omi-ctl

#!/bin/bash
# omi-ctl — drive the running Omi desktop app via its local automation bridge.
#
# The bridge (DesktopAutomationBridge.swift) auto-enables on every non-production
# bundle and listens on 127.0.0.1:47777. It lets an agent jump straight to any
# screen and read app state without clicking through the UI.
#
# Env: OMI_AUTOMATION_PORT (default 47777) — set this per bundle if running
#      several named test bundles side-by-side (each needs its own port).
#      OMI_AUTOMATION_TOKEN can override the per-launch token file.
set -euo pipefail

PORT="${OMI_AUTOMATION_PORT:-47777}"
BASE="http://127.0.0.1:${PORT}"
TOKEN="${OMI_AUTOMATION_TOKEN:-}"
# shellcheck source=automation-token-path.sh
source "$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)/automation-token-path.sh"
TOKEN_FILE="$(omi_automation_token_file "$PORT")"

load_token() {
  if [ -z "$TOKEN" ] && [ -f "$TOKEN_FILE" ]; then
    TOKEN="$(tr -d '\r\n' < "$TOKEN_FILE")"
  fi
  if [ -z "$TOKEN" ]; then
    echo "omi-ctl: missing automation token; set OMI_AUTOMATION_TOKEN or wait for $TOKEN_FILE" >&2
    exit 1
  fi
}

curl_bridge() {
  load_token
  curl -fsS -H "Authorization: Bearer $TOKEN" "$@"
}

cmd="${1:-help}"; shift || true
case "$cmd" in
  state)
    curl_bridge "$BASE/state"; echo ;;
  health)
    # Health intentionally uses the unauthenticated identity route so callers
    # see backend and negotiated runtime diagnostics, not the auth-only state
    # snapshot served at the same path.
    curl -fsS "$BASE/health"; echo ;;
  log-path)
    curl -fsS "$BASE/health" | python3 -c '
import json, sys
health = json.load(sys.stdin)
path = health.get("logFilePath")
if not isinstance(path, str) or not path:
    raise SystemExit("omi-ctl: health did not provide logFilePath")
print(path)
' ;;
  navigate)
    target="${1:?usage: omi-ctl navigate <screen> [settings-section] [--show]}"; shift || true
    section=""
    activate=false
    for value in "$@"; do
      if [ "$value" = "--show" ]; then
        activate=true
      elif [ -z "$section" ]; then
        section="$value"
      else
        echo "omi-ctl: unexpected navigate argument: $value" >&2
        exit 2
      fi
    done
    body="{\"target\":\"$target\",\"activateApp\":$activate"
    [ -n "$section" ] && body="$body,\"settingsSection\":\"$section\""
    body="$body}"
    curl_bridge -X POST "$BASE/navigate" -d "$body" >/dev/null
    # The POST snapshot races the animated tab switch — settle, then report fresh state.
    sleep 0.3
    curl_bridge "$BASE/state"; echo ;;
  ui)
    mode="${1:-}"
    activate=false
    [ "${2:-}" = "--activate" ] && activate=true
    params=""
    [ -n "$mode" ] && params="\"mode\":\"$mode\",\"activate\":\"$activate\""
    body="{\"name\":\"set_automation_ui_presentation\""
    [ -n "$params" ] && body="$body,\"params\":{$params}"
    body="$body}"
    curl_bridge -X POST "$BASE/action" -d "$body"; echo ;;
  open-conversation)
    id="${1:?usage: omi-ctl open-conversation <id> [--transcript] [--show]}"; shift || true
    show=false
    activate=false
    for value in "$@"; do
      case "$value" in
        --transcript) show=true ;;
        --show) activate=true ;;
        *) echo "omi-ctl: unexpected open-conversation argument: $value" >&2; exit 2 ;;
      esac
    done
    curl_bridge -X POST "$BASE/conversation/open" \
      -d "{\"conversationId\":\"$id\",\"showTranscript\":$show,\"activateApp\":$activate}"; echo ;;
  actions)
    curl_bridge "$BASE/actions"; echo ;;
  action)
    name="${1:?usage: omi-ctl action <name> [key=value ...]}"; shift || true
    params=""
    for kv in "$@"; do
      key="${kv%%=*}"; val="${kv#*=}"
      [ -n "$params" ] && params="$params,"
      params="$params\"$key\":\"$val\""
    done
    body="{\"name\":\"$name\""
    [ -n "$params" ] && body="$body,\"params\":{$params}"
    body="$body}"
    curl_bridge -X POST "$BASE/action" -d "$body"; echo ;;
  wait-ready)
    tries="${1:-30}"
    for _ in $(seq 1 "$tries"); do
      s=$(curl_bridge "$BASE/state" 2>/dev/null || true)
      # `appState: main` is a UI projection, not an authentication guarantee.
      # The authenticated state snapshot is the only readiness contract exposed
      # by DesktopAutomationBridge: require a live, signed-in, fully-onboarded
      # owner snapshot after restoration has completed.
      if printf '%s' "$s" | python3 -c '
import json
import sys

