CLI reference
Flags, output formats, meta commands and exit codes.
jevql [flags] [dbname | postgres://...]
jevql serve [--listen 127.0.0.1:7433] [--token SECRET] [dbname | postgres://...]Flags may appear before or after the positional argument, as with psql.
Connection
| Flag | Meaning |
|---|---|
-h host -p port -U user -d dbname |
as psql; PGHOST, PGPORT, PGUSER, PGDATABASE, PGPASSWORD are honoured |
postgres://... positional |
a connection URI; DATABASE_URL is used when nothing else is given |
-w / -W |
never prompt for a password / always prompt |
Running statements
| Flag | Meaning |
|---|---|
-c "sql" |
run one or more statements and exit |
-f file |
run a file and exit |
| stdin | piped input is run like -f |
--explain |
print the plan and a cost estimate; no TypeSafe calls |
--timing |
print elapsed time after each statement |
-v |
verbose: collect SQL, batches, tokens, cost |
Output
| Flag | Meaning |
|---|---|
| default | aligned table like psql |
--expanded / -x |
one record per block |
--csv |
CSV with header |
--json |
an array of objects |
--json-table |
one JSON document per statement, the same shape jevql serve returns (protocol) |
Colour is used only when stdout is a terminal.
TypeSafe and budget
| Flag | Default | Meaning |
|---|---|---|
--api-key |
$TYPESAFE_API_KEY |
never printed |
--api-url |
https://api.typesafe.ai/v1/systemone |
or $TYPESAFE_API_URL |
--model |
jev-latest |
|
--threshold |
0.5 |
default for jev(); or $JEV_THRESHOLD |
--batch-size |
40 |
rows per request |
--concurrency |
6 |
parallel requests |
--max-rows |
2500 |
abort before any HTTP call if the collect exceeds this; 0 disables |
--max-chars |
0 |
abort if row objects exceed this many characters |
--columns a,b |
columns to send for alias-form calls | |
--cache path |
~/.cache/jevql/cache.db |
|
--no-cache |
Meta commands
Available in the REPL and in scripts, one per line:
| Command | |
|---|---|
\q |
quit |
\timing [on|off] |
toggle timing |
\x [on|off] |
expanded output |
\d [name] \dt \dv \dn \l |
describe, via the catalogs |
\set JEV_THRESHOLD 0.7 |
also JEV_MODEL, JEV_BATCH_SIZE, JEV_MAX_ROWS, TYPESAFE_API_KEY |
\cache [stats|clear] |
inspect or empty the answer cache |
\explain <query> |
like --explain for one statement |
\i file |
run a file |
help |
the built-in summary |
Other backslash commands print “not implemented” instead of reaching the server.
Exit codes
| Code | Meaning |
|---|---|
0 |
ok |
1 |
SQL or usage error |
2 |
TypeSafe API error or a budget guard fired |
The footer
After a jev statement, the REPL (and -v) prints:
jev: 129 judged, 4 req, 82 cache hits, 21.0k in tokens, $0.0009, 1.02sjudged counts distinct row-and-question pairs, cache hits included. Cost is estimated at $0.042 per million input tokens from the usage the API reports.