Getting Started

Getting Started

This guide installs Jarvis Line from a tagged GitHub release, runs guided setup, and verifies that your agent and TTS configuration agree.

Requirements

• Python 3.10 or newer
• A supported AI agent or an adapter that can emit Jarvis Line events
• A TTS backend available on your platform
• macOS for the currently validated end-to-end experience

Windows and Linux support is available as Preview. See Platform and Agent Support.

1. Install A Tagged Release

python3 -m pip install "git+https://github.com/reitenji/jarvis-line.git@v0.8.3"
jarvis-line --version

Using a tag makes the installation reproducible. Installing from the default branch gives you unreleased development changes.

2. Run Guided Setup

jarvis-line setup

The setup flow asks for:

• spoken language
• TTS backend
• final-only or commentary-and-final speech
• target agent
• project or global instruction scope
• optional attention alerts and Codex hooks
• optional voice test

Review the final plan before applying it. Kokoro model downloads and licenses remain explicit; setup does not silently download large model files.

For a non-interactive local default:

jarvis-line setup --default

3. Add The Agent Instruction

Jarvis Line does not edit AGENTS.md, CLAUDE.md, or GEMINI.md by default. Setup prints the command that generates the right block. Review that output, then paste it into the project or global instruction file your agent reads.

Codex or a generic AGENTS.md target:

jarvis-line instructions print agents --language "English"

Claude:

jarvis-line instructions print claude --language "English"

Gemini:

jarvis-line instructions print gemini --language "English"

Use a full language name. The instruction language and selected TTS voice must match. See Agent Instructions.

4. Install The Codex Hooks

jarvis-line install codex

The installer adds or refreshes Jarvis Line's SessionStart and supported attention hook entries without requiring you to hand-edit hook JSON.

Claude, Gemini, and other agents currently use instructions and the explicit event protocol rather than a bundled native hook installer.

5. Verify The Runtime

jarvis-line doctor
jarvis-line status
jarvis-line tts test --text "Jarvis line test is ready."
jarvis-line emit \
  --source manual \
  --session setup-check \
  --phase final \
  --line "The final completion chime is ready."

Expected result:

• doctor reports the config and required integration checks
• status reports watcher, worker, queue, TTS, and speech mode state
• the TTS test prints a playback result and you hear the short test line
• the explicit final event plays the optional completion chime, then the final line

jarvis-line tts test tests only the selected voice. It does not play the final chime because it is not a final agent event.

If the test command works but agent speech does not, confirm that the instruction block is in the instruction file the active agent actually reads. Then see Troubleshooting.

Updates

Check without changing the installation:

jarvis-line update check

Apply the configured Git-based update:

jarvis-line update apply

Run jarvis-line doctor and a TTS test after an update. Release notes may call out config migrations or integration changes.

Removal

Remove the Codex hook first:

jarvis-line uninstall codex
jarvis-line stop

Then uninstall the Python package with the same Python environment used during installation. Jarvis Line intentionally leaves user configuration and optional model files in place so upgrades do not erase them unexpectedly.