# Command line Generated from the help of `elabora` by `pnpm run docs`: `elabora help` prints the same, and `elabora --help` one group. `[dir]` is a course folder (the current one by default). Run it from a checkout with `node bin/elabora.js`, or through the desktop app's bundled command. ## Project ```text init --id --title [--lang fr] Scaffold a course project (course.json, plan.json, AGENTS.md) [--audience <who>] [--business-goal <goal>] What the author already said of it: the plan's audience, business goal and [--duration 10-20|5|30+] [--presentation duration in minutes, the presentation, the LMS, and notes for agents in slides|scroll] [--lms a,b] [--notes <text>] AGENTS.md (the tone, the brand wanted, what is left to the agent) [--template <id>] Start from a module template (elabora templates): its charter and its screens with placeholder texts, in --lang when it is translated into it templates [--json] The module templates: a charter (theme, fonts, components, an animated backdrop and its variants) and 14 screens of placeholder texts each validate [dir] [--profile a,b] [--release] [--json] Check format, plan conformity, pedagogy, accessibility, budgets project [dir] [--client] [--screen <id>] Markdown view of the course (agent map, or client review) migrate [dir] [--dry-run] [--json] Write the course in the current format (older ones are read migrated): one change after a backup version, undone by elabora undo rename [dir] <old-id> <new-id> [--json] Rename a section, screen, layer, block or event: its references, the screen file and the state lock follow (learner progress keeps its key) ignore [dir] <code> <key> [--restore] [--json] When the author asks: ignore a warning that has a key (elabora validate prints it), as « Ignorer » in the editor does (course.json ignored, D31); it comes back when what it found changes. --restore: back ``` ## Projects (the app's library: one plain folder per project; archives to exchange them) ```text projects list [--json] [--validate] Projects of the library, newest first: title, course, date, size, validation, folder (--validate checks those not checked yet) projects path [<id>] [--set <dir> [--move|--copy]] Folder of the library, or of a project (id or its first characters). --set: use another folder; --move takes the projects along; --copy moves them by copying and checking every file (another disk) projects create <course-id> --title <t> [--language fr] [--template <id>] Create a course in the library, blank or from a module template (elabora templates) projects duplicate <id> [--title <t>] Copy a project under a new id (same course id; the history stays) projects rename <id> --title <t> Change the title of a project (an undoable edit of course.json) projects trash <id> Move a project, history included, to the library's trash projects trash list|restore <entry>|empty [--older-than <days>] What the trash holds; put a project back; delete for good export [dir] [--out <file>] [--history] [--sources] [--json] Write a project archive (default dist/<course-id>.elabora): the project's files and, with --history, its journal and backups; with --sources, the author's sources (sources/, may hold personal data) import <archive|folder> [--replace [<id>] | --copy] [--json] Import into the library: an archive (every file checked) or a folder (copied). When the library holds the course already, say --replace (the replaced files go to the trash, the history stays) or --copy import <file> [dir] [--title <t>] [--id <id>] [--lang fr] [--choose <decision>=<choice>,…] [--json] Beta: a Storyline 360 project (.story), a course Storyline published (folder or zip, web or LMS) or a Rise 360 export, as a new module: in dir, or else the library. Writes import-report.json (every loss, the choices to make); --choose takes one of its choices: presentation=scroll, backgrounds=keep, timing=steps, fonts=theme, media-accessibility=errors ``` ## History (changes made through Elabora tools and by agents; AI work is grouped in batches) ```text history [dir] [--limit 20] [--json] Changes grouped by batch of AI work: when, who, the request, what changed, what was undone history [dir] --chats [--limit 20] [--json] The turns of Elabora's chat: the tools refused to the agent (file edits, shell), what it did instead, what each turn cost, and whether refusals call for a new Elabora tool undo [dir] Undo the last change made through Elabora tools or by an agent undo [dir] --change <id|seq> Undo one change (an undo too: that redoes it) undo [dir] --batch [<id>] Undo a batch of AI work (default: the last one). Whatever changed since is kept, and listed mark [dir] "<request>" [--chat <id>] Label the next batch of AI work with the request that starts it. --chat: a turn of Elabora's chat, whose agent's changes all join it Claude Code: a UserPromptSubmit hook running elabora mark --stdin labels each batch with the prompt (read from the hook's JSON) ``` ## The author's sources (never published) ```text sources list [dir] [--json] The author's sources (sources/: a brief, a deck, recordings; never published), each with its readable version: ready, partial (its