Question 1

Does Writed connect to any server?

Accepted Answer

Only once, and only for AI model files. On first visit, your browser downloads the embedding and chat models from public CDNs (Hugging Face, jsDelivr). Those CDNs see your IP and browser — the same as visiting any website — and never see your journal. After that one-time fetch, Writed runs fully offline.

Question 2

Does Writed upload my journal anywhere?

Accepted Answer

No. Entries are stored in your browser's IndexedDB. The embedding model and chat model both run on your device. Open DevTools → Network and you'll see no upload traffic — only the initial one-time model download.

Question 3

Do I need an account?

Accepted Answer

No. Writed does not have user accounts, sign-up, sign-in, or email collection. Open the app and start writing.

Question 4

Which AI models run in the browser?

Accepted Answer

Embeddings: BGE-small-en (33 MB). Chat: Llama 3.2 3B (~1.8 GB, WebGPU) on machines with a GPU, or Qwen 2.5 1.5B (~1 GB, WASM) on CPU-only machines. Writed auto-detects.

Question 5

How much disk space does Writed use?

Accepted Answer

Your journal entries are a few KB each. Model weights live in the browser cache: ~1–2 GB total depending on which model your device supports. Weights are cached after the first download, so the app runs offline afterwards.

Question 6

Can I export my entries?

Accepted Answer

Yes. Writed exports everything as a zip of plain Markdown files at any time. The format is standard `.md` with YAML frontmatter — readable in any text editor or imported into Obsidian, Logseq, etc.

Question 7

Does it work offline?

Accepted Answer

Yes. After the first visit downloads the models, Writed runs fully offline. The page is a static asset; no server roundtrip is needed for writing or chatting.

Question 8

What about mobile?

Accepted Answer

Writed works on modern mobile browsers, but the WebGPU chat model is desktop-first. On phones, the CPU fallback (Qwen 1.5B) runs but is slower. We recommend laptop/desktop for the best experience.

Question 9

What's Pro and when is it coming?

Accepted Answer

Pro adds optional end-to-end encrypted cloud sync across devices and encrypted backups. The local app stays free forever. Pro launches in v2 — join the waitlist on the homepage.

Question 10

How does the chat understand my journal?

Accepted Answer

When you save an entry, Writed splits it into chunks and computes a vector embedding for each one locally. When you ask a question, it retrieves the most relevant chunks using hybrid search (semantic + keyword + recency boost + date filter) and the chat model answers using those chunks as grounded context.

Frequently asked questions