsible-use, CODEOWNERS, and gitattributes files.

README.md

@@ -15,8 +15,14 @@ Router project by default, or Vite React when requested.
 The compiler is not an LLM page author. It is a capture-to-code pipeline: same
 frozen capture in, byte-stable app out.
 
+> **"Cloning" here means generating a codebase from a live URL — not `git clone`.**
+> You don't need an existing repository, and you don't need the site's source. Point
+> ditto.site at a public URL and it writes you a fresh project from what the page
+> renders in a browser.
+
 Read the public development and evaluation method in
-[docs/METHODOLOGY.md](docs/METHODOLOGY.md).
+[docs/METHODOLOGY.md](docs/METHODOLOGY.md). For a map of all the docs, see
+[docs/README.md](docs/README.md).
 
 ## Usage
 
@@ -35,6 +41,11 @@ curl -sS -X POST "https://api.ditto.site/v1/signup/request" \
 The emailed verification link lands on `/api-key?token=...`, which calls
 `POST /v1/signup/verify` and displays the new `dtto_live_...` key once.
 
+> **Keys are secrets.** Put your key in an environment variable (`export
+> DITTO_API_KEY=dtto_live_...`) and reference `$DITTO_API_KEY` in every command —
+> never paste the raw key inline, where it lands in shell history, logs, or a chat.
+> Don't commit it. You can rotate a key anytime from the dashboard.
+
 ### REST API
 
 Start a clone job:
@@ -56,20 +67,52 @@ curl -sS -X POST "$DITTO_API_URL/v1/clones" \
   }'
 ```
 
-The service returns either an inline result or a queued job:
+The service returns either a queued job or an inline result. A finished result
+is a file map — every generated file keyed by its app-relative path:
 
 ```json
-{ "jobId": "job_123", "status": "queued" }
+{
+  "jobId": "job_123",
+  "status": "succeeded",
+  "files": {
+    "package.json": { "type": "text", "content": "{ ... }", "bytes": 812, "sha256": "..." },
+    "src/app/page.tsx": { "type": "text", "content": "export default ...", "bytes": 2048, "sha256": "..." },
+    "public/assets/logo.png": { "type": "binary", "url": ".../files/public/assets/logo.png", "bytes": 5123, "sha256": "..." }
+  }
+}
 ```
 
-Poll and download the generated app:
+**Turn that JSON into a project on disk** with the official unpacker — from a
+checked-out `ditto.site` repo with dependencies installed, pipe the response
+straight in with no temp file:
+
+```bash
+curl -sS -X POST "$DITTO_API_URL/v1/clones" \
+  -H "authorization: Bearer $DITTO_API_KEY" \
+  -H "content-type: application/json" \
+  -d '{"url":"https://example.com/","options":{"mode":"single"}}' \
+  | npm run --silent unpack -- - ./out
+```
+
+`npm run unpack -- <clone.json|-> <out-dir>` writes the text files inline and
+materializes binary assets (inline base64, or fetched from their `url` using
+`$DITTO_API_URL` / `$DITTO_API_KEY`). The CLI package is repo-local until the
+npm distribution story is ready, so do not use `npx ditto` yet. See
+[`packages/cli`](packages/cli/README.md) for options.
+
+If you got back a queued job (`{ "jobId": "job_123", "status": "queued" }`),
+poll it, then unpack the finished result — or download the whole app as one
+archive:
 
 ```bash
 JOB_ID="job_123"
 
+# poll status, then unpack the finished file map
 curl -sS -H "authorization: Bearer $DITTO_API_KEY" \
-  "$DITTO_API_URL/v1/clones/$JOB_ID"
+  "$DITTO_API_URL/v1/clones/$JOB_ID/result" \
+  | npm run --silent unpack -- - ./out
 
+# ...or grab the whole app as a single archive
 curl -L -H "authorization: Bearer $DITTO_API_KEY" \
   "$DITTO_API_URL/v1/clones/$JOB_ID/bundle?format=tgz" \
   -o ditto-clone.tgz
@@ -139,6 +182,8 @@ src/app/page.tsx, and src/app/ditto.css.
 ### Local CLI
 
 ```bash
+# this git clone gets the ditto.site tool itself — the URL you clone into a
+# codebase comes later, as the argument to `npm run clone`.
 git clone https://github.com/ion-design/ditto.site.git
 cd ditto.site
 
@@ -148,7 +193,17 @@ npx playwright install chromium
 npm run clone -- https://example.com/ --out=./output
 ```
 
-The generated app lands under `output/<site>/app`.
+The generated app lands under `output/<site>/app`. On success the CLI prints a
+copy-paste-safe summary — a single quoted `cd … && npm install && npm run dev`
+line and pointers to the safe-to-edit files (`src/app/content.ts`,
+`src/app/components/`; the app's `AGENTS.md` has the full guide).
+
+To skip the copy-paste entirely and go straight to a running preview:
+
+```bash
+npm run clone -- https://example.com/ --serve        # clone, npm install, npm run dev
+npm run clone -- https://example.com/ --open         # ...and open the browser too
+```
 
 Common local variants:
 
