funkpower
c071f84829
citefind: the references in a bibliography, and what could not be read off them
...
Cuts a reference list into entries — at the note numbers where there are any,
at the layout where there are not — and reads each as far as its punctuation
allows. Fields that would not parse are named in `unsettled` rather than
guessed at, and `--prompt` asks a model about exactly those and no others.
Over 202 real reference sections from the corpus: 8,566 references, 95% with
a title, 82% with a year, 46% settling every field with no model involved.
A title is refused where the entry was never divided, so no record reports
its own text back as the work it cites.
`ibid.` is flagged and left alone — it points at a work, and the entry does
not say which. A rule of dashes is filled in, because the style defines it as
the author above.
chunkmd gains `note_starts` and `cites_something`: recognising a page of
notes and splitting one are the same knowledge, and two copies of that
pattern would drift.
2026-09-20 10:38:48 -04:00
funkpower
30ff3361ee
recipes: a catalogue out of what the files already say
2026-09-20 10:03:20 -04:00
funkpower
d24e37bd12
langid and tocfind: what language a book is in, and what it says its chapters are
2026-09-20 10:02:15 -04:00
funkpower
05759ab902
bookmeta: the metadata a file already carries, and a question about the rest
2026-09-20 09:41:37 -04:00
funkpower
0b849c39c4
cueseg: chapter by vectors somebody else made, where there are some
...
The hashed bag of words asks only whether two stretches repeat the same
words, and cannot tell that ships and vessels are one subject. An embedding
can. segment_with takes one vector per window and falls back to the words
when it is given none, or the wrong number of them — a caller that embedded
some other recording's windows is asking about that recording.
windows_of gives the text to embed, in the order the vectors come back in,
so nobody has to reimplement the windowing pass to use this.
Bound in all three languages, and on the command line as --windows and
--embeddings, which pipe through emb. The C ABI version goes to 2:
catena_segment_cues gained an argument.
2026-09-20 09:12:04 -04:00
funkpower
dd09499e68
bindings: cueseg, so a transcript can be chaptered in process
2026-09-20 09:00:26 -04:00
funkpower
f9f809f701
chunkmd: the half of a word boundary that applies before a digit
2026-09-20 07:18:34 -04:00
funkpower
84d95b1022
mdscrub: a page anchor the engine cannot backtrack on
2026-09-19 21:01:24 -04:00
funkpower
577b381868
bindings: how to verify a swap, without the scaffolding for one
2026-09-19 20:43:16 -04:00
funkpower
645116d5ed
shadow: run both implementations over a corpus and diff them
2026-09-19 20:40:15 -04:00
funkpower
31f7c3f87c
recipes: the ocr flag depends on which verdict it was
2026-09-19 19:49:20 -04:00
funkpower
d47881ef3b
idnorm and bibfetch: an identifier read, and a register asked
2026-09-19 19:19:54 -04:00
funkpower
4d14544409
bibclean: an abstract that arrived as an index
2026-09-19 19:09:27 -04:00
funkpower
e6c51a5943
parity: the vocabulary rules gained a field
2026-09-19 19:07:43 -04:00
funkpower
484017fc18
vocabmerge: a subject scheme, supplied rather than compiled in
2026-09-19 19:06:57 -04:00
funkpower
05b363ea8a
seriesof: the series a book belongs to, and which volume this is
2026-09-19 19:03:47 -04:00
funkpower
e9d22ee1fa
bindings: typecheck the declarations the package ships
2026-09-19 18:57:24 -04:00
funkpower
48600bd90c
bindings: a C ABI and an npm package alongside the wheel
2026-09-19 18:40:08 -04:00
funkpower
cf126dddc3
bindings: the libraries in process, for Python
2026-09-19 17:54:18 -04:00
funkpower
a80f8a024d
chunkmd: say rfind, which is what this was doing
2026-09-19 17:36:18 -04:00
funkpower
e811a4d724
ci: authenticate the clone; the repository is private
2026-09-19 17:30:58 -04:00
funkpower
b32fc8b6a1
recipes: cut a line about itself
2026-09-19 17:28:04 -04:00
funkpower
8556f7ae84
