CLI reference¶
Every command, with the summary tsdive --help prints. Each command links to its usage line, its options and an example run on the demo archives.
The examples run in a directory where tsdive demo data wrote those archives, with the repository's examples/plans/demo.toml.
| Command | Summary |
|---|---|
tsdive demo |
write the demo archives into DIR (default tsdive-demo/) to try the rest |
tsdive ingest |
build an archive from an export; --init-meta META.json writes the metadata template first; one archive per column of a wide export; one archive per tag of a long export, the tag of each row in column COL |
tsdive profile |
data physics and statistics for a window |
tsdive segment |
regimes read off the samples themselves, for a tag with no MODE tag to key them |
tsdive screen |
flag samples outside a MAD baseline, or a per-regime one with --mode |
tsdive spc |
individuals chart; every rule reports its own hits, zero included |
tsdive mspc |
PCA T2 and SPE over aligned tags |
tsdive compare |
what changed between two periods: tags, pairs of tags, and their joint structure |
tsdive switchback plan |
a balanced random schedule of settings A and B, with its power over a history |
tsdive switchback analyze |
the difference between settings A and B under the plan, by randomization inference |
tsdive run |
walk one plan over several archives into one evidence ledger |
tsdive report-html |
render a static HTML evidence snapshot |
tsdive --help¶
tsdive - data-quality profiling and monitoring for process time series
commands, in the order an archive walks them:
demo [DIR] write the demo archives into DIR
(default tsdive-demo/) to try the rest
ingest <csv|parquet> --out ARCHIVE --meta META.json
build an archive from an export;
--init-meta META.json writes the
metadata template first
ingest <csv|parquet> --wide --out DIR --meta-dir DIR
one archive per column of a wide export
ingest <csv|parquet> --tag-col COL --out DIR --meta-dir DIR
one archive per tag of a long export,
the tag of each row in column COL
profile <parquet> [--window START/END] data physics and statistics for a window
segment <parquet> [--window START/END] regimes read off the samples themselves,
for a tag with no MODE tag to key them
screen <parquet> --baseline START/END --window START/END [--mode PARQUET]
flag samples outside a MAD baseline, or a
per-regime one with --mode
spc <parquet> --baseline START/END --window START/END
individuals chart; every rule reports its
own hits, zero included
mspc <parquet...> --baseline START/END --window START/END
PCA T2 and SPE over aligned tags
compare <parquet...> --before START/END --after START/END
what changed between two periods: tags,
pairs of tags, and their joint structure
switchback plan --window START/END --block PT1H --washout PT10M --seed N
-o PLAN.json [--history PARQUET --history-window START/END]
a balanced random schedule of settings A
and B, with its power over a history
switchback analyze <parquet...> --plan PLAN.json --target TAG [--covariate TAG]
the difference between settings A and B
under the plan, by randomization inference
run <plan.toml> [-o DIR] [--strict] walk one plan over several archives into
one evidence ledger
report-html <parquet...> [--window START/END] [-o FILE]
render a static HTML evidence snapshot
options, on every command:
--json one JSON object on stdout instead of
text (the analysis commands and ingest),
opening with result_kind and
tsdive_version
--no-color plain text; NO_COLOR does the same
--version print the tsdive version
windows, wherever START/END appears above:
START/END, START/PT5H, PT5H/END, or a date for one whole UTC day
exit status:
0 the answer was printed
2 usage error or invalid input
3 refusal: a typed tsdive error, printed
as [ErrorName] on stderr and, under
--json, as one object on stdout