@@ -159,6 +214,10 @@ npm run clone -- https://example.com/ --framework=vite
 npm run validate-site -- runs/site-example.com/<timestamp>
 ```
 
+Without `--out`, runs land under `runs/<site>/<timestamp>/` and a stable
+`runs/<site>/latest` symlink always points at the newest clone, so scripts and
+`cd` targets don't depend on the timestamp.
+
 ### Local REST And MCP Service
 
 Quick inline mode, with no database:
@@ -241,6 +300,7 @@ minting is intentional.
 | --- | --- |
 | `compiler/` | deterministic capture, inference, generation, and validation |
 | `packages/core/` | compiler adapter and file-map helpers |
+| `packages/cli/` | `ditto` CLI — unpack a clone result JSON into a project tree |
 | `packages/api/` | Hono REST API and MCP server |
 | `packages/db/` | Drizzle schema, migrations, repository, and queue wrapper |
 | `packages/storage/` | local and S3/R2 artifact storage |

compiler/README.md

@@ -16,6 +16,8 @@ npm install
 npx playwright install chromium
 
 npm run clone -- https://example.com/
+npm run clone -- https://example.com/ --serve   # then npm install + npm run dev
+npm run clone -- https://example.com/ --open    # ...and open the browser too
 npm run clone -- https://example.com/ --mode=multi --styling=tailwind
 npm run clone -- https://example.com/ --mode=single --framework=vite
 npm run clone-site -- https://example.com/
@@ -34,6 +36,13 @@ npm run typecheck
 Root-level scripts forward to these commands, so `npm run clone -- <url>` works
 from the repository root too.
 
+("Clone" here = generating a codebase from a live URL, not `git clone`; no source
+repo required.) On success the CLI prints a copy-paste-safe summary: a single
+quoted `cd … && npm install && npm run dev` line plus safe-to-edit pointers. Pass
+`--serve` to run install + dev automatically, or `--open` to also launch the
+browser. Without `--out`, a `runs/<site>/latest` symlink always points at the
+newest run so paths aren't timestamp-fragile.
+
 Multi-page generation defaults to the fast no-validation path. Use `--validate`
 when the clone command itself should run the full build/render/gates QA pass, or
 run `validate-site` separately. `--concurrency` controls source route capture;

compiler/src/cli.ts

@@ -1,11 +1,12 @@
 #!/usr/bin/env -S npx tsx
 import { basename, dirname, join, resolve, sep } from "node:path";
-import { cpSync, existsSync, readdirSync, readFileSync, rmSync, statSync, writeFileSync } from "node:fs";
+import { cpSync, existsSync, readdirSync, readFileSync, rmSync, statSync, symlinkSync, writeFileSync } from "node:fs";
 import { captureSite, REQUIRED_VIEWPORTS, SAMPLE_VIEWPORTS, type CaptureResult } from "./capture/capture.js";
 import { fileURLToPath } from "node:url";
 import { generateAll } from "./generate/pipeline.js";
 import { refineSizing } from "./generate/refineSizing.js";
 import { writeJSON, writeText, ensureDir, readJSON, fileExists } from "./util/fsx.js";
+import { doneSummary, serveApp } from "./cliSummary.js";
 import type { AppFramework } from "./generate/app.js";
 
 export type CloneOptions = {
@@ -34,6 +35,9 @@ export type CloneResult = {
   sourceDir: string;
   appDir: string;
   sourceUrl: string;
+  /** A stable, timestamp-free path to the app (via the `runs/<site>/latest` symlink),
+   *  present in runs-layout mode when the symlink could be created. */
+  stableAppDir?: string;
 };
 
 export function siteIdFromUrl(url: string): string {
@@ -462,16 +466,34 @@ export async function runClone(opts: CloneOptions): Promise<CloneResult> {
 
   writeText(join(runDir, "logs", "compiler.log.jsonl"), logEvents.map((e) => JSON.stringify(e)).join("\n") + "\n");
 
+  let stableAppDir: string | undefined;
   if (out) {
     // Publish the deliverable to <siteName>/app; keep working artifacts in .clone.
     exportApp(appDir, out.appDir);
     logBoth({ event: "exported", app: out.appDir });
   } else {
-    // latest pointer (runs/ layout, used by --reuse / regen)
+    // latest pointer (runs/ layout, used by --reuse / regen) + a `latest` symlink so the
-    writeJSON(join(runsDir, siteId, "latest.json"), { runDir, ts: timestamp() });
+    // app has a stable, timestamp-free path that survives copy-paste.
+    stableAppDir = writeLatestPointer(runsDir, siteId, runDir);
   }
 
-  return { runDir, sourceDir, appDir: out ? out.appDir : appDir, sourceUrl: opts.url };
+  return { runDir, sourceDir, appDir: out ? out.appDir : appDir, sourceUrl: opts.url, stableAppDir };
+}
+
+/** Record the newest run for a site in the runs layout: a `latest.json` breadcrumb (used by
+ *  `--reuse`/regen) and a `latest` symlink → runDir. Returns the stable app path when the
+ *  symlink was created (symlinks can be unavailable, e.g. Windows without privilege — non-fatal). */
+export function writeLatestPointer(runsDir: string, siteId: string, runDir: string): string | undefined {
+  writeJSON(join(runsDir, siteId, "latest.json"), { runDir, ts: timestamp() });
+  const link = join(runsDir, siteId, "latest");
+  try {
+    // Refresh an existing symlink; refuse (via non-recursive rm) to clobber a real directory.
+    rmSync(link, { force: true });
+    symlinkSync(runDir, link, "junction");
+    return join(link, "generated", "app");
+  } catch {
+    return undefined;
+  }
 }
 
 function copySourceRef(from: string, to: string): void {
@@ -543,12 +565,22 @@ async function main(): Promise<void> {
   const args = process.argv.slice(2);
   const url = args.find((a) => !a.startsWith("--"));
   if (!url) {
-    console.error("usage: clone-static <url> [--mode=single|multi] [--styling=tailwind|css] [--framework=next|vite] [--out=<dir>]");
+    console.error("usage: clone-static <url> [--mode=single|multi] [--styling=tailwind|css] [--framework=next|vite] [--out=<dir>] [--serve] [--open]");
     process.exit(1);
   }
   const mode = parseProductMode(args);
   const styling = parseProductStyling(args);
   const framework = parseProductFramework(args);
+  // --serve installs deps + starts the dev server after cloning; --open also launches the browser.
+  const open = hasAnyFlag(args, ["--open"]);
+  const serve = open || hasAnyFlag(args, ["--serve"]);
+  const finish = async (res: { appDir: string; stableAppDir?: string }) => {
+    if (serve) {
+      await serveApp(res.appDir, { open });
+    } else {
+      process.stderr.write(doneSummary({ url, appDir: res.appDir, framework, stableAppDir: res.stableAppDir }));
+    }
+  };
   const vpArg = firstFlagValue(args, ["--dev-viewports", "--viewports"]);
   const runsArg = firstFlagValue(args, ["--dev-runs", "--runs"]);
   // --out=<dir>: clean <dir>/<siteName>/{app,.clone} layout (default: ./output when bare --out).
@@ -604,7 +636,8 @@ async function main(): Promise<void> {
       tier,
       log: (e) => console.log(JSON.stringify(e)),
     });
-    console.log(JSON.stringify({ event: "done", runDir: res.runDir, app: res.appDir }));
+    console.log(JSON.stringify({ event: "done", runDir: res.runDir, app: res.appDir, stableApp: res.stableAppDir }));
+    await finish(res);
     return;
   }
 