transcript to come), stale, missing, failed, native (read as it is: images, text, a PDF) or unsupported sources add [dir] <file|folder>… [--no-read] [--json] Copies files into sources/ (a file already there is not copied again, a name taken is numbered, one Windows refuses is made portable), then writes their readable versions (--no-read: later) sources read [dir] <file> [--no-refresh] [--json] A source's readable version (Markdown), written first when it is missing or out of date: a Word document with its comments and tracked changes, a deck's slides in reading order with their notes and pictures, a recording's transcript (heard on this computer, the model downloaded once, 671 MB); an image or a PDF: its path sources refresh [dir] [<file>…] [--force] [--json] Writes the readable versions missing, out of date or waiting for a transcript; --force writes them all again sources remove [dir] <file> [--json] Deletes a source and its readable version for good: no backup holds sources Nothing leaves the computer: a client's brief may hold personal data. A deck gives its slides' texts, notes and pictures (the slides themselves are not drawn: add the deck as a PDF too to see them); a file in an older office format (.doc, .rtf, .odt, .ppt, .odp, .key) is read once saved as .docx or .pptx; agents read PDFs themselves. ``` ## Media (video, audio and images of assets/) ```text media import [dir] <file…> [--to <folder>] [--name <name>] [--no-wait] [--json] Bring media into assets/ (video, audio, images): converted in the background to web masters (MP4 H.264/AAC, M4A, WebP; SVG as is), metadata such as GPS positions removed, the original never copied. Prints the paths at once, then waits (--no-wait: a worker goes on) media optimize [dir] [<path>…|--all] [--json] Convert media of assets/ that are not masters (W_MEDIA_NOT_OPTIMIZED); references follow a new extension, in one undoable change media captions [dir] <block> <file.vtt|.srt> [--name <name>] [--json] Captions of a video block (E_VIDEO_CAPTIONS_MISSING): the file written under assets/captions/ (SubRip converted to WebVTT), named by the block, in one undoable change media status [dir] [--wait] [--json] Media jobs (queued, running, done, failed) and media to optimise media cancel [dir] <job> Cancel a media job media worker [dir] Run the project's media jobs until none is left (import --no-wait) ``` ## Narration (the voice of narrated screens, and their timeline) ```text timeline [dir] [screen] [--language <lang>] [--json] A timed screen's timeline as text: its steps and their times, each line with its speaker, start, end and audio (ok, missing, stale), the blocks of each step, the timed blocks with their moments and effects, and the events its cues fire; then each layer that has its own timeline. Without a screen: every narrated screen and its audio timing [dir] <block> [--in <moment>] [--out <moment>] [--effect <name>] [--duration <s>] [--direction <side>] [--out-effect <name>] [--out-duration <s>] [--out-direction <side>] [--clear in|out|all] [--json] When a block comes and goes on its screen's timeline, in one undoable change. A moment: l2 (a line starts), l2:end, a word of it l2:imprimante (l2:imprimante:end, #2 its second time, +0.3s after it), 1.5s or 1.5s@2 from a step's start on a screen without narration, step2, step2:end. Effects: none, fade, rise, slide, zoom, wipe, draw, or the theme's; sides: left, right, up, down narration import [dir] <screen> <line> <file> [--language <lang>] [--keep-silence] [--json] A recorded human voice for a line: converted (M4A, AAC, mono), its silence at both ends trimmed, written to assets/narration/<language>/<screen>/<line>.m4a with its sidecar, in one undoable change (W_NARRATION_AUDIO_MISSING, _STALE) narration import-files [dir] [--screen <id>] <file>… [--language <lang>] [--keep-silence] [--json] Recordings of several lines at once: each file goes to the line its name gives (l2.wav, accueil-l2.m4a, or 02.wav: the second line of --screen), the others to the line of --screen whose text the voice says (with the word aligner installed); one undoable change a line narration from-plan [dir] <screen> [--replace] [--json] The screen's script from the voice-over the validated plan gives the items it covers (plan items' voiceover), a line each; --replace: a screen that has a narration loses it and its audio narration transcribe [dir] <file> [--language <lang>] [--service local|elevenlabs] [--spend <percent>] [--json] What a recording says, punctuated, its words with their times: on this computer by default (Parakeet, downloaded once, 671 MB; free, nothing leaves the computer), or by ElevenLabs with the author's key: from 1 % of the quota it says the share (exit code 3) and runs with --spend narration add-from-audio [dir] <screen> <file> [--text <text>] [--after <line>|--first] [--step <n>] [--speaker <id>] [--layer <id>] [--language <lang>] [--keep-silence] [--service local|elevenlabs] [--spend <percent>] [--json] A new line made from a recording: its text what the recording says (transcribed, or --text), its audio the recording (trimmed, encoded), its words timed when