Skip to content
Luotsi Luotsi

Quickstart

This is the command detail behind First Five Minutes. Use that page when you are evaluating Luotsi for the first time; use this page when you want the command sequence in more detail.

0. Ask for the five-minute plan

Section titled “0. Ask for the five-minute plan”
Terminal window
luotsi quickstart
luotsi quickstart --human
luotsi quickstart-verify
luotsi quickstart --device <serial> --package <app.id> --artifacts artifacts/first-run
luotsi quickstart --artifacts artifacts/first-run --write-json --write-markdown

The command returns a JSON envelope with a five-minute path, concrete commands, proof_checks, an AI-agent prompt, an evaluation proof pack, and the local-first/artifact-first boundary that explains where Luotsi fits beside automation frameworks and hosted device services. Use proof_checks as the compact install/device/artifact/device-truth/replay-packet checklist for deciding whether the first five minutes produced usable evidence; each proof check reports whether it is ready_to_run, needs_input, or ready_after_artifact. Run luotsi quickstart-verify with the same inputs when you want the checklist grouped into ready, blocked, and later proof commands before starting the first-run path; it also writes the local plan/proof-pack handoff artifacts and reports the local proofs that passed. Use --human when you want the same plan as compact terminal text while you are getting oriented. Use --write-json --write-markdown to persist quickstart-plan.json, quickstart-plan.md, evaluation-proof-pack.json, and evaluation-proof-pack.md in the artifact root when a human or AI operator needs a durable handoff. The proof pack treats run-summary.json and run-summary.md from replay packet as the replay evidence gate before a run is production-ready. If you do not pass --device, the first command is luotsi doctor. This allows Luotsi to list adb-visible devices and return the exact selected-device next command.

1. Select or explain the target device

Section titled “1. Select or explain the target device”
Terminal window
luotsi doctor
luotsi devices
luotsi device-status --device <serial>

Use the next_command returned by bare doctor, or use the adb transport selector from luotsi devices / adb devices -l as your --device value.

2. Run the onboarding check

Section titled “2. Run the onboarding check”
Terminal window
luotsi doctor --device <serial>
luotsi doctor --device <serial> --fix

This verifies adb, device readiness, and live-view prerequisites. If you only care about the view stack, you can also use:

Terminal window
luotsi view setup --device <serial>

3. Choose a workflow

Section titled “3. Choose a workflow”

Operator loop

Section titled “Operator loop”
Terminal window
luotsi view --device <serial>

Use this when you want a local SDL mirror, operator hotkeys, screenshots, or manual interaction.

Agent loop

Section titled “Agent loop”
Terminal window
luotsi inspect --device <serial>

Use this when you want structured JSONL snapshots and command results for an agent-driven exploration loop.

Repeatable automation loop

Section titled “Repeatable automation loop”
Terminal window
luotsi scenario-init --file scenarios/smoke.json --name "smoke"
luotsi scenario-validate --path scenarios
luotsi run --path scenarios --device <serial> --report-junit junit.xml

Use this when you want a versioned scenario playbook (tracked in version control) with timing, artifacts, and CI-friendly reporting.

4. Keep the artifacts

Section titled “4. Keep the artifacts”

Most commands and sessions write useful output under an artifact root. Override it when you want a stable location:

Terminal window
luotsi view --device <serial> --artifacts ./artifacts/live-view
luotsi run --file scenarios/smoke.json --device <serial> --artifacts ./artifacts/smoke-run

5. Go deeper

Section titled “5. Go deeper”