@@ -625,7 +658,8 @@ async function main(): Promise<void> {
     reuseSource,
     log: (e) => console.log(JSON.stringify(e)),
   });
-  console.log(JSON.stringify({ event: "done", runDir: res.runDir, app: res.appDir }));
+  console.log(JSON.stringify({ event: "done", runDir: res.runDir, app: res.appDir, stableApp: res.stableAppDir }));
+  await finish(res);
 }
 
 if (import.meta.url === `file://${process.argv[1]}`) {

compiler/src/cliSummary.ts

@@ -0,0 +1,112 @@
+/**
+ * Human-facing success output for the `clone-static` CLI.
+ *
+ * The CLI streams machine-readable JSON events to stdout; this module renders the
+ * copy-paste-friendly summary a person actually reads at the end, plus the optional
+ * `--serve` runner that installs deps and starts the dev server for them.
+ *
+ * Why this exists: a bare `runs/<domain>/<timestamp>/generated/app` path pasted into a
+ * terminal wraps mid-word, so `cd` fails and the follow-up `npm run dev` runs in the
+ * wrong directory. The preview command here is a single quoted line that survives
+ * wrapping.
+ */
+import { spawn } from "node:child_process";
+import { platform } from "node:os";
+
+export type DoneSummaryInput = {
+  url: string;
+  /** The exact generated app directory for this run. */
+  appDir: string;
+  framework: "next" | "vite";
+  /** Stable path (a `runs/<site>/latest` symlink target) preferred as the `cd` target when present. */
+  stableAppDir?: string;
+};
+
+/** Double-quote a path for a POSIX shell so spaces and wrapping don't break copy-paste. */
+export function shellQuote(p: string): string {
+  return `"${p.replace(/(["\\$`])/g, "\\$1")}"`;
+}
+
+/** The success block. The preview command is one quoted line on purpose. */
+export function doneSummary(input: DoneSummaryInput): string {
+  const root = input.framework === "next" ? "src/app" : "src";
+  const cdTarget = input.stableAppDir ?? input.appDir;
+  const lines: string[] = [
+    "",
+    `✓ Done — cloned ${input.url}`,
+    "",
+    "  Preview it locally (copy-paste this one line):",
+    "",
+    `    cd ${shellQuote(cdTarget)} && npm install && npm run dev`,
+    "",
+    "  What's safe to edit — full guide in AGENTS.md inside the app:",
+    `    • page copy & content  →  ${root}/content.ts`,
+    `    • components           →  ${root}/components/`,
+    "",
+    "  Or re-run with --serve to install deps and start the dev server for you",
+    "  (add --open to launch the browser too).",
+  ];
+  if (input.stableAppDir && input.stableAppDir !== input.appDir) {
+    lines.push("");
+    lines.push(`  The path above is a stable pointer to the newest clone. This exact run:`);
+    lines.push(`    ${input.appDir}`);
+  }
+  return lines.join("\n") + "\n";
+}
+
+function openBrowser(url: string): void {
+  const cmd = platform() === "darwin" ? "open" : platform() === "win32" ? "cmd" : "xdg-open";
+  const args = platform() === "win32" ? ["/c", "start", "", url] : [url];
+  try {
+    const child = spawn(cmd, args, { stdio: "ignore", detached: true });
+    child.on("error", () => {}); // opening a browser is best-effort
+    child.unref();
+  } catch {
+    /* best-effort */
+  }
+}
+
+function runToCompletion(cmd: string, args: string[], cwd: string): Promise<void> {
+  return new Promise((resolvePromise, reject) => {
+    const child = spawn(cmd, args, { cwd, stdio: "inherit", shell: platform() === "win32" });
+    child.on("error", reject);
+    child.on("exit", (code) =>
+      code === 0 ? resolvePromise() : reject(new Error(`\`${cmd} ${args.join(" ")}\` exited with code ${code}`)),
+    );
+  });
+}
+
+/**
+ * Run `npm install` then `npm run dev` in the generated app (foreground — `npm run dev`
+ * keeps running until the user stops it). With `open`, launches the default browser at
+ * the dev URL once the server prints it.
+ */
+export async function serveApp(appDir: string, opts: { open: boolean }): Promise<void> {
+  process.stderr.write(`\n→ Installing dependencies in ${appDir}\n`);
+  await runToCompletion("npm", ["install"], appDir);
+
+  process.stderr.write(`→ Starting the dev server (Ctrl-C to stop)\n\n`);
+  await new Promise<void>((resolvePromise, reject) => {
+    const dev = spawn("npm", ["run", "dev"], {
+      cwd: appDir,
+      stdio: opts.open ? ["inherit", "pipe", "inherit"] : "inherit",
+      shell: platform() === "win32",
+    });
+    if (opts.open && dev.stdout) {
+      let opened = false;
+      dev.stdout.on("data", (chunk: Buffer) => {
+        process.stdout.write(chunk);
+        if (opened) return;
+        const m = String(chunk).match(/https?:\/\/(?:localhost|127\.0\.0\.1)(?::\d+)?\S*/i);
+        if (m) {
+          opened = true;
+          openBrowser(m[0]);
+        }
+      });
+    }
+    dev.on("error", reject);
+    dev.on("exit", (code) =>
+      code === 0 || code === null ? resolvePromise() : reject(new Error(`\`npm run dev\` exited with code ${code}`)),
+    );
+  });
+}

compiler/src/site/cloneSite.ts

@@ -21,7 +21,7 @@ import { structurallySimilar } from "./signature.js";
 import { generateSiteApp, routeToSegment, routeKey, type RouteArtifact } from "./generateSite.js";
 import { detectSharedChrome, chromeSignatureId } from "./sharedLayout.js";
 import { validateSite, type SiteReport } from "./validateSite.js";
-import { siteIdFromUrl, namedOutDirs, exportApp } from "../cli.js";
+import { siteIdFromUrl, namedOutDirs, exportApp, writeLatestPointer } from "../cli.js";
 import { writeJSON, readJSON, ensureDir, fileExists, writeText } from "../util/fsx.js";
 import { seoInventoryToMarkdown } from "../generate/seo.js";
 import type { AppFramework } from "../generate/app.js";
@@ -56,6 +56,8 @@ export type CloneSiteResult = {
   plan: RoutePlan;
   routes: RouteArtifact[];
   siteReport?: SiteReport;
+  /** Stable, timestamp-free path to the app (via the `runs/<site>/latest` symlink), when created. */
+  stableAppDir?: string;
 };
 
 function timestamp(): string {
@@ -236,8 +238,10 @@ export async function runCloneSite(opts: CloneSiteOptions): Promise<CloneSiteRes
     siteReport = await validateSite(runDir, { tier: opts.tier ?? "stage2", routeConcurrency: opts.validationConcurrency, viewportConcurrency: opts.viewportConcurrency, log });
   }
 
+  let stableAppDir: string | undefined;
   if (out) { exportApp(appDir, out.appDir); log({ event: "exported", app: out.appDir }); }
-  return { runDir, appDir: out ? out.appDir : appDir, siteId, plan, routes, siteReport };
+  else { stableAppDir = writeLatestPointer(runsDir, siteId, runDir); }
+  return { runDir, appDir: out ? out.appDir : appDir, siteId, plan, routes, siteReport, stableAppDir };
 }
 
 type ManifestForRegen = {

compiler/test/cliSummary.test.ts

@@ -0,0 +1,53 @@
+import { describe, it } from "node:test";
+import assert from "node:assert/strict";
+import { doneSummary, shellQuote } from "../src/cliSummary.js";
+
+describe("shellQuote", () => {
+  it("wraps a path in double quotes so spaces survive copy-paste", () => {
+    assert.equal(shellQuote("/Users/x/runs/site/20260701/generated/app"), '"/Users/x/runs/site/20260701/generated/app"');