the aligner is installed, in one undoable change; after the last line by default, in the step and with the speaker of the line before it; --layer: a line of that layer's narration narration retranscribe [dir] <screen> <line> [--dry-run] [--text <text>] [--service local|elevenlabs] [--spend <percent>] [--json] The line's text written again from its recording (a recording says something else than the script): its old and new text printed, then written in one undoable change, the recording kept and its words timed again; translations of the line become « changed since ». --dry-run: print only narration keep [dir] <screen> <line> [--language <lang>] [--json] The line's text was corrected (a checked transcription, a typo) and its audio still says it: the audio stays up to date, its words follow narration split [dir] <screen> <line> [--json] Cut a step at a line (W_SEGMENT_TOO_LONG): the line and those after it in its step start the next step, so Continue paces them; later steps, their blocks and the anchors that name them move one on, in one undoable change narration align [dir] [screen] [--line <ids>] [--language <lang>] [--force] [--json] When each word of the voiced lines is said (forced alignment): written into their sidecars, where word anchors find their times. Downloads the word aligner of the language once (French 355 MB, English 95 MB); lines made or imported afterwards are timed at once. --line l1,l2 (with a screen); --force: lines already timed too. Says which words the voice may have swallowed ``` ## Voices (generated narration) ```text voice list [--language <lang>] [--json] The voices a narration line can be generated with: id, name, register and pace, licence of the recording it comes from, and whether its model is on this computer (else the size downloaded on first use) voice models [--language <lang>] [--json] The local voices by language and their model: on this computer (its size on disk) or not (the size downloaded on first use), being downloaded now (by the app, a command or an agent) or in use (the voices' part of elabora downloads list) voice install <language> [--force] [--json] Downloads the model of a language's local voices once for this computer (fr 354 MB, en 126 MB), with its progress; --force removes it first and downloads it again (a reinstall) voice uninstall <language> [--json] Removes the model of a language's local voices from this computer (elabora downloads uninstall voices-<language>), refused while lines are said with it or it is being downloaded; the lines already made keep their audio, and the next generation downloads it again voice set [dir] <speaker> <voice> [--language <lang>] [--provider elevenlabs] [--name <n>] [--model <id>] [--stability <0-1>] [--similarity <0-1>] [--json] The voice of a speaker of course.json in a language (speakers[].voices[<lang>]): its lines made with another voice are then out of date voice generate [dir] <screen> [--line <id,…>] [--all] [--voice <id>] [--language <lang>] [--take <n>] [--dialogue] [--spend <percent>] [--json] Generates the lines of a narrated screen with their speaker's voice (by default those missing or out of date; --all: every line): each a new take, written to assets/narration/<lang>/<screen>/ with its sidecar (generated voice, credit), one undoable change a line. --voice sets that voice for the speaker of the lines first; --take puts back an earlier take of one line voice takes [dir] <screen> <line> [--language <lang>] [--json] The takes of a line kept on this computer: which one is its audio, which can be put back (--take) voice search [words] [--language <lang>] [--gender female|male] [--age young|middle_aged|old] [--accent <a>] [--use-case <u>] [--mine|--library] [--model <m>] [--high-quality] [--ready] [--page <n>] [--json] ElevenLabs voices (D29): the account's own, then the library's thousands, most used first; each with its accent, age, use, a recording to hear it (free), its price over the model's, its quality and whether it is ready for the model (Eleven v4 unless --model) in the language: prefer « checked in <lang> ». --high-quality: professional clones that passed ElevenLabs' check only; --ready: those checked with the model in the language voice estimate [dir] <screen> [--line <id,…>] [--all] [--dialogue] [--language <lang>] [--json] What generating would cost before anything is spent: the share of the monthly quota, characters and credits per line (about, at most), lines made free (local voices, earlier takes given back), the runs said as one exchange (--dialogue) voice account [--check] [--json] The ElevenLabs plan, the share of its monthly quota used, what a line costs of it, the key (masked) and where it comes from; --check tries each permission Elabora needs without spending voice key set|remove|status The ElevenLabs API key for the command line and agents: set reads it from the input (never an argument), keeps it readable by you only in Elabora's local data; ELEVENLABS_API_KEY overrides it A paid voice's cost is told as a share of the author's monthly quota (D30). Under 1 % of it, voice generate runs; from 1 %, it refuses with the estimate (exit code 3) and runs with --spend <percent> once the author agreed to that share. --dialogue says the lines of several speakers that follow each other in a step as one exchange, cut back into lines (their pauses from it). Stage directions in brackets ([chuchote], [rit]) are played by Eleven v4 voices and never shown. ``` ## Sounds ```text sound search [dir] <words…> [--source library|openverse] [--category <c>] [--kind effect|ambience] [--max-duration <s>] [--json] Sounds to add: Elabora's library « Sons courants » (CC0, offline, the default), or Freesound and Wikimedia Commons through Openverse (no account; CC0, public domain and CC BY only) sound add [dir] <library id | openverse:<id> | file> [--id <id>] [--trim <start>:<end>] [--fade <s>] [--loop] [--caption "<text>"] [--license <id> --author <name> --title <t> --url <https://…>] Copy a sound into assets/sounds/<id> with its sidecar (source, author, licence); a file needs --license (own: the author's) sound list [dir] [--json] The course's sounds: licence, credit, where each is used sound generate [dir] "<description>" [--duration <s>] [--loop] [--influence 0-1] [--id <id>] [--caption "<text>"] [--estimate] [--spend <percent>] [--json] Make a sound effect with ElevenLabs (the author's key): from 1 % of the monthly quota, ask the author, then --spend ``` ## Dialogues (branching conversations held in a block) ```text dialogue show [dir] <block> [--json] A dialogue block as a script: each node with its lines (name, stage direction, line), its options with their quality, points, effects and « then … » (where each one goes), their reply and feedback, then the endings. Stable text: it also serves diffs dialogue play [dir] <block> [--choose b,a,a] [--json] « Jouer un chemin »: the path of these choices (option ids, one for each decision met, retries included), its lines and notes, the ending, the score, the decisions and the duration (the lines' audio, else estimated from the text). Exit code 1 when a choice is not offered (it says which were) dialogue paths [dir] <block> [--limit <n>] [--json] Every path: how many, the best and worst score and points, the endings reached and never reached, the nodes never reached, and each path with its choices, ending, score and duration (2000 paths at most) dialogue export [dir] <block> --to html|pdf [--out <file>] [--date YYYY-MM-DD] [--json] The dialogue for the client: its map (every node and ending, placed by Elabora), then its script, as one page (html) or printed (pdf, by Elabora's or Google Chrome's Chromium); dist/<course>-<block>.<format> by default ``` ## Optional downloads (models downloaded once per computer) ```text downloads list [--kind <kind>] [--language <lang>] [--json] What Elabora downloads once per computer on first use (never in a course): the local voices and the word aligners, by language, what each is for and what happens without it, on this computer (its size on disk) or not (the size to download), being downloaded now (by the app, a command or an agent) or in use downloads install <id> [--force] [--json] Downloads one now (voices-fr, aligner-en…), with its progress; --force removes it first and downloads it again (a reinstall) downloads uninstall <id> [--json] Removes one from this computer, refused while it is in use (lines said, words placed) or being downloaded; what modules already contain stays, and the next use downloads it again elabora voice models, voice install and voice uninstall are the local voices' part of these (voice uninstall <lang> is downloads uninstall voices-<lang>). ``` ## Build and test ```text build [dir] [--profile a,b|all] [--release] [--out <dir>] [--dir] [--debug] [--media-quality high|medium|low] [--languages all] [--no-wait] [--json] [--target scorm12|web] [--progress local|none] [--navigation authored|free] [--noindex] [--url <address>] [--image <file>] [--review-origin <origin>]… Validate, then compile a deterministic SCORM 1.2 package + proof dossier. Media ship at delivery.mediaQuality (default high: the masters); lower qualities ship renditions, encoded once and cached. Waits for media jobs (--no-wait: refuses while they run). --languages all: one package of every language, the course's own and each translation that is ready and approved; learners switch from the player's « Langue » and keep their progress (D59). --target web: a module for a plain web server, no LMS (a folder to serve and its ZIP, <course>-<version>-web). --progress local (default: kept in the browser, resumed at the next visit) or none (nothing kept); --navigation authored (default: the course's rules) or free (every screen open, no lock); --noindex keeps search engines out; --url: the address of its index.html (canonical, sharing previews); --image: a PNG, JPEG or WebP shown when the link is shared. A showcase: --progress none --navigation free; a publication: the defaults. --review-origin (repeatable): a module for the Revue, shown in an iframe by a page of that origin (https://revue.elabora.studio; http on localhost for development): it carries the review bridge (elabora-review/1, docs/review-bridge.md) and talks to that page