Question 1

Is this going to ship my source code to an API?

Accepted Answer

Only what survives your local filters, and for each survivor only the path relative to the root you gave, the file name and extension, the size and the modification time — plus, with --content, the first 8 KiB if the file is text, with binaries detected and skipped. Nothing else leaves the machine. Symlinks are never followed, so --content cannot read outside the tree through one, and --dry-run shows how many entries would be sent and what it would cost without sending anything.

Question 2

How accurate is a probability-based match, and can I tune it?

Accepted Answer

jev returns a probability, not a verdict, so the trade-off is yours: raise --threshold for precision, lower it for recall. The default is 0.7, and --explain prints the probability next to each path, with --kind adding a source, test, config, docs, data, binary or other label. The README's honest guidance is that descriptions in English work best.

Question 3

What does a full scan cost, and is it slow on a large repo?

Accepted Answer

A metadata-only query runs roughly 400 tokens per file and a --content query a few thousand, at $0.042 per million input tokens with output free, so scanning a 1,000-file project stays well under a cent. Requests run concurrently — 20 at a time by default through --concurrency — and the README says a few thousand files take seconds. Use --dry-run for an estimate before spending anything.

Question 4

Should this replace rg, or live next to it?

Accepted Answer

Next to it. jfind takes the same predicates as find, prints the same kind of output and is meant to sit in the same pipelines, which is why exit codes follow grep and -print0 feeds xargs. Reach for it when what you know about a file is what it does rather than what it is called; grep and ripgrep remain faster for literal text and regex. Also worth knowing: the PyPI package is jfind-cli because jfind was taken, though the command and module are both jfind.

Question 5

How do I install it?

Accepted Answer

uv tool install jfind-cli, or pipx or pip; it needs Python 3.13+ and a TypeSafe key in TYPESAFE_API_KEY or a .env file. The PyPI package is jfind-cli while the command and module are jfind.

Question 6

What classic find options does it carry?

Accepted Answer

-name, -type, -maxdepth, -newer and --hidden, with .gitignore handling, plus -print0 for xargs, --concurrency defaulting to 20 and --model defaulting to jev-latest.

Question 7

How does it treat symlinks and platforms?

Accepted Answer

Symlinks are never followed, like find -P, so --content cannot read outside the tree through one, and it works the same on macOS, Linux and Windows, with hidden always meaning dot-prefixed.

Jfind

About this project

Highlights

Questions about this entry

More Jev resources like this

From the Jev guides

From the Jev community