+    assert.equal(shellQuote("/Users/x/My Sites/app"), '"/Users/x/My Sites/app"');
+  });
+
+  it("escapes characters the shell would otherwise interpret inside double quotes", () => {
+    assert.equal(shellQuote('/a/"b"/$c/`d`/e\\f'), '"/a/\\"b\\"/\\$c/\\`d\\`/e\\\\f"');
+  });
+});
+
+describe("doneSummary", () => {
+  const base = { url: "https://academyux.com/", appDir: "/runs/academyux.com/20260701-173012/generated/app", framework: "next" as const };
+
+  it("emits a single quoted preview line (no mid-word wrap on the path)", () => {
+    const out = doneSummary(base);
+    const line = out.split("\n").find((l) => l.includes("cd "));
+    assert.ok(line, "has a cd line");
+    // one line, path is quoted so a wrapping terminal can't split the command
+    assert.match(line!, /cd "\/runs\/academyux\.com\/20260701-173012\/generated\/app" && npm install && npm run dev/);
+  });
+
+  it("points at the Next safe-edit areas from AGENTS.md", () => {
+    const out = doneSummary(base);
+    assert.match(out, /src\/app\/content\.ts/);
+    assert.match(out, /src\/app\/components\//);
+    assert.match(out, /AGENTS\.md/);
+  });
+
+  it("uses the Vite src root when framework is vite", () => {
+    const out = doneSummary({ ...base, framework: "vite" });
+    assert.match(out, /cd .* && npm install && npm run dev/);
+    assert.match(out, /• page copy & content {2}→ {2}src\/content\.ts/);
+    assert.match(out, /• components {11}→ {2}src\/components\//);
+  });
+
+  it("prefers the stable path as the cd target and still shows the exact run", () => {
+    const out = doneSummary({ ...base, stableAppDir: "/runs/academyux.com/latest/generated/app" });
+    const cdLine = out.split("\n").find((l) => l.includes("cd "))!;
+    assert.match(cdLine, /cd "\/runs\/academyux\.com\/latest\/generated\/app"/);
+    assert.match(out, /This exact run/);
+    assert.match(out, /\/runs\/academyux\.com\/20260701-173012\/generated\/app/);
+  });
+
+  it("mentions the --serve / --open shortcut", () => {
+    assert.match(doneSummary(base), /--serve/);
+    assert.match(doneSummary(base), /--open/);
+  });
+});

compiler/test/runsLayout.test.ts

@@ -0,0 +1,31 @@
+import { describe, it } from "node:test";
+import assert from "node:assert/strict";
+import { mkdtempSync, mkdirSync, existsSync, readlinkSync, realpathSync } from "node:fs";
+import { tmpdir } from "node:os";
+import { join } from "node:path";
+import { writeLatestPointer } from "../src/cli.js";
+
+describe("writeLatestPointer", () => {
+  it("writes latest.json and a `latest` symlink to the newest run, refreshing it in place", () => {
+    const runs = mkdtempSync(join(tmpdir(), "ditto-runs-"));
+    const siteId = "example.com";
+
+    const run1 = join(runs, siteId, "20260701-100000");
+    mkdirSync(join(run1, "generated", "app"), { recursive: true });
+    const stable1 = writeLatestPointer(runs, siteId, run1);
+
+    assert.ok(existsSync(join(runs, siteId, "latest.json")), "writes latest.json breadcrumb");
+    assert.ok(stable1 && existsSync(stable1), "returned stable app path resolves through the symlink");
+    assert.equal(realpathSync(readlinkSync(join(runs, siteId, "latest"))), realpathSync(run1));
+
+    // A second run must re-point the existing symlink, not throw on the collision.
+    const run2 = join(runs, siteId, "20260701-200000");
+    mkdirSync(join(run2, "generated", "app"), { recursive: true });
+    const stable2 = writeLatestPointer(runs, siteId, run2);
+
+    assert.equal(realpathSync(readlinkSync(join(runs, siteId, "latest"))), realpathSync(run2));
+    assert.ok(stable2 && existsSync(stable2));
+    // The stable path is timestamp-free (goes through `latest`, not the run's timestamp).
+    assert.match(stable2!, /example\.com\/latest\/generated\/app$/);
+  });
+});

docs/README.md

@@ -0,0 +1,57 @@
+# ditto.site Documentation
+
+The central index for ditto.site's docs. Start here.
+
+> **First, the word "clone."** In ditto.site, *cloning* means **generating a
+> codebase from a live URL** — it is **not** `git clone`. You do **not** need an
+> existing Git repository, and you do **not** need the target site's source code.
+> You give ditto.site a public URL; it captures what the page renders in a browser
+> and writes you a fresh, runnable project. (The only `git clone` involved is
+> optionally cloning *this tool's* repository to run the compiler locally.)
+
+## Get started
+
+| I want to… | Go to |
+| --- | --- |
+| Understand what ditto.site is and see it run | [Project README](../README.md) |
+| Call the hosted **REST API** or **MCP** server | [Project README → Usage](../README.md#usage), [SERVICE.md](SERVICE.md) |
+| Turn a clone result JSON into files on disk | [Repo-local unpack CLI](../packages/cli/README.md) |
+| Run the **compiler locally** from the command line | [compiler/README.md](../compiler/README.md) |
+| Deploy the service | [DEPLOY.md](DEPLOY.md) |
+| Read the development & evaluation method | [METHODOLOGY.md](METHODOLOGY.md) |
+| Understand responsible-use boundaries | [RESPONSIBLE_USE.md](RESPONSIBLE_USE.md) |
+| Cut a release | [RELEASING.md](RELEASING.md) |
+
+## API keys are secrets
+
+Keys look like `dtto_live_...`. Keep them in an environment variable and reference
+it in commands:
+
+```bash
+export DITTO_API_KEY="dtto_live_..."
+curl -sS -H "authorization: Bearer $DITTO_API_KEY" "$DITTO_API_URL/v1/clones"
+```
+
+Never paste a raw key inline (it leaks into shell history, logs, and chat), and
+never commit one. Rotate a leaked key anytime from the dashboard. See
+[SERVICE.md](SERVICE.md) for the full auth model.
+
+## The short version of the workflow
+
+1. **Clone** a URL → `POST /v1/clones` (API) or `npm run clone -- <url>` (local CLI).
+2. **Get the app** → unpack the result JSON with the [repo-local unpack CLI](../packages/cli/README.md),
+   download the `bundle?format=tgz` archive, or read files from `runs/<site>/latest/`.
+3. **Preview it** → `cd` into the app and `npm install && npm run dev` (or let the
+   local CLI do it for you with `--serve` / `--open`).
+4. **Edit safely** → each generated app ships an `AGENTS.md` describing what's safe
+   to change (copy in `src/app/content.ts`, components in `src/app/components/`, etc.).
+
+## Full doc list
+
+- [SERVICE.md](SERVICE.md) — REST + MCP service reference (endpoints, options, env vars).
+- [DEPLOY.md](DEPLOY.md) — production deployment.
+- [METHODOLOGY.md](METHODOLOGY.md) — how the compiler is developed and evaluated.
+- [RESPONSIBLE_USE.md](RESPONSIBLE_USE.md) — acceptable-use boundaries.
+- [RELEASING.md](RELEASING.md) — release process.
+- [../compiler/README.md](../compiler/README.md) — local compiler commands.
+- [../packages/cli/README.md](../packages/cli/README.md) — the repo-local unpack CLI.

docs/SERVICE.md

@@ -60,6 +60,21 @@ curl -s -X POST localhost:8787/v1/clones -H 'content-type: application/json' \
   -d '{"url":"https://example.com/","options":{"mode":"single","styling":"tailwind"}}' | jq '.files | keys'
 ```
 