only; progress none, navigation free and noindex by default test [dir] [--package <zip>] [--profile a,b|all] [--mode iframe,popup…] [--browser chromium,electron,chrome,msedge,playwright,firefox,webkit] [--parallel 4] [--no-a11y] [--json] Run the package in the LMS emulator (profiles × launch modes × browsers); 4 runs side by side by default, fewer on a loaded machine preview [dir] [--port 4807] [--profile <id>] [--navigation authored|free] Live preview (standalone and in the LMS emulator): pages reload after each rebuild; free objects show in the in-app browsers of agent apps. The emulator plays the course's LMS (--profile: another); --navigation free opens every screen, whatever the course's rules look [dir] [--screen <id>,…|all] [--size 1600x900,375x812] [--state initial|answered] [--step <n>] [--layer <id>] [--at <moment> | --sheet] [--scroll <px|N%|block>,…] [--block <id> [--scale 2] [--act <action>]…] [--viewport] [--motion] [--language <code>] [--controls] [--select <css>] [--out shot.jpg | --out-dir <dir>] [--json] | --url | --stop See a screen while designing, in well under a second: a look server kept warm for the course (one browser, no build) renders it from the files as they are, with the player's header and footer; free objects an edit did not touch keep running. JPEG (PNG: --out x.png), in .elabora/look/ by default. Says whether a slide fits. --url: a page that follows the looks and the files, for a browser pane. --stop: stop the server (it stops itself after 15 minutes without a look). --at: a timed screen at a moment, the window with its bar: l2 (a line starts, its caption shown), l2:end, l2:imprimante (a word), l2:imprimante+0.3s, 1.5s@2, step2:end, 12.5 (seconds). --sheet: a contact sheet, one capture per mark of its timeline (start, each block that comes or goes, each cue, each step's end), captioned with the anchor and what happens; on a screen without a timeline, the page down its length (0, 25, 50, 75, 100 % or --scroll's). With --layer, a layer that has its own timeline: --at and --sheet are on it. --scroll: the window there, animations on (px, % of the page, or a block's id); several: a sheet down the page. --block: a close-up of the block (PNG up to 2 Mpx, else JPEG; in view unless --scroll says where), --scale 2 drawn twice as large. With --scroll, --block or --motion it says what moves: each block in the window (or the block) visible, clipped or outside, its animations and transitions, their timeline and progress, and the scroll-driven ones that CANNOT FINISH (the page ends before their range). --act (repeatable) drives the free object of --block from inside its frame before the capture: click:<text>, click:role=<role>[name=<n>], key:<key>, drag:x,y->x,y (0-1 across the object); it reports the object's state and events, and the object keeps that state until the screen changes. On a dialogue block, choose:<option> (one per decision met, retries included, as dialogue play) shows its thread at rest after those choices, its end in the window: notes, the next options, or the debrief. On a screen with free objects it says each control the page covers (a click on its middle would not reach it); --controls lists every control with its box in the window, instead of looking for them. --select <css>: the computed styles of what it matches (a layout's title, an ornament) and its parent's layout screenshot [dir] [--screen <id>,…|all] [--size 1600x900,375x812] [--state initial,answered] [--step <n>] [--layer <id>] [--at 0,2000,4000|l1,l2:end | --scroll 0,25%,50%,1800] [--motion] [--viewport] [--language <code> [--allow-missing]] [--from <screen>] [--out shot.png | --out-dir <dir>] Render screens to PNG once images and free objects are ready. answered: every activity shows its tallest feedback. --at: filmstrip of the animations (ms), or of a narrated screen as each line starts (l2) or ends (l2:end), with --layer of that layer's own timeline; with --from <screen>, the passage from that screen (times from the move, animations on: continuity, the backdrop); --scroll: filmstrip down the page (px or % of it: scroll-driven animations and objects). One contact sheet. Several shots share one build and one browser; --out names one image, so several shots are refused with it (exit code 2): --out-dir gets one file each fit [dir] [--screen <id>,…] [--reference 1600x900] [--minimum 1280x720] [--phone 375x812|none] [--language <code> [--allow-missing]] [--fresh] [--json] Measure every slide (all steps shown) at the fit sizes: what scrolls, by how much, and what moves once its activities are answered. In scroll presentation: the length of each page and the contrast of its text, down the whole page. On a phone (375×812 unless sizes are given): the contrast of the text down the page and the player's bars that go onto two lines. --language: the translation (--allow-missing: a draft, missing texts in the source language). A screen unchanged since its last measure (its files, the theme, components, objects, Elabora, the browser) is not measured again: its fit is reused (--fresh measures every screen again) inspect <package.zip> Read any SCORM package: manifest, files, size diff <old.report.json> <new.report.json> [--profile a,b] What an update does to learners in progress, LMS by LMS profiles [--json] List executable LMS profiles ``` ## Hosting: account (elabora.studio) ```text account login --email <address> [--device <name>] [--language fr|en] [--handle <name>] [--json] Signs in to Elabora's hosting: a code is sent to the address and read from the input (never an argument); the session is kept readable by you only in Elabora's local data. At the first sign-in, --handle (or the input) chooses the account's address. The device is named « Elabora CLI · macOS » (or Windows, Linux), never after your computer; --device names it as you like account login --invitation <file>|- --space <space> [--json] Keeps a tester's invitation key instead (from a file, or the input with -), after checking it opens its space account handle <name> [--change] [--json] Chooses the account's address: https://<name>.elabora.page/. --change renames it (once in 30 days, no redirect: the old address stops at once, links to it break, learners start their progress again; the old name is held a year) account status [--json] Who is signed in (the token masked, where it is kept), the space, the account's limits and usage account logout [--json] Ends the session on the server and forgets it here (an invitation key or an AI key is only forgotten) account ai-key create [--name <name>] [--replace-session] [--json] Makes a read-only key for your AI with your session and keeps it readable by you only: it lists reviews and reads their comments, nothing else (agents' read tools take it first). --replace-session: forget the session file then, so that what agents can reach here only reads (sign in again to publish or answer) account ai-key list [--json] The account's AI keys (never a key), the one kept here marked account ai-key revoke <aik_…> [--json] Revokes an AI key at once (and forgets it here if kept) ``` ## Hosting: publishing and review ```text publish [dir] [--private] [--name <publication>] [--image <file>] [--index] [--progress local|none] [--navigation authored|free] [--languages all] [--review-origin <origin>] [--json] Builds the course for the web and publishes it in your space; only the files the space lacks are sent (a second publication of the same content sends nothing). Public: https://<space>.elabora.page/ <name>/ (default name: the course id), noindex unless --index. --private: a review build (<id>-review by default) that only a review shows; a new version goes to its open reviews, even with red checks (listed: they stop the web, not reviewers). --languages all: one package of every approved language, which learners switch in the player (D59; public only). A folder elabora build --target web wrote is published as it is publication versions <publication> [--json] The versions the hosting keeps of a publication (the current one marked): older ones stay at their pinned addresses publication withdraw <publication> <version> --yes [--json] Takes one older version down for good while the publication goes on: its pinned address answers 404, a review that showed it lists it as withdrawn (its comments stay), the files only it used are deleted. The current version is refused: publish a newer one first review create [dir] [--name <publication>] [--version <v>] [--title <t>] [--note <text>] [--link-out <file>] [--json] Shares a private publication for review: prints the reviewers' link once (or writes it to --link-out, readable by you only). --note: a word to reviewers, under « <your name> vous invite à relire ce module » on the review page (500 characters) review list [dir] [--all] [--json] The course's reviews (--all: the space's): state, versions, open and resolved comments, reviewers review link <review> [--link-out <file>] [--json] A new link for reviewers: the old one stops at once review close <review> [--read-only] [--json] Closes a review (nothing shown any more), or makes it read-only review open <review> [--json] Opens a closed or read-only review again, under the same link review delete <review> --yes [--json] Deletes a review, its comments and reviewers, and its private publication when no other review shows it review comments [dir] [--review <review>] [--since <rev>] [--state open|resolved] [--screen <id>] [--json] The reviewers' comments placed in the course: the screen (key, title, source file), the block (id, type, pointer), the step, layer, narration line and word, the version commented and whether a newer one or the screen's source changed since, the author, the text, the state and the replies. --since <rev>: only what changed review reply <review> <comment> [text] [--name <n>] [--json] Answers a thread (#4 or cm_…) as the author (the text, or the input); --name: the name shown (default: the account's) review resolve <review> <comment> [--json] Marks a thread resolved review reopen <review> <comment> [--json] Opens a resolved thread again The API is https://api.elabora.studio (ELABORA_CLOUD_URL elsewhere). A tester's key in a file also serves with ELABORA_CLOUD_KEY_FILE and ELABORA_CLOUD_SPACE. ``` ## Design ```text