docs: fully document the bring-your-own-Bible corpus format

Expand "Bring your own Bible" with the details a user needs to actually do it:
- the six-column TSV format with three real, tab-marked example rows and a
  column legend;
- which columns are load-bearing -- clectio matches on Book+Chapter+Verse
  (cols 1,4,5) and prints col 6; Abbrev/BookNum (cols 2,3) are ignored but must
  be present;
- the Vulgate-versification requirement made explicit (a differently numbered
  Bible misaligns), with the "start from vul.tsv, swap the text column" tip;
- pointer to the four ready corpora shipped with lectio, with a runnable drb
  example;
- a "switching back to the Vulgate" note (the git-tracked text_*.lz blob is
  overwritten by a CORPUS= build).

Docs only; no code change.

Claude-Session: https://claude.ai/code/session_01S1CD1u3tgSS4xNu6WHfp5a

1 files changed, 51 insertions, 9 deletions

diff --git a/README.md b/README.md
index eb763e2..0de5efe 100644
--- a/README.md
+++ b/README.md
@@ -45,17 +45,59 @@ One form per binary; for both, build twice. The interface is English only.
 
 ## Bring your own Bible
 
-The default build embeds the Latin Vulgate. To compile with another
-translation, hand `make` a corpus in the same 6-column TSV format
-(`Book<TAB>Abbrev<TAB>BookNum<TAB>Chapter<TAB>Verse<TAB>Text`), using Vulgate
-chapter/verse numbering:
+The default build embeds the Latin **Vulgate**. clectio compiles in only the
+~12,000 verses the lectionary actually cites (never a whole Bible), so it stays
+well under 1 MiB with any text. To build with a different translation, point
+`make` at a corpus and it repacks exactly those verses from it:
 
-    make CORPUS=/path/to/mybible.tsv
+    make CC=gcc CORPUS=/path/to/mybible.tsv
 
-clectio compiles in only the ~12000 verses the lectionary actually cites, so the
-binary stays well under 1 MiB whatever the Bible. A verse the corpus lacks prints
-blank; it never fails the day. (The default build needs only a compiler; a
-`CORPUS=` rebuild also needs `gzip`, which `mktext` uses to pack the text.)
+You already have four corpora in the required format: lectio ships them under
+`internal/bible/` — Vulgate (`vul`), Douay-Rheims (`drb`), Wujek (`wuj`) and
+Greek (`grb`). For an English build:
+
+    make CC=gcc CORPUS=~/git/projects/lectio/internal/bible/corpora_optional/drb.tsv
+
+### Corpus format
+
+A plain UTF-8 file, **six columns separated by a single TAB, one verse per
+line**. These three real rows show the shape (arrows mark the tabs):
+
+    Genesis → Ge   → 1  → 1  → 1  → In the beginning God created heaven, and earth.
+    John    → John → 43 → 20 → 1  → AND on the first day of the week, Mary Magdalen …
+    1 John  → 1Jn  → 62 → 5  → 11 → And this is the testimony, that God hath given …
+
+    column: 1 Book   2 Abbrev  3 BookNum  4 Chapter  5 Verse  6 Text
+
+clectio matches a citation to a row on **Book + Chapter + Verse** (columns 1, 4,
+5) and prints column 6. What that means for your file:
+
+  * **Book (col 1)** must be the full name the calendar's keys use — e.g.
+    `1 John`, `Song of Songs`, `Numbers`, not `1Jn` or `Cant`. Copy the exact
+    spellings from any shipped `*.tsv` (the safest start is to take `vul.tsv` and
+    replace only the text column).
+  * **Chapter and Verse (cols 4, 5)** must follow **Vulgate numbering** — the
+    versification the lectionary is keyed to. A Bible numbered differently
+    (modern Psalm numbers, split or renamed books) will silently misalign, so a
+    translation not already Vulgate-numbered has to be re-numbered first.
+  * **Abbrev (col 2) and BookNum (col 3)** are ignored by clectio, but the
+    columns must still be present so every line has six fields (a line with fewer
+    is skipped).
+  * **Text (col 6)** is the verse; runs of whitespace are collapsed to one space.
+
+A verse the corpus does not contain prints blank — it never fails the day; the
+build just warns how many verses were missing. The default Vulgate build needs
+only a C compiler; a `CORPUS=` rebuild also needs `gzip` (which `mktext` uses to
+pack the text).
+
+### Switching back to the Vulgate
+
+The shipped `text_of.lz` / `text_ef.lz` hold the packed Vulgate text and are
+git-tracked; a `CORPUS=` build overwrites the one for the form you built. To
+restore it:
+
+    git checkout text_of.lz          # or text_ef.lz for an EF build
+    make CC=gcc clean && make CC=gcc
 
 ## Coverage