+To turn that `files` map into a project on disk, pipe the response into the
+repo-local `ditto` CLI (`packages/cli`) instead of inspecting the JSON by hand:
+
+```bash
+curl -s -X POST localhost:8787/v1/clones -H 'content-type: application/json' \
+  -d '{"url":"https://example.com/","options":{"mode":"single"}}' \
+  | npm run --silent unpack -- - ./out
+# binary assets: set DITTO_API_URL (and DITTO_API_KEY when authenticated) so the
+# CLI can fetch each file's reference URL; --no-fetch writes only the text tree.
+```
+
+The CLI package is intentionally private for now; run this command from a
+checked-out `ditto.site` repo with dependencies installed. Do not use
+`npx ditto` until this package is published.
+
 ## REST surface
 
 ```
@@ -78,6 +93,11 @@ GET    /healthz                    → { ok: true }  (unauthenticated)
 
 `/v1/clones*` and `/mcp` are authenticated when `API_KEYS` is set or DB-backed
 keys exist. Use `Authorization: Bearer <key>` or `x-api-key: <key>`.
+
+> **Treat keys as secrets.** In any snippet you copy or share, template the key as
+> an environment variable (`Authorization: Bearer $DITTO_API_KEY`) rather than an
+> inline `dtto_live_...` token — inline keys leak into shell history, logs, and
+> pasted transcripts. Never commit a key; rotate a leaked one from the dashboard.
 Signup routes are intentionally public only when `SIGNUP_ENABLED=true` **and**
 `DATABASE_URL` is set. Direct `POST /v1/signup` mints a `dtto_live_...` key
 immediately when `SIGNUP_DIRECT_ENABLED=true`. For public production signup,

package-lock.json

@@ -1878,6 +1878,10 @@
         "node": ">= 0.8"
       }
     },
+    "node_modules/ditto": {
+      "resolved": "packages/cli",
+      "link": true
+    },
     "node_modules/drizzle-kit": {
       "version": "0.31.10",
       "resolved": "https://registry.npmjs.org/drizzle-kit/-/drizzle-kit-0.31.10.tgz",
@@ -3802,6 +3806,17 @@
         "typescript": "5.7.3"
       }
     },
+    "packages/cli": {
+      "name": "ditto",
+      "version": "0.1.0",
+      "license": "MIT",
+      "bin": {
+        "ditto": "bin/ditto.mjs"
+      },
+      "engines": {
+        "node": ">=20"
+      }
+    },
     "packages/core": {
       "name": "@cloner/core",
       "version": "0.1.0",

package.json

@@ -41,7 +41,8 @@
     "dev:api": "npm run dev --workspace @cloner/api",
     "dev:worker": "npm run dev --workspace @cloner/worker",
     "db:generate": "npm run generate --workspace @cloner/db",
-    "db:migrate": "npm run migrate --workspace @cloner/db"
+    "db:migrate": "npm run migrate --workspace @cloner/db",
+    "unpack": "node packages/cli/bin/ditto.mjs unpack"
   },
   "devDependencies": {
     "tsx": "4.22.4",

packages/cli/README.md

@@ -0,0 +1,50 @@
+# ditto (CLI)
+
+The official ditto.site command-line helper. Turns a clone **result JSON** —
+the giant blob you get back from `POST /v1/clones` or
+`GET /v1/clones/:id/result` — into an actual project tree on disk.
+
+Zero dependencies. Needs Node >= 20.
+
+This workspace is currently private and repo-local. Run these commands from the
+repository root after `npm install`; do not use `npx ditto` until the package is
+published.
+
+## Unpack
+
+```bash
+# from a saved file
+npm run unpack -- clone.json ./out
+
+# straight from curl, no temp file
+curl -sS -X POST "$DITTO_API_URL/v1/clones" \
+  -H "authorization: Bearer $DITTO_API_KEY" \
+  -H "content-type: application/json" \
+  -d '{"url":"https://example.com/","options":{"mode":"single"}}' \
+  | npm run --silent unpack -- - ./out
+```
+
+`npm run unpack -- <clone.json|-> <out-dir>`:
+
+- writes every text file from its inline `content`,
+- materializes binary assets from inline base64 when present, otherwise fetches
+  them from their reference `url`,
+- refuses paths that escape `<out-dir>` and verifies `sha256` when the result
+  carries it.
+
+### Binary assets
+
+Clone results return binaries by reference (`{ "type": "binary", "url": ... }`)
+rather than inlining megabytes of base64. To fetch them, the unpacker needs to
+know where the API lives:
+
+| Source | Flag | Env |
+| --- | --- | --- |
+| Base URL for relative asset URLs | `--base-url <url>` | `DITTO_API_URL` |
+| Bearer key for authenticated APIs | `--api-key <key>` | `DITTO_API_KEY` |
+
+Use `--no-fetch` to write only the text tree and list the binaries as skipped.
+To grab everything in one shot instead, download the archive directly:
+`GET /v1/clones/<id>/bundle?format=tgz`.
+
+Run `npm run unpack -- --help` for the full option list.

packages/cli/bin/ditto.mjs

@@ -0,0 +1,253 @@
+#!/usr/bin/env node
+// ditto — official command-line helper for ditto.site.
+//
+// `ditto unpack <clone.json> <out-dir>` turns the JSON returned by
+// `POST /v1/clones` (or `GET /v1/clones/:id/result`) into a real project tree
+// on disk: text files are written from their inline `content`, and binary
+// assets are materialized from inline base64 or fetched from their reference
+// URL. Zero dependencies — just Node >= 20.
+
+import { createHash } from "node:crypto";
+import { mkdir, readFile, writeFile } from "node:fs/promises";
+import { dirname, join, relative, resolve, sep } from "node:path";
+
+const USAGE = `ditto — unpack a ditto.site clone result into a project tree
+
+Usage:
+  ditto unpack <clone.json|-> <out-dir> [options]