assets [dir] [--only <name>] [--sheet <file.png>] Run generators/*.mjs: regenerate figures from code, theme tokens and plan. --sheet: a contact sheet of what they wrote, on the course's surface colour theme upgrade [dir] [--dry-run] [--json] Put the theme's rules under the components (theme.cascade components-over-theme, D32), the course's look kept: what the theme sets on properties components also set goes into @layer overrides. One undoable change player [dir] [--enable <items>] [--disable <items>] [--move <item:bar[:index]>] [--set <item.option=value>] [--position <bar=window|page>] [--token <player.name=value>] [--logo <asset|none>] [--reset] [--link-terms [<ids>]] [--add-resource <file>…] [--title <t>] [--description <d>] [--target scorm12|web] [--json] The learner's player (D58): its two bars of items (logo, title, progress, spacer, outline, controls, resources, glossary, credits, display, language, previous, next, exit), what shows and why not, its tokens and findings. With changes: one undoable change. The spacer is one per bar, always on: --move spacer:top:3 says where the start of the line ends. --set controls.speeds=0.75-1.5, outline.label=section, controls.transcript=false, logo.plate=true, progress.parts=false, progress.percent=false, glossary.linkTerms=false; --token player.top.background=#FFFFFF (=default removes it). --link-terms links the glossary's terms (all, or the ids given) the first time each is in each screen. --add-resource copies any file into assets/resources/ and adds it to the resources, titled by --title, else the PDF's own title, else its name (one undoable change each). Resources, glossary, credits: elabora tools call elabora_player . --from <arguments.json> ``` ## Fonts ```text font list [dir] [--category <c>] [--json] The vetted fonts shipped with Elabora (offline): family, category, weights, styles, size, licence; ✔ marks those of the course font add [dir] <family> [--weights 400,700] [--styles normal,italic] [--for body,heading] [--subsets latin,latin-ext] Copy a vetted font and its licence into theme/fonts/<id>/, declare its faces in course.json and give its stack to tokens (--for body, heading, mono, display… or none; default: what the list says); latin, and latin-ext when a language of the course needs it font add [dir] --from <Fontsource package folder or .tgz> [--family <name>] [--weights …] [--styles …] [--subsets …] [--for …] font add [dir] --file <a.woff2,b.woff2> --license <OFL.txt> [--family <name>] [--for …] Any other font, with its licence: npm pack @fontsource-variable/<id>, or WOFF2 files (weights and style read from the files) ``` ## Components ```text component list [dir] Block types of the course: built-in and components/<id>/ component new [dir] <id> [--title <t>] [--level block|screen] component fork [dir] <parent> <id> [--title <t>] Fork a component (built-in or course); the fork remembers its parent component preview [dir] <id> [--width 375] [--out file.png] Specimen of its examples, rendered component migrate [dir] [<id>] [--from <version>] Run migrations.mjs steps on every instance (all or nothing) component add [dir] <id> --library <dir> [--as <id>] [--upgrade] [--force] Copy (or upgrade) a library component component publish [dir] <id> --library <dir> Copy a course component into a library block duplicate [dir] <block> [--id <new id>] Copy a block (and the blocks of its slots) block detach [dir] <block> Replace a component instance by its native fallback block continue [dir] <block> --from <block> | --cut [--duration <ms>] Continuity between screens (D65): the block carries on from a block of another screen (a free object keeps running, another block glides), or no longer does; --duration: the passage to it (default 700) convert [dir] <screen> <block> --to <type>[:<presentation>] [--dry-run] [--lang fr] [--json] « Transformer en… »: the block as another type or presentation, its id, items and translations kept; without --to, the possible targets ``` ## Free objects ```text object new [dir] <id> [--title <t>] [--objectives OBJ-1,OBJ-2 | --decoration] Scaffold a free object for plan objectives, or with --decoration a decoration of the course backdrop (theme.backdrop.object): a 2D canvas that follows the course (screen and section rank, direction), out of the player's bars object map [dir] <id> [--tests] [--json] What to read before changing an object: each code file's sections (« // ── Name ── ») with their lines and functions, the props (type, choices, default), words, objects, state, events and the names of the tests (--tests: each test's steps on one line). Then read the lines needed object test [dir] <id> [--update-snapshots] [--browser chromium|electron|chrome|msedge|playwright|firefox|webkit] [--only <test name>] [--force] Gates 2–6 of the free-object contract (result kept in object.gates.json); an object unchanged since its last run here is not run again (--force runs it) object replay [dir] <id> [--test <name>] [--block <id> | --screen <id>] [--size 1600x900] [--out sheet.png] [--json] Replay a test of object.tests.json in the built course, on the page that places the object, in