@lla-ma/cli

The lla.ma command-line tool — deploy, logs, env, and db from the terminal. The vercel/supabase of the lla.ma suite.

Status (milestone 4 + deploy). Auth (login/logout/whoami) and directory linking (link/unlink) are live. deploy and logs are live and hit endpoints confirmed against llama-apps — deploy is git-connected (the CLI triggers; the apps control plane clones → builds → runs), and logs streams the build over SSE (--watch to follow live). list, inspect, remove, and env are wired up but call a project API shape not yet confirmed against the deployed backend (see the note under Commands). db and rollback are planned — see Roadmap.

lla-ma is the canonical command; llma is a shorter alias for the same tool.

Commands

Identity & linking

lla-ma login                      log in with an email one-time code
lla-ma logout                     log out and clear the stored session
lla-ma whoami                     show the current account
lla-ma link [url]                 link this directory to a project / host-set
lla-ma unlink                     remove the link from this directory

Apps — deploy & inspect

lla-ma deploy [--branch <name>]   deploy the linked project (git → build → run)
lla-ma logs <deployment> [-w]     stream a deployment's build logs (--watch to follow)
lla-ma list (ls)                  list deployments for the linked project
lla-ma inspect <deployment>       show details for a deployment
lla-ma remove (rm) <deployment>   remove a deployment
lla-ma env ls                     list environment variables
lla-ma env add <key> [value]      add or update an environment variable
lla-ma env rm <key>               remove an environment variable
lla-ma env pull [file]            write env vars to a local file (default .env.local)

Board (kanban → git)

lla-ma branch <key>               cut/adopt a ticket's branch and switch to it (e.g. CLI-5)

Reads the kanban board on admin.lla.ma using your lla-ma login session and runs local git — no GitHub token in the CLI. If the ticket already has a branch (cut from the board's button) it adopts it; otherwise it asks the board to cut one, then git fetch + git switch onto it. Run it inside the ticket's repo.

Global

--json                            machine-readable JSON instead of human text
--endpoint <url>                  override the linked API endpoint for one command
-h, --help                        full command list
-v, --version                     print version

Backend confirmation. deploy (POST /api/projects/:id/deploy) and logs (GET /api/logs/:id, SSE) target routes confirmed in llama-apps. list, inspect, remove, and env currently assume an apps project API (/api/projects/:project/deployments/…) that isn't confirmed against the deployed backend yet — they may 404 until the CLI paths or the apps routes are reconciled.

Roadmap

Planned, not yet built:

• rollback <deployment> — re-point production to a previous deployment (apps).
• db group — db ls, db query "<sql>", db tables against the base (database) host through the gateway.
• db push — apply local migrations (backlog).
• Non-interactive / CI auth — account-scoped, revocable CLI tokens (today's management auth is the interactive OTP session cookie only).

Develop

npm install
npm run dev -- whoami     # run from source (no build) via tsx
npm run build             # compile TypeScript to dist/
npm link                  # dev: symlink `lla-ma` / `llma` onto your PATH (tracks the working tree)
npm run release           # install a self-contained global copy (build → pack → install -g)

Use npm link while iterating (it points at this checkout, so it breaks if node_modules/dist get wiped); use npm run release to install a standalone global build that survives independent of the working tree.