+
+Arguments:
+  <clone.json>   Path to the saved JSON (POST /v1/clones or GET .../result),
+                 or "-" to read the JSON from stdin (e.g. piped from curl).
+  <out-dir>      Directory to write the project into (created if missing).
+
+Options:
+  --base-url <url>   Base URL used to resolve relative binary asset URLs.
+                     Defaults to $DITTO_API_URL.
+  --api-key <key>    Bearer key sent when fetching binary assets.
+                     Defaults to $DITTO_API_KEY.
+  --no-fetch         Do not fetch binary assets over the network; report them
+                     as skipped instead. Text files are still written.
+  --quiet            Only print the final summary line.
+  -h, --help         Show this help.
+
+Examples:
+  npm run unpack -- clone.json ./out
+
+  curl -sS -X POST "$DITTO_API_URL/v1/clones" \\
+    -H "authorization: Bearer $DITTO_API_KEY" \\
+    -H "content-type: application/json" \\
+    -d '{"url":"https://example.com/","options":{"mode":"single"}}' \\
+    | npm run --silent unpack -- - ./out
+
+`;
+
+/** Print to stderr and exit non-zero. */
+function fail(msg) {
+  process.stderr.write(`ditto: ${msg}\n`);
+  process.exit(1);
+}
+
+/** Minimal flag parser: pulls known --flags out, returns { positionals, flags }. */
+function parseArgs(argv) {
+  const positionals = [];
+  const flags = { fetch: true, quiet: false };
+  for (let i = 0; i < argv.length; i++) {
+    const a = argv[i];
+    switch (a) {
+      case "-h":
+      case "--help":
+        flags.help = true;
+        break;
+      case "--no-fetch":
+        flags.fetch = false;
+        break;
+      case "--quiet":
+        flags.quiet = true;
+        break;
+      case "--base-url":
+        flags.baseUrl = argv[++i];
+        break;
+      case "--api-key":
+        flags.apiKey = argv[++i];
+        break;
+      default:
+        if (a.startsWith("--")) fail(`unknown option: ${a}`);
+        positionals.push(a);
+    }
+  }
+  return { positionals, flags };
+}
+
+/** Read the whole of stdin as a string. */
+async function readStdin() {
+  const chunks = [];
+  for await (const chunk of process.stdin) chunks.push(chunk);
+  return Buffer.concat(chunks).toString("utf8");
+}
+
+/** Accept either the full result envelope ({ files: {...} }) or a bare file map. */
+function extractFiles(doc) {
+  if (doc && typeof doc === "object" && doc.files && typeof doc.files === "object") {
+    return doc.files;
+  }
+  // A bare file map: values look like clone file entries.
+  if (doc && typeof doc === "object") {
+    const vals = Object.values(doc);
+    if (vals.length && vals.every((v) => v && typeof v === "object" && ("content" in v || "url" in v || "type" in v))) {
+      return doc;
+    }
+  }
+  return null;
+}
+
+/** Resolve `path` under `outDir`, refusing anything that escapes the tree
+ *  (absolute paths, `..`, leading slashes) — clone results are untrusted input. */
+function safeJoin(outDir, path) {
+  const cleaned = String(path).replace(/^[/\\]+/, "");
+  const dest = resolve(outDir, cleaned);
+  const rel = relative(outDir, dest);
+  if (rel === "" || rel.startsWith("..") || rel.split(sep).includes("..")) {
+    throw new Error(`refusing to write outside output dir: ${path}`);
+  }
+  return dest;
+}
+
+/** Decode a binary entry's inline bytes, if present (base64 or plain). */
+function inlineBytes(entry) {
+  if (typeof entry.base64 === "string") return Buffer.from(entry.base64, "base64");
+  if (typeof entry.content === "string") {
+    const enc = entry.encoding || (entry.type === "binary" ? "base64" : "utf8");
+    return Buffer.from(entry.content, enc === "base64" ? "base64" : "utf8");
+  }
+  return null;
+}
+
+function sha256(buf) {
+  return createHash("sha256").update(buf).digest("hex");
+}
+
+async function fetchBinary(url, baseUrl, apiKey) {
+  const abs = /^https?:\/\//i.test(url) ? url : baseUrl ? new URL(url, baseUrl).toString() : null;
+  if (!abs) throw new Error("relative asset URL but no --base-url / $DITTO_API_URL set");
+  const headers = apiKey ? { authorization: `Bearer ${apiKey}` } : {};
+  const res = await fetch(abs, { headers });
+  if (!res.ok) throw new Error(`GET ${abs} -> ${res.status} ${res.statusText}`);
+  return Buffer.from(await res.arrayBuffer());
+}
+
+async function unpack(positionals, flags) {
+  const [input, outDir] = positionals;
+  if (!input || !outDir) fail("unpack needs <clone.json|-> and <out-dir>\n\n" + USAGE);
+
+  const raw = input === "-" ? await readStdin() : await readFile(input, "utf8").catch((e) => fail(`cannot read ${input}: ${e.message}`));
+  let doc;
+  try {
+    doc = JSON.parse(raw);
+  } catch (e) {
+    fail(`input is not valid JSON: ${e.message}`);
+  }
+
+  if (doc && (doc.status === "queued" || doc.jobId) && !doc.files) {
+    fail(
+      `this looks like a queued job (${doc.jobId ? `jobId ${doc.jobId}` : doc.status}), not a finished result.\n` +
+        `Poll GET /v1/clones/<id>/result until it has a "files" map, then unpack that.`,
+    );
+  }
+
+  const files = extractFiles(doc);
+  if (!files) fail(`no "files" map found in the input — is this a clone result JSON?`);
+
+  const outAbs = resolve(outDir);
+  const baseUrl = flags.baseUrl || process.env.DITTO_API_URL || "";
+  const apiKey = flags.apiKey || process.env.DITTO_API_KEY || "";
