captoys/diigo-dump/README.md

diigo-dump

Cloudflare Worker port of diigo-dump.mjs in the mm-diigo repo (projects/diigo-bak/diigo-dump.mjs) — dump Diigo bookmarks, most-recently-updated first, via the browser session cookie rather than the static API key.

What it does

One page: a page-number field and a dump button. Submitting POSTs to the worker, which fetches

https://www.diigo.com/interact_api/load_user_items
    ?sort=updated&count=24&page_num=<page>

with the session cookie and shows status, total, and the JSON items in a read-only <pre>. GET / is inert — dumps happen only on explicit submits.

Structure: WebApp as a Durable Object (webapp worker)

The credentialed authority is a Durable Object instance from the separately deployable ../webapp/ worker (service binding WEBAPP → its WebAppFactory entrypoint; diigo knowledge = only the mint call in src/index.ts). The instance is configured once — diigo's URL + seed headers (cookie, UA, x-requested-with) + 500ms rate gate — and persists its init in DO storage: it survives restarts, and diigo-dump or any other worker can reconnect purely by id. The cookie crosses the wire exactly once, at mint time.

Setup (dev):

cd diigo-dump
npx wrangler dev -c wrangler.jsonc -c ../webapp/wrangler.jsonc
# → http://localhost:8787 (webapp is secondary, binding-only)

npm run diigo-login >> .dev.vars   # askpass for the cookie, mints the
                                   # instance, writes DIIGO_COOKIE +
                                   # DIIGO_WEB_ID; then restart wrangler
# or, no askpass: open http://localhost:8787/_connect and paste the
# cookie into the mint form; it prints the DIIGO_WEB_ID line to add

Deployed:

cd ../webapp && npx wrangler deploy
cd ../diigo-dump && npx wrangler deploy
npx wrangler secret put DIIGO_COOKIE   # re-run diigo-login only when
npx wrangler secret put DIIGO_WEB_ID   # the cookie itself expires

Re-login is needed only when diigo invalidates the cookie; re-minting (POST /_mint with the new cookie) yields a fresh id and no other changes.

Provenance notes: the user-agent / x-requested-with headers and the interact_api endpoint mirror the original script; the fluent client from WebApp.js is restored in src/webapp.ts with WebApp and the rate-limit wrapper as RpcTarget subclasses. src/diigo.ts holds makeDiigoAPI and the dumpBookmarks generator; DiigoService.open() hands an RPC caller a rate-limited stub with the cookie attached only on the server side, and promise pipelining makes a whole pathjoin().query().get() chain a single round trip. One deliberate divergence from the original: url is a private field, not a public property — on an RpcTarget, a public property is readable over RPC (a property read is an argument-less method call), and a query string can carry credentials, so a public url would leak them to any stub holder.

RPC use (from another worker, same account)

// wrangler.jsonc of the caller
"services": [{ "binding": "DIIGO", "service": "diigo-dump", "entrypoint": "DiigoService" }]

import { dumpBookmarks } from "./diigo.js";

const stub = env.DIIGO.open(); // rate-limited, cookie stays server-side
// pipeline: one round trip for the whole chain
const first = JSON.parse(await stub.pathjoin("load_user_items")
	.query({ sort: "updated", count: 24, page_num: 1 }).get());
// or run the whole dump in this worker's address space:
for await (const item of dumpBookmarks({ diigoAPI: stub, pages: 2 })) {
	console.log(item);
}

Configuration

• Secret DIIGO_COOKIE: the Cookie header from diigo.com devtools.
  • Local dev: .dev.vars (gitignored). Note wrangler reads it only at startup — restart after edits.
  • Deployed: npx wrangler secret put DIIGO_COOKIE.

Run

cd diigo-dump
npx wrangler dev -c wrangler.jsonc -c ../webapp/wrangler.jsonc
# → http://localhost:8787; webapp rides along, binding-only
# (first run: npm run diigo-login >> .dev.vars, or the /_connect page;
# restart wrangler after adding DIIGO_WEB_ID)

Status as of 2026-10-06: manual integration test passes — 200, total: 1674. See the project ladder in ../capsite-checklist.md; context in ../capsite-notes.md. This is the cookie path; the API-key path (EmDash plugin) is still 401.