Skip to the text
Parseh
پParseh guide Reference›The daily loops
On this page

The daily loops

Most days with Parseh are one of a handful of routines. Here they are in order, a few steps each, with the buttons named as the pages name them. The sections of this guide explain every step at length; this page is the map.

1. Starting and stopping#

  1. Start Parseh: ./serve.sh on Linux, a double-click on Parseh.command on a Mac, on serve.bat on Windows.
  2. Open https://localhost:7654/ — or, from a phone or another computer on your network, the address Parseh printed when it started. The first time, each browser warns about the certificate: accept it once.
  3. On the hub, pick your language’s chip: every door then opens on that language, and counts what it holds in it.
  4. When you are done, ⏻ stop. If something is still running, it tells you what first; wait for the Working… pill to go, then stop.

2. Reading a book#

  1. Books on the hub, then the book’s card.
  2. Play the narration, and the page follows the voice. A click on a chunk plays it; hover puts the gloss in a cloud over the chunk you are reading instead of between the lines.
  3. A word you want to keep: Alt-click or Ctrl-click it, and the card sheet opens already filled in — to Anki, to an exercise deck, or as markdown for a document.
  4. A chunk nobody has glossed yet: switch on dictionary in the header (it is there once reading help has a dictionary for the language).

3. Watching a video#

  1. Videos on the hub, then the video.
  2. The transcript runs under the video, the line being spoken marked; a click on a line replays it, and the gloss is where you hover.
  3. The same Alt-click makes a card, with the caption as its context and the word’s own sound cut with cut the audio… — from the film, in any browser, for a film on this machine; for a YouTube video, recorded from the tab as it plays, which only Chrome and Edge on a computer can do.

4. Making cards and taking them to Anki#

  1. Make cards as you read and watch (above): they wait in Parseh’s card store.
  2. If you have edited anything inside Anki since the last time, sync first. Skipping this is the one way to lose an edit.
  3. Build the deck: build .apkg in a card sheet (with the deck chosen in it), or step 5 of the sync. The deck downloads, rebuilt from the store; in Anki, File → Import… it. Cards are matched by their identifier, so the ones already there are updated in place and keep their history.

5. After a session of editing inside Anki#

  1. In Anki, File → Export…, Anki Deck Package (.apkg), with Include media ticked if your cards have pictures or recordings.
  2. Anki on the hub: drop the export on step 2.
  3. Preview the changes (step 3), read them, then Apply to the store (step 4).
  4. Step 5: download the rebuilt deck and import it into Anki. Five minutes, and the two sides agree again.

6. Studying exercises#

  1. Exercises on the hub: each deck’s card says how many exercises are new, learning and due for review.
  2. Open the deck and Study now. Answer and Check (or Enter), or turn a flashcard with Show answer; ⤢ Enlarge shows a card as large as the window holds.
  3. Rate it — Again, Hard, Good, Easy, or the keys 1 to 4 — and the next one comes. When nothing is left, the page says when the next one is due.
  4. To go over some exercises without touching their schedule, select them on the deck’s page and Cram exercises.

7. Writing in the studio#

  1. Studio on the hub. LLM prompt: pick the target language, type your question, Copy prompt + question, and paste it to any model.
  2. The model writes a .md file: Upload .md / .zip it in the library — or Paste LLM answer when the model answered in the chat.
  3. Or write your own: + New opens a page that shows everything a document can hold in your language; keep what you need, Save & view.
  4. Read it, tag it, and Build PDF when you want paper — with PDF options ▾ for large print or black and white.
  5. Its exercises go to a deck with + Deck, or all at once with + Add all exercises.

8. Adding a video#

  1. Add a video on the video index. Say where it is — From YouTube or A film already on this machine — and who writes the glosses.
  2. An LLM — I paste the answer back: copy the prompt the page makes, paste it to any model, paste the answer back; it is checked before anything is written.
  3. Nobody — I gloss it in the player: the video opens with every caption cut into chunks and every gloss blank, and you fill them in with the pencil.

9. Adding a book#

  1. Add a book on the library page, and say what you are doing.
  2. Write it here, by hand: paste a chapter; the book is drafted with the text divided and every gloss blank, and you gloss it in the reader.
  3. Add to a book already here: more text onto the end of a book on the shelf, as a new chapter or more of the last one.
  4. Let an agent make it: choose the original with the file picker and press make the book’s folder; open that folder in the agent you use and tell it read AGENTS.md and begin. The book is on the library, marked being made, and its reader has a making panel to look at it now, to write what to change from now on, and to finish it when the agent is done.

10. Glossing with an LLM#

  1. In a book’s reader, gloss with an LLM in the header (or gloss around here with an LLM… in a chunk’s sheet); in a video’s player, gloss with an LLM in the bar, then a click on the first caption and one on the last.
  2. Pick the stretch, and copy the prompt. Paste it into any chatbot.
  3. Paste its whole reply into the LLM’s answer and press fill from the answer. Only chunks nobody has glossed are filled; the report says what was filled, kept and dropped.
  4. A gloss you want written again: delete gloss in its sheet or ✎ form first, then copy the prompt.

See Glossing a stretch with an LLM and Glossing captions with an LLM.

11. Writing one yourself, in the files#

  1. download in the reader’s header, or ⤓ in the player’s bar: the authored files as one zip.
  2. Work on them in a text editor — this is where find-and-replace across a chapter belongs — and zip the folder again.
  3. Drop the zip on Bring a book back at the foot of the library (or Bring a video back on the video index). It is checked before it is installed; a book then waits on the shelf for its build button.

12. Keeping everything safe#

Each shelf has a Backup and a Load from backup: the studio’s library, the books (Backup every book), the videos (Backup every video) and the exercise decks. Loading one brings each item back as itself, and what is already there is kept unless you choose to replace it.

Two of them do not hold everything:

Backup every book
leaves every book’s narration out, to save space: each book is packed as everything but the recording. Copy each book’s audio/ folder yourself; after a restore, put it back in the book’s folder, then press build (or rebuild) on the book’s card in the library.
Backup every video
does carry a film kept on this machine; a YouTube video’s picture and sound stay on YouTube.

Your Anki cards go to Anki. The file layout says where everything is on the disk.

13. Updating Parseh#

  1. ⚙ Settings → Updating Parseh, on the computer Parseh runs on. Check now, then Download the version it names — or choose a release’s zip you already have.
  2. Read what Ready to install says: which way it goes, what it writes and deletes, any file changed by hand. Press Update to… and Yes.
  3. Wait for the page to come back by itself. The update adds what the environment lacks, compiles the guide and rebuilds the readers; The last update at the top says what it did, and Back to where you were returns you to your page.
  4. What changed, version by version says what is new.

A Parseh installed before it could update itself moves into a fresh install once, by hand (Updating Parseh). And whenever a page says the environment lacks a package, run the installer again — ./install.sh, Parseh.command or install.bat — which adds what is missing, and start Parseh again.