+  const log = (m) => {
+    if (!flags.quiet) process.stderr.write(m + "\n");
+  };
+
+  let written = 0;
+  let bytes = 0;
+  const skipped = [];
+
+  for (const [path, entry] of Object.entries(files)) {
+    if (!entry || typeof entry !== "object") {
+      skipped.push({ path, reason: "malformed entry" });
+      continue;
+    }
+    const isBinary = entry.type === "binary" || entry.kind === "binary";
+    let buf;
+
+    if (isBinary) {
+      buf = inlineBytes(entry);
+      if (!buf) {
+        if (!flags.fetch) {
+          skipped.push({ path, reason: "binary asset (fetch disabled)" });
+          continue;
+        }
+        if (typeof entry.url !== "string") {
+          skipped.push({ path, reason: "binary asset with no inline bytes or url" });
+          continue;
+        }
+        try {
+          buf = await fetchBinary(entry.url, baseUrl, apiKey);
+        } catch (e) {
+          skipped.push({ path, reason: `could not fetch asset: ${e.message}` });
+          continue;
+        }
+      }
+    } else {
+      buf = Buffer.from(typeof entry.content === "string" ? entry.content : "", "utf8");
+    }
+
+    if (typeof entry.sha256 === "string" && entry.sha256 && sha256(buf) !== entry.sha256) {
+      log(`  ! ${path}: sha256 mismatch`);
+      fail(`${path} failed sha256 integrity check`);
+    }
+
+    const dest = safeJoin(outAbs, path);
+    await mkdir(dirname(dest), { recursive: true });
+    await writeFile(dest, buf);
+    written++;
+    bytes += buf.length;
+    log(`  + ${path}`);
+  }
+
+  const kb = (bytes / 1024).toFixed(1);
+  log("");
+  process.stdout.write(`Wrote ${written} file${written === 1 ? "" : "s"} (${kb} KB) to ${relative(process.cwd(), outAbs) || "."}\n`);
+
+  if (skipped.length) {
+    process.stderr.write(`\n${skipped.length} file${skipped.length === 1 ? "" : "s"} skipped:\n`);
+    for (const s of skipped) process.stderr.write(`  - ${s.path}: ${s.reason}\n`);
+    const anyAsset = skipped.some((s) => /asset|binary/.test(s.reason));
+    if (anyAsset) {
+      process.stderr.write(
+        `\nTip: binary assets are referenced by URL. Set $DITTO_API_URL (and $DITTO_API_KEY if\n` +
+          `the API is authenticated) so ditto can fetch them, or download the whole app in one\n` +
+          `shot with GET /v1/clones/<id>/bundle?format=tgz.\n`,
+      );
+    }
+  }
+}
+
+async function main() {
+  const argv = process.argv.slice(2);
+  const { positionals, flags } = parseArgs(argv);
+  const command = positionals.shift();
+
+  if (flags.help || !command) {
+    process.stdout.write(USAGE);
+    // No command at all is a misuse (exit 1); an explicit --help is success.
+    process.exit(flags.help ? 0 : 1);
+  }
+
+  switch (command) {
+    case "unpack":
+      await unpack(positionals, flags);
+      break;
+    default:
+      fail(`unknown command: ${command}\n\n${USAGE}`);
+  }
+}
+
+main().catch((e) => fail(e?.stack || String(e)));

packages/cli/package.json

@@ -0,0 +1,30 @@
+{
+  "name": "ditto",
+  "version": "0.1.0",
+  "private": true,
+  "description": "Official ditto.site command-line helper: unpack a clone JSON result into a project tree on disk.",
+  "license": "MIT",
+  "author": "ion-design",
+  "repository": {
+    "type": "git",
+    "url": "https://github.com/ion-design/ditto.site.git",
+    "directory": "packages/cli"
+  },
+  "bugs": {
+    "url": "https://github.com/ion-design/ditto.site/issues"
+  },
+  "homepage": "https://github.com/ion-design/ditto.site#readme",
+  "type": "module",
+  "bin": {
+    "ditto": "./bin/ditto.mjs"
+  },
+  "files": [
+    "bin"
+  ],
+  "engines": {
+    "node": ">=20"
+  },
+  "scripts": {
+    "test": "node --test test/*.test.mjs"
+  }
+}

packages/cli/test/unpack.test.mjs

@@ -0,0 +1,186 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+import { spawn } from "node:child_process";
+import { createHash } from "node:crypto";
+import { mkdtemp, readFile, rm, stat, writeFile } from "node:fs/promises";
+import { tmpdir } from "node:os";
+import { fileURLToPath } from "node:url";
+import { dirname, join } from "node:path";
+
+const BIN = fileURLToPath(new URL("../bin/ditto.mjs", import.meta.url));
+
+function sha256(s) {
+  return createHash("sha256").update(Buffer.from(s)).digest("hex");
+}
+
+/** Run the CLI. `stdin` (optional) is written to the child's stdin. */
+function run(args, { stdin, env } = {}) {
+  return new Promise((resolvePromise) => {
+    const child = spawn(process.execPath, [BIN, ...args], {
+      env: { ...process.env, ...env },
+    });
+    let out = "";
+    let err = "";
+    child.stdout.on("data", (d) => (out += d));
+    child.stderr.on("data", (d) => (err += d));
+    child.on("close", (code) => resolvePromise({ code, out, err }));
+    if (stdin !== undefined) child.stdin.end(stdin);
+    else child.stdin.end();
+  });
+}
+
+async function withTmp(fn) {
+  const dir = await mkdtemp(join(tmpdir(), "ditto-cli-"));
+  try {
+    return await fn(dir);
+  } finally {
+    await rm(dir, { recursive: true, force: true });
+  }
+}
+
+test("unpack: writes a text file tree from a result envelope", async () => {
+  await withTmp(async (dir) => {
+    const content = "export default function Page() { return null; }\n";
+    const doc = {
+      jobId: "job_1",
+      files: {