readme: point at the recipes from under the tools
2026-09-19 17:27:53 -04:00
funkpower
acfab066ed
docs: recipes, and where these tools meet a database
2026-09-19 17:21:07 -04:00
funkpower
3a7dea7d95
ci: the image the runner already pulls
2026-09-19 17:17:37 -04:00
funkpower
ce14545553
ci: ask for the label the runner actually serves
2026-09-19 17:14:10 -04:00
funkpower
9fd654ad5f
docs: the manual-page flag, and the records that belong to one tool
2026-09-19 17:06:17 -04:00
funkpower
619672d403
pdfgutter: never rewrite a page tree it cannot account for
2026-09-19 16:51:12 -04:00
funkpower
a03dcd400c
avplan: paths an encoder can actually read, and a pipeline that runs
2026-09-19 15:31:17 -04:00
funkpower
ab33ac4693
every tool writes its own manual page
2026-09-19 15:25:13 -04:00
funkpower
df2a62ec84
pdfgutter: drop a doctest marker from prose
2026-09-19 15:20:12 -04:00
funkpower
3c697bee50
pdfgutter: two facing pages photographed as one, found and cut
2026-09-19 15:19:35 -04:00
funkpower
49f12c9718
avplan: whether a recording plays, and the one call that fixes it
2026-09-19 15:07:33 -04:00
funkpower
ea0a3694b1
mdsplit: the works inside one file, and the books inside scripture
2026-09-19 14:58:54 -04:00
funkpower
fb63b3f9ac
note where the records that belong to one tool are described
2026-09-19 14:38:21 -04:00
funkpower
9bf59f6ebb
vocabmerge: the near-duplicate headings a catalogue accumulates
2026-09-19 14:37:22 -04:00
funkpower
739e6c82e9
lints
2026-09-19 14:30:22 -04:00
funkpower
3c16892ebf
bibclean: a name from the book's own pages, where the record has none
2026-09-19 14:29:12 -04:00
funkpower
14e1fcf3ef
bibclean: slugs, and whether the one a record has still names it
2026-09-19 14:23:36 -04:00
funkpower
7fa610c578
each crate carries its own readme, and the contract describes rather than instructs
2026-09-19 14:15:01 -04:00
funkpower
2e2988b1b1
docs: a page for every tool and every library, with real output
2026-09-19 13:09:20 -04:00
funkpower
b0743d5bf7
a missing word, a hidden dependency, and a second pass that changed things
2026-09-19 13:00:29 -04:00
funkpower
6e26d5cf4b
bibclean: record faults, weighed, with the repairs that need no lookup
2026-09-19 12:50:53 -04:00
funkpower
c6575b2d42
emb: lines into vectors, against a server that owns the model
2026-09-19 12:43:20 -04:00
funkpower
1071f0bed3
booksniff, and one place for the parts every tool shares
2026-09-19 12:40:20 -04:00
funkpower
36afe3350b
cuepack: cues into passages that keep their timestamps
2026-09-19 12:32:41 -04:00
funkpower
b6602ee1e9
rrf: rank fusion without the database
2026-09-19 12:29:54 -04:00
funkpower
339f402610
ported patterns mean in rust what they mean in python
2026-09-19 12:26:41 -04:00
funkpower
17a03b0087
cueseg: topic chapters from a transcript, and a worked pipeline in the readme
2026-09-19 12:21:35 -04:00
funkpower
c7d0767376
cuekit and cuenorm: three subtitle formats, one cue stream
2026-09-19 11:54:36 -04:00
funkpower
33c2e40a1a
pdftriage: text layer verdicts, including the reflected-font recovery
2026-09-19 11:46:34 -04:00
funkpower
bd0ae11d01
mdscrub: decode character references, keep block boundaries, narrow the fence
2026-09-19 11:36:31 -04:00
funkpower
1b3c002b3d
mdscrub: residual html, control characters, headings, page markers
2026-09-19 11:30:05 -04:00
funkpower
e0473fb8ae
chunkmd: page and chapter aware chunking, roles, footnote flag
2026-09-19 11:22:56 -04:00
funkpower
acf8605d37
format
2026-09-19 11:12:23 -04:00
funkpower
6d9b7415dd
corpus-core: make the stream contract hold under hostile input
2026-09-19 11:11:58 -04:00
funkpower
6c2f5352a0
one gate: check.sh, run locally and by the workflow
2026-09-19 11:07:52 -04:00
funkpower
d8929371b4
parity harness, pinned to the python tokeniser over 10k lines
2026-09-19 11:06:55 -04:00
funkpower
0208b14b5d
textkit: token counting and the two segmenters
2026-09-19 11:04:27 -04:00
funkpower
00f4c38816
workspace, licence, and the three record types
2026-09-19 11:00:40 -04:00
funkpower