chassidushub.org

Search the catalog of Chabad Torah and media (sefarim, sichos, letters, farbrengens, recordings, scans) by name or date, in Hebrew or English; or, with where="words", find the lines of scans and paragraphs of texts and transcripts that hold the words.

One item of the catalog with all its data, by its id (rh-…) or its readable path (/likkutei-sichos/12/3, /events/5742-05-10).

What is under an item, in order: a sefer's sichos or letters, a text's paragraphs, a farbrengen's recordings, a printing's scans, a set's items.

Every printing of a sefer in one call, in volume order: each with its volume, its kind, its path if it has one, and where its scans or PDF come from (HebrewBooks, a Drive file of אוצרות הרבי…), with the links.

The words of an item: a sicha's or letter's text (by its unit or text id), a page of a scan's text (OCR, proofread line by line), or a recording's transcript with when each paragraph is heard.

Suggest a correction to one item: the fields to change (a field set to null is removed), with a short title and why.

A sefer's shaar file, the README of a sefer: a header of fixed fields between --- lines (shaar: 1, title, title-en, subtitle, subtitle-en, by: rh-… one line an author, by-line, by-line-en, genre), then sections under fixed headings (## על הספר | About, ## סדר הספר | Structure, ## הדפסות | Printings, ## מקורות | Sources, ## הערות | Notes).

Send a sefer's whole shaar file (as get_shaar gives it, changed) for review.

Issues people opened about the catalog (a wrong fact, a missing page, a bad scan…), newest first: open ones by default, or about one item.

Open an issue about an item or the catalog: what is wrong or missing, for people to look into.

Add new items, or change or delete several at once, as one suggestion: what the site's editor does, many items at a time.

Add or replace a few segments in the words of many items at once, as one suggestion: a line a source left out, a heading, a machine's reading of a scan.

Suggestions people and bots sent, newest first: open ones (waiting for review, or sent back) by default.

One suggestion: its title, status and author, whether you may approve it (and if not, why), its checks that did not pass, and what it changes, a page of items at a time (each item's changed fields, before and after, without long words).

What approving a suggestion would do now, changing nothing: its stage (draft, sent_back, blocked, clashes, ready, merged, withdrawn), whether it would merge, whether you may approve it, each field that clashes with a later change on the site (the site's value and the suggestion's), what would stop the merge (failed checks, items that would not be valid, links to items that are gone), live items it would leave pointing at nothing, and other open suggestions changing the same items.

Your own suggestion (or any, as a steward): the site's later changes merged into it field by field, so it no longer clashes and shows only what it still changes; its checks run again.

Approve a suggestion sent for review, merging it into the catalog, when you may (its sets' keepers, or a steward).

Close (withdraw) a suggestion without merging it: your own, or any as a steward, for one that is no longer needed.

Reopen a closed (withdrawn) suggestion: it is open for review again, and its checks run again.

Combine several of your own suggestions not yet approved into one, as a pull request holds many commits, so they are reviewed and approved at once and never clash with each other.

Send a suggestion back to its author without approving it, saying what should change (keepers of its sets, stewards).

Ask ChassidusHub's machines to read a scan (kind ocr: Hebrew OCR, line by line, so its words can be searched and proofread) or to transcribe a recording (kind transcript: the Rebbe's Yiddish heard by a model trained on his voice, synced paragraph by paragraph).

What waits for the machines (OCR of scans, transcripts of recordings), in the order they will be taken, and what they did lately; or the requests for one item.

Farbrengens whose machine transcripts have paragraphs no person has checked, scans read by OCR with pages nobody proofread, and pages whose words a machine read (a subject index read from its scan) with segments nobody checked, the newest first, with how many in all.

How near the next transcription model is (farbrengens and hours people have checked since the last one, against the goal, and which farbrengens to check next, those before 5740 most wanted), and how much training data people have made by checking transcripts: clips and hours (gold: words and timing checked by a person; silver: words checked, timing by machine), train and test hours, what was left out and why.

The catalog as a tree, for organizing it: the top sets (no root), or one set or sefer, with the sets under it, the sefarim and other items in it (official sefarim first, in their order, then additions, marked {addition: kind to id}) and how much each holds (sets, items, units).

See what a plan of organizing operations would change, item by item, without saving anything: the fields that change, the paths that move and redirect, notes.

Send a plan of organizing operations (as preview_organize takes them) as one suggestion: several moves, renames and merges reviewed together.

Move items under a new parent: sefarim (or any items) into a set (with from, out of that set; mode "only" makes it their one set); to null with from takes them out of that set; a set under another set, or to the top (to null); sichos (units) to another sefer, at the end or at a position, their paths moving along; a printing to another sefer, a scan to another printing, a recording to another farbrengen.

Move items one level up the tree: a set to its parent's parent (the top, when its parent is a top set); a sefer out of a set into that set's parent set (from says which set, when it is in several).

Give an item (a set, a sefer, a sicha…) a new name in Hebrew and/or English, and optionally a new slug (the last part of its path; a set's slug too) or a whole new path.

Put siblings in order: a sefer's sichos, the sefarim of a set, the sets under a set (parent null for the top sets).

Make a new set, under a parent set or at the top, kept like its parent (the same keepers and policy), optionally moving items into it at once.

Remove a set that holds nothing (no sets under it, no items in it; move them out first).

Merge a duplicate into the item to keep (both of one type: two sefarim, two sets, two sichos…): everything under or pointing at the duplicate moves to the one kept, what it lacks is taken from the duplicate (external ids, sets, editions joined), and the duplicate is deleted, its paths leading to the one kept.

The catalog's tree is built of the official sefarim (the Rebbeim's sefarim and the other official sets); any other book is an addition to one: a commentary, an index, a book about it, a collection from it.