+        "package.json": { type: "text", content: '{"name":"x"}\n', bytes: 13, sha256: sha256('{"name":"x"}\n') },
+        "src/app/page.tsx": { type: "text", content, bytes: content.length, sha256: sha256(content) },
+      },
+    };
+    const jsonPath = join(dir, "clone.json");
+    const out = join(dir, "out");
+    await writeFile(jsonPath, JSON.stringify(doc));
+
+    const res = await run(["unpack", jsonPath, out]);
+    assert.equal(res.code, 0, res.err);
+    assert.equal(await readFile(join(out, "package.json"), "utf8"), '{"name":"x"}\n');
+    assert.equal(await readFile(join(out, "src/app/page.tsx"), "utf8"), content);
+    assert.match(res.out, /Wrote 2 files/);
+  });
+});
+
+test("unpack: reads JSON from stdin with '-'", async () => {
+  await withTmp(async (dir) => {
+    const doc = { files: { "a.txt": { type: "text", content: "hi" } } };
+    const out = join(dir, "out");
+    const res = await run(["unpack", "-", out], { stdin: JSON.stringify(doc) });
+    assert.equal(res.code, 0, res.err);
+    assert.equal(await readFile(join(out, "a.txt"), "utf8"), "hi");
+  });
+});
+
+test("unpack: materializes inline base64 binary assets", async () => {
+  await withTmp(async (dir) => {
+    const bytes = Buffer.from([0x89, 0x50, 0x4e, 0x47, 0x0d, 0x0a]);
+    const doc = {
+      files: {
+        "public/logo.png": {
+          type: "binary",
+          content: bytes.toString("base64"),
+          encoding: "base64",
+          bytes: bytes.length,
+          sha256: createHash("sha256").update(bytes).digest("hex"),
+        },
+      },
+    };
+    const out = join(dir, "out");
+    const res = await run(["unpack", "-", out], { stdin: JSON.stringify(doc) });
+    assert.equal(res.code, 0, res.err);
+    const written = await readFile(join(out, "public/logo.png"));
+    assert.deepEqual([...written], [...bytes]);
+  });
+});
+
+test("unpack: fetches binary assets by URL and resolves relative URLs against base", async () => {
+  await withTmp(async (dir) => {
+    const asset = Buffer.from("PNGDATA");
+    const { createServer } = await import("node:http");
+    const server = createServer((req, res) => {
+      if (req.url === "/v1/clones/job_1/files/public/img.png" && req.headers.authorization === "Bearer k") {
+        res.writeHead(200);
+        res.end(asset);
+      } else {
+        res.writeHead(404);
+        res.end("no");
+      }
+    });
+    await new Promise((r) => server.listen(0, r));
+    const port = server.address().port;
+    try {
+      const doc = {
+        files: {
+          "public/img.png": {
+            type: "binary",
+            url: "/v1/clones/job_1/files/public/img.png",
+            bytes: asset.length,
+            sha256: createHash("sha256").update(asset).digest("hex"),
+          },
+        },
+      };
+      const out = join(dir, "out");
+      const res = await run(["unpack", "-", out], {
+        stdin: JSON.stringify(doc),
+        env: { DITTO_API_URL: `http://127.0.0.1:${port}`, DITTO_API_KEY: "k" },
+      });
+      assert.equal(res.code, 0, res.err);
+      assert.deepEqual([...(await readFile(join(out, "public/img.png")))], [...asset]);
+    } finally {
+      server.close();
+    }
+  });
+});
+
+test("unpack: --no-fetch skips remote binaries but still writes text", async () => {
+  await withTmp(async (dir) => {
+    const doc = {
+      files: {
+        "index.html": { type: "text", content: "<h1>hi</h1>" },
+        "public/img.png": { type: "binary", url: "/v1/clones/x/files/public/img.png" },
+      },
+    };
+    const out = join(dir, "out");
+    const res = await run(["unpack", "-", out, "--no-fetch"], { stdin: JSON.stringify(doc) });
+    assert.equal(res.code, 0, res.err);
+    assert.equal(await readFile(join(out, "index.html"), "utf8"), "<h1>hi</h1>");
+    await assert.rejects(stat(join(out, "public/img.png")));
+    assert.match(res.err, /skipped/);
+  });
+});
+
+test("unpack: refuses path traversal", async () => {
+  await withTmp(async (dir) => {
+    const doc = { files: { "../escape.txt": { type: "text", content: "x" } } };
+    const out = join(dir, "out");
+    const res = await run(["unpack", "-", out], { stdin: JSON.stringify(doc) });
+    assert.notEqual(res.code, 0);
+    assert.match(res.err, /outside output dir/);
+  });
+});
+
+test("unpack: fails clearly on a queued job with no files", async () => {
+  await withTmp(async (dir) => {
+    const doc = { jobId: "job_9", status: "queued" };
+    const out = join(dir, "out");
+    const res = await run(["unpack", "-", out], { stdin: JSON.stringify(doc) });
+    assert.notEqual(res.code, 0);
+    assert.match(res.err, /queued job/);
+  });
+});
+
+test("unpack: reports sha256 mismatch as a failure", async () => {
+  await withTmp(async (dir) => {
+    const doc = { files: { "a.txt": { type: "text", content: "hi", sha256: "deadbeef" } } };
+    const out = join(dir, "out");
+    const res = await run(["unpack", "-", out], { stdin: JSON.stringify(doc) });
+    assert.notEqual(res.code, 0);
+    assert.match(res.err, /integrity check/);
+    await assert.rejects(stat(join(out, "a.txt")));
+  });
+});
+
+test("help: exits 0 and prints usage", async () => {
+  const res = await run(["--help"]);
+  assert.equal(res.code, 0);
+  assert.match(res.out, /ditto unpack/);
+});