try:
    envelope = json.load(sys.stdin)
    snapshot = envelope["result"]
except (json.JSONDecodeError, KeyError, TypeError):
    raise SystemExit(1)

ready = (
    envelope.get("ok") is True
    and isinstance(snapshot, dict)
    and snapshot.get("appState") == "main"
    and snapshot.get("isSignedIn") is True
    and snapshot.get("hasCompletedOnboarding") is True
    and snapshot.get("isRestoringAuth") is False
    and snapshot.get("snapshotStale") is False
)
raise SystemExit(0 if ready else 1)
'; then
        echo "$s"
        exit 0
      fi
      sleep 0.5
    done
    echo "omi-ctl: timed out waiting for a live signed-in owner-ready main state" >&2; exit 1 ;;
  screens)
    # Every token here must resolve in ChatFirstRoute.automationVisibilityDestination. `focus` and
    # `insight` were listed after their pages were deleted, so `omi-ctl navigate insight` posted a
    # target the app answered "ok" to and then did nothing with — the failure mode the bridge exists
    # to avoid. `goals` lost its "(chat-first)" note when the second shell did: there is one shell,
    # and `help` opens Settings > About, where getting help from a person lives.
    echo "dashboard|home  conversations  chat  memories  tasks  goals  rewind  apps|integrations  settings  permissions  help" ;;
  *)
    cat <<EOF
omi-ctl — drive the Omi desktop app via the local automation bridge

Usage:
  omi-ctl state                          App state snapshot (selected tab, auth, onboarding)
  omi-ctl health                         Bundle, backend, runtime, and exact current log identity
  omi-ctl log-path                       Print the exact current bundle-and-launch app log path
  omi-ctl navigate <screen> [section] [--show]
                                         Jump cursor-free; --show brings the app forward
  omi-ctl ui [quiet|interactive|normal] [--activate]
                                         Read or change non-production test-window presentation
  omi-ctl open-conversation <id> [--transcript] [--show]
  omi-ctl wait-ready [tries]             Block until a live signed-in owner-ready main state (0.5s/try)
  omi-ctl screens                        List valid screen targets
  omi-ctl actions                        List semantic actions the app exposes
  omi-ctl action <name> [key=value ...]  Run a semantic action (cursor-free, in-process)

Examples:
  omi-ctl navigate rewind
  omi-ctl navigate rewind --show          # opt in to foreground presentation
  omi-ctl ui quiet                        # park rendered windows out of the way
  omi-ctl navigate settings rewind       # Settings page, Rewind sub-section
  omi-ctl wait-ready && omi-ctl navigate memories
  omi-ctl actions                        # discover available actions
  omi-ctl action refresh_all_data
  omi-ctl action toggle_transcription enabled=false

Env: OMI_AUTOMATION_PORT (default 47777), OMI_AUTOMATION_TOKEN, OMI_AUTOMATION_TOKEN_FILE
EOF
    ;;
esac