mydev/README.md

# MyDev

Local dev-environment monitor. AWS auth state, docker containers, git repo status.
Two modes: one-shot commands, or a live refreshing TUI dashboard.

## Build

```sh
go build -o mydev .
# optional: put it on PATH
mv mydev ~/go/bin/   # or /usr/local/bin
```

## Configure (required for `status` / `dash`)

Repo locations live in a config file — no default, no flag. First run:

```sh
mydev config init     # writes a starter config
mydev config path     # prints its location
```

Config lives at `~/.config/mydev/config.yaml` (honors `$XDG_CONFIG_HOME`):

```yaml
repos_roots:            # scanned one level deep for git repos
  - /Users/you/code
  - /Users/you/work     # multiple roots allowed
dev_dir:    /Users/you/code/dev      # commands/awsmfa, docker-compose.yml, bin/*.sh
legacy_dir: /Users/you/code/legacy   # rundev.sh for prod-data tools
```

If the file is missing, commands that need it exit with an error telling you to
run `config init`. New settings get added as keys here later.

## Use

```sh
# AWS MFA (via dev/commands/awsmfa — the real session, not `aws sts`)
mydev auth                 # status: account + expiry
mydev auth login           # re-authenticate (MFA prompt)

# Dev docker-compose stack (dev/docker-compose.yml)
mydev dev start            # auths first, then up -d --force-recreate
mydev dev stop | restart | remove
mydev dev status           # are running containers current with the dev repo?
mydev dev status --fetch=false   # skip the git fetch (no network)

# Prod data via legacy rundev.sh (mystore:devtool:*)
mydev data table  <args>
mydev data snapshot <args>
mydev data userfiles <args>

# DB access requests (dev/bin/*.sh)
mydev db bast              # request_bast_access.sh
mydev db grant <arg>       # grant_request.sh
mydev db list  [arg]       # grant_list.sh

# Misc
mydev port 3306            # what's listening on a port (lsof)
mydev docker               # container table
mydev status               # one-shot overview (auth + docker + repos)
mydev dash [--interval 5]  # live TUI dashboard
```

In the dashboard you can run actions without leaving it:

| key | action |
|-----|--------|
| `r` | refresh now |
| `a` | `auth login` (MFA prompt) |
| `s` / `x` / `R` | dev `start` / `stop` / `restart` |
| `:` | command palette — run **any** subcommand, with args |
| `q` / `ctrl+c` | quit |

Quick keys cover the common no-arg actions and show a `y`/`n` confirm first.
For anything that takes arguments, press `:` to open the command palette and
type a full command line, e.g. `data table mytenant`, `db grant 1234`,
`port 3306` — Enter runs it, Esc cancels.

The palette browses the live command tree as you type: it lists the root
commands first, the subcommands once you pick a group (`dev ` → start/stop/…),
and the usage hint for a leaf command (`port ` → `port <number>`). `Tab`
completes the first match. The list is generated from cobra itself, so it always
matches the real commands — no separate help text to keep in sync.

Commands whose output is already on the dashboard are hidden and refused: the
top-level `status` and the bare `auth` status. `auth login`, `docker`, and
`dev status` stay (actions, or more detail than the panels show).

Commands run in one of two modes, chosen automatically:

- **Interactive** (`auth login`, `dev start`/`restart`, all `data …` fetches) —
  need stdin or produce long live logs/prompts. The dashboard suspends and hands
  over the full terminal, then resumes.
- **Captured** (everything else: `status`, `db grant`, `db list`, `port`, …) —
  the dashboard stays up showing `running <cmd>… [c]ancel`. While it runs, other
  launches are blocked (so you can't stack commands); `c` cancels it (kills the
  process and returns to the menu, no result shown). When it finishes normally,
  the output shows in a scrollable result pane.
  `↑/↓` scroll, `esc`/`enter` close. Long lines wrap to the terminal width (so
  things like a generated password aren't cut off), and re-wrap when you resize.

Either way it re-invokes this same binary's subcommands, so behavior matches the
CLI exactly. The terminal is cleared when the dashboard starts, before each
interactive command, and on exit — so you never see stale output from a previous
command (or after quitting).

The AWS panel turns red
(`⚠ expiring`) when the session has under 30 min left. The DOCKER header shows a
container-freshness badge (`✓ all N current` / `⚠ N/M stale`) — computed locally
each refresh, no network. `dev status` adds a git-fetch check on top (is the dev
repo itself behind origin); the dashboard deliberately omits that to stay fast.

## How it works

Shells out to the real `aws`, `docker`, `git` CLIs — no SDKs, no creds handling
of its own. All probe logic lives in `internal/checks/` as plain functions; both
the one-shot commands and the TUI call the same functions, so behavior never drifts.

```
main.go              entrypoint
cmd/                 cobra subcommands (auth, dev, data, db, port, docker, status, dash, config)
internal/checks/     auth.go, docker.go, repos.go — read-only probes (capture output)
internal/config/     config.go — load/init the YAML config + path accessors
internal/run/        run.go — stream a command's stdio to the terminal (auth/compose/data)
```

Probes (`internal/checks`) capture output for display; actions that prompt or
stream (MFA, compose, data fetch) go through `internal/run` so they reach the
terminal directly. The `dev start`/`restart` commands auto re-auth first,
mirroring the old `myd` shell function.

## Extend

Add a probe: write a func in `internal/checks/`, then either a new file in `cmd/`
for a one-shot subcommand, or wire it into `dashModel` in `cmd/dash.go` (add a
field, a `fetchX` command, a message type, and a panel).

Ideas: ECS/EKS context, VPN state, kubectl context, disk space, brew outdated.

## License

MIT — see [LICENSE](LICENSE).