Write HTML. Render video. Built for agents.
JavaScript aiai-agentsanimationclaude-codeffmpegframeworkgsaphtmlhtml-to-videomcpmotion-graphicspuppeteerrenderingtypescriptvideo-editor
/**
* Content-addressing for a project directory, shared by the distributed-render
* adapters' `deploySite` verbs.
*
* Each adapter uploads a project under `…/sites/<siteId>/` (AWS: a file list,
* Cloud Run: a tarball) and skips content already uploaded. `siteId` is a
* SHA-256 over the project's files so identical content always maps to the
* same key — that contract has to be byte-identical across adapters, which is
* exactly why it lives here rather than being copy-pasted per adapter.
*/
import { closeSync, openSync, readdirSync, readSync, realpathSync, statSync } from "node:fs";
import { createHash } from "node:crypto";
import { join } from "node:path";
import { PLAN_PROJECT_DIR_SKIP_SEGMENTS } from "./plan.js";
// Bump when the hashed set changes so uploads cached under an older scheme are never reused.
const SITE_ID_SCHEME = "site-v2\0";
const READ_CHUNK_BYTES = 256 * 1024;
/**
* The files a site upload ships, as `/`-separated paths under `projectDir`. Symlinks are followed
* (root included), top-level {@link PLAN_PROJECT_DIR_SKIP_SEGMENTS} are left out as the plan stage
* does, and a dangling symlink throws.
*/
export function listSiteFiles(projectDir: string): string[] {
const files: string[] = [];
const ancestors = new Set([realpathSync(projectDir)]);
function walk(dir: string, rel: string): void {
for (const name of readdirSync(dir).sort()) {
if (rel === "" && PLAN_PROJECT_DIR_SKIP_SEGMENTS.has(name)) continue;
const full = join(dir, name);
const relPath = rel === "" ? name : `${rel}/${name}`;
let stat;
try {
stat = statSync(full);
} catch (err) {
throw new Error(`[listSiteFiles] cannot read ${relPath} (dangling symlink?)`, {
cause: err,
});
}
if (stat.isFile()) {
files.push(relPath);
} else if (stat.isDirectory()) {
const real = realpathSync(full);
// A folder linking back to an ancestor is skipped rather than walked forever.
if (ancestors.has(real)) continue;
ancestors.add(real);
walk(full, relPath);
ancestors.delete(real);
}
}
}
walk(projectDir, "");
return files;
}
/**
* SHA-256 over a scheme tag plus each of {@link listSiteFiles}'s paths and
* bytes → 16-character hex prefix. The prefix is the `siteId`.
*
* The hash includes the relative path plus every byte of each file, so a
* same-bytes rename still yields a fresh id. We trim to 16 chars because the
* full 64 isn't useful in an object key for legibility. Pass `files` when the
* caller already listed them, so the hashed set is the shipped set.
* Files are read in fixed-size chunks, so no file is ever held whole in memory.
*/
export function hashProjectDir(
projectDir: string,
files: readonly string[] = listSiteFiles(projectDir),
): string {
const hash = createHash("sha256").update(SITE_ID_SCHEME);
const chunk = Buffer.allocUnsafe(READ_CHUNK_BYTES);
for (const rel of files) {
hash.update(rel);
hash.update("\0");
const fd = openSync(join(projectDir, rel), "r");
try {
let read: number;
while ((read = readSync(fd, chunk, 0, chunk.length, null)) > 0) {
hash.update(chunk.subarray(0, read));
}
} finally {
closeSync(fd);
}
}
return hash.digest("hex").slice(0, 16);
}