real time: a contact sheet of its snapshot steps, and its expectations checked. The backdrop's decoration replays behind the course with its bars, from --screen (default: the first one) ``` ## Contract (the plan and its versions) ```text plan status [dir] [--json] The plan's versions (made, sent, validated) and what changed since plan write [dir] (--from <plan.json> | --patch <patch.json>) Write the plan at any time (the whole plan, or a JSON Patch of it), its schema checked first; items moved take their screens along. Agents never change a text kept « à l'identique » (verbatim) plan text [dir] <TXT-1|block> --text <wording> Reword a validated text of the plan (by its id, or a block that shows it) plan voiceover [dir] <item> (--from <file.txt> | --clear) The voice-over of an item, one line per line of the file (« marc: Bonjour… » for a speaker); elabora narration from-plan makes it a screen's script plan diff [dir] [--from <Vn|validated|latest>] [--to <Vn|current>] [--json] What changed in the plan, element by element (by default since the validated version, to the working copy) plan revert [dir] <key> [--version <Vn>] Undo one change of the diff (item:ITEM-3, text:TXT-2…): back to the validated version (or the one given) plan note [dir] <key> (--text <note> | --clear) A note for the client on a change in no version yet: frozen with the next version, a comment in Word plan version [dir] [--note <text>] A person's: freeze the working copy as the next version (V1, V2…) plan approve [dir] [--version <Vn>] --by "<name, organisation>" [--reference <ref>] [--date YYYY-MM-DD] A person's: the client validated a version (the latest by default) plan export [dir] [--version <Vn|current>] [--format pdf|docx|html] [--out <file>] [--date YYYY-MM-DD] [--json] The plan's document, what the client reads, comments and signs: the latest version (the working copy without one), its changes since the validated version as tracked changes in Word; PDF (default; printed by Elabora's or Google Chrome's Chromium), Word or one HTML page, in dist/<course>-plan.<format> by default. Records the send on the version; never makes one ``` ## Translation (one package per language; translations live in the source course, translations/<lang>.xlf) ```text translate export [dir] --to <lang> Create or update the XLIFF 2.0 file: every string, what changed since translate status [dir] [--to <lang>] Translated, to translate and stale units; broken maths or placeholders translate reuse [dir] --to <lang> --from <course or .xlf> [--json] Fill the units to translate (or stale) whose source text another course already translated, word for word (a module made from the same template: --from its folder), in one undoable change translate build [dir] --to <lang> --out <dir> [--allow-missing] [--json] Generate the translated course (then elabora build it as any course) translate approve [dir] --to <lang> --by "<name, organisation>" [--reference <ref>] Approve the translated plan (human decision, kept with the source) ``` ## Interoperability ```text lesson-md export [dir] [--out <dir>] Export simple blocks to LESSON.md 1.9 (Slate), with a loss report lesson-md import [dir] <lesson.md> --covers ITEM-n [--id <screen-id>] [--section <id>] [--after <screen-id>] ``` ## Agents ```text mcp [dir…] [--read-only] [--tools <name>,…] MCP server on stdio for a course folder; several folders, or a folder of courses, give a course argument to each operation; outside a course, only the docs, skills and library. No folder: every course Elabora knows (its library, the folders the app opened), read again at each call. --tools: only the named agent tools (unknown names are refused) codes [--json] Catalogue of diagnostic codes docs [<page>] [--section <heading>] [--full] The documentation as Markdown, without a checkout: the map of pages, a page (D29: a decision; a long one as its outline), or one section rules [dir] [--json] The rules for agents of this elabora (what a course's AGENTS.md holds between its markers), and where the course's stand rules update [dir] [--force] [--json] Bring the course's AGENTS.md rules up to date (done by itself while nobody edits them); --force replaces edited ones, the notes after « elabora:rules end » kept. One undoable change ``` ## Agent operations (shared with MCP) ```text tools skills [<name> [<reference>]] [--json] List bundled skills, or read one skill/on-demand recipe (no course needed) tools guide [--json] Agent workflows, transport choices and human decisions (no course needed) tools list [<tool>] [--read-only] [--tools <name>,…] [--json] Discover agent operations and their JSON input schemas (no course needed) tools call <tool> [dir] [--args '<json>' | --from <arguments.json>|-] [--read-only] [--raw] Call the same operation as MCP, with the same validation. Arguments default to {}: --args takes them inline, --from from a file (- reads stdin). Prints the tool's own JSON (its text when it answers in text); images go to files under the course's .elabora/look/, their paths in "images". --raw prints the MCP result (content, isError) instead. Exit 0 success, 1 refused, 2 bad input ```