TALOS

A terminal dashboard for a directory full of git repos. it tells you which ones need pulling, which ones need pushing, which ones have uncommitted work, and which ones aren't actually git repos at all. Originally a bash script, now in Rust.

layout

• src/main.rs — clap CLI, rayon thread-pool setup, hands off to the UI.
• src/ui.rs — ratatui dashboard, key handling, background scan thread.
• src/scan.rs — parallel directory walk, git plumbing, fetch cache.
• src/repo.rs — Repo and State types and their colours.

install

You will need:

• a rust toolchain on edition 2024 (rustc 1.85+). easiest path is rustup.
• git on $PATH — talos shells out to git for every operation.
• macOS or Linux. the fetch shim points GIT_ASKPASS at /usr/bin/true and the cache lives under $HOME/.cache/talos.

from this directory:

cargo build --release

the binary lands at target/release/talos. drop it on your PATH (cp target/release/talos ~/.local/bin/) or call it directly.

first run

git clone https://github.com/<you>/talos.git
cd talos
cargo run --release -- --target ~/path/to/your/repos

If you'd like to install TALOS so it is reachable without navigating to the script folder, you can use:

cargo install --path .

The default target (the directory that holds all your git repositories) is ~/PANTHEON. To make it your own, pass --target and point at any directory whose immediate children are git clones. talos only walks one level deep and hidden subdirectories are skipped, so nesting like ~/code/work/foo won't get picked up unless you point at ~/code/work directly.

use

talos                       # scan ~/PANTHEON, fetch + status
talos --target ~/code       # scan somewhere else
talos --no-fetch            # skip the network round-trip (fast, stale data)
talos --force-fetch         # ignore the fetch cache and refetch everything
talos --fetch-ttl 300       # only refetch a repo if last fetch was >5min ago
talos --refresh 60          # seconds between background auto-rescans (default 30)
talos -j 4                  # fewer parallel workers (default 16)

keys, once the dashboard is up:

• r — rescan now (respects the fetch cache).
• f — force-fetch every repo on the next rescan.
• q / Esc / Ctrl-C — quit.

example output:

╺┳╸┏━┓╻  ┏━┓┏━┓
 ┃ ┣━┫┃  ┃ ┃┗━┓
 ╹ ╹ ╹┗━╸┗━┛┗━┛

Pull: 2/23  Push: 0/23  Dirty: 1/23  No upstream: 0/23

PATH: /Users/you/VAULT

#    repo               state       *  +/-    last commit       Σ
---- ------------------ ----------- -  ------ ----------- -------
1    REPO_01             synced      *  0/0    08/06/26          2
2    REPO_02             synced         0/0    20/01/26          5
...
22   REPO_22             pull           0/1    29/04/26         17
23   REPO_23             pull           0/44   29/04/26        922

what the columns mean

column	meaning
#	row index
repo	directory name
state	upstream relationship (see below)
*	magenta * when the working tree has uncommitted changes
+/-	commits ahead / commits behind upstream
last commit	committer date of HEAD
Σ	total commits on the current branch

states:

• synced (green) — matches upstream, nothing to do
• pull (yellow) — behind upstream, run git pull
• push (red) — ahead of upstream, run git push
• diverged (yellow) — both ahead and behind; merge or rebase required
• no-upstream (cyan) — branch has no remote tracking
• not-repo (default) — directory isn't a git repository

dirty (the *) is tracked independently of state. a repo can be synced * (in sync with origin but with local edits) or pull * (behind upstream and dirty), etc.

how it works

per repo, in parallel across a thread pool (default 16 workers):

1. git fetch --quiet origin — unless --no-fetch, the cache entry is fresh, or the remote is unreachable. credential prompts are disabled (GIT_TERMINAL_PROMPT=0) and network timeouts are tight (~10s) so a dead remote can't hang the dashboard.
2. local git rev-parse / status --porcelain / rev-list --left-right to compute branch, dirtiness, and ahead/behind.
3. git log -1 + git rev-list --count for the last-commit date and total-commit columns.

fetch cache

each successful fetch touches a zero-byte marker file at ~/.cache/talos/<repo>.fetch. on the next run, if the marker's mtime is younger than --fetch-ttl (default 60s), the fetch is skipped. clear it with rm -rf ~/.cache/talos or override with --force-fetch.