#!/usr/bin/env bash
#
# opencode-builder.sh
#
# Script-driven orchestrator for the auto-agents system.
# Queries the OpenCode server fleet, calculates worker pool gaps,
# launches one-shot supervisors sequentially to spawn new workers,
# and runs periodic health checks directly.
#
# Signal handling:
#   Ctrl+C once  — finishes the in-flight request, then stops the loop
#   Ctrl+C twice — (within 2 seconds) kills the request and exits immediately
#

set -uo pipefail

# ── Configuration (override via environment) ──────────────────────────────
PORT="${OPENCODE_PORT:-4096}"
HOST="${OPENCODE_HOST:-127.0.0.1}"
BASE="http://${HOST}:${PORT}"
HEALTH_TIMEOUT=60
MAX_IDLE_PER_MINUTE=10

# Worker pool sizing
total_max_workers="${CA_MAX_PARALLEL_WORKERS:-4}"
# Calculate per-type max workers (same logic as auto-agents)
merge_max_workers=$(awk "BEGIN {print int((${total_max_workers} + 7) / 8)}")
imp_max_workers="${total_max_workers}"
rev_max_workers=$(awk "BEGIN {print int((${total_max_workers} + 1) / 2)}")

# Health check intervals (seconds)
DOOM_CHECK_INTERVAL=900   # 15 minutes
DOOM_AGE_THRESHOLD=1800   # 30 minutes
CONTINUE_BURST_WINDOW=60  # seconds
CONTINUE_BURST_LIMIT=5
SLEEP_WHEN_IDLE=30        # seconds
SUPERVISOR_TIMEOUT=300    # 5 minutes max for a one-shot supervisor

# Script paths
SCRIPT_DIR="/app/.opencode/skills/auto-agents-system/scripts"
SESSION_LIST="npx --yes tsx ${SCRIPT_DIR}/session_list.ts"
SESSION_FIND_PREFIX="npx --yes tsx ${SCRIPT_DIR}/session_find_by_prefix.ts"
SESSION_START="npx --yes tsx ${SCRIPT_DIR}/session_start.ts"
SESSION_MESSAGES="npx --yes tsx ${SCRIPT_DIR}/session_messages.ts"
SESSION_DELETE="npx --yes tsx ${SCRIPT_DIR}/session_delete.ts"
SESSION_STOP="npx --yes tsx ${SCRIPT_DIR}/session_stop.ts"

# ── Internal state ────────────────────────────────────────────────────────
SERVER_PID=""
CURL_PID=""
STOP_LOOP=false
LAST_SIGINT=0
OWN_SERVER=false
N=0
IDLE_EVENTS_FILE=""

# In-memory tracking for health checks
declare -A QUESTION_COUNT
declare -A CONTINUE_TIMES
LAST_DOOM_CHECK=0

# ── Helpers ───────────────────────────────────────────────────────────────
die() { printf "ERROR: %s\n" "$*" >&2; exit 1; }
log() { printf "[%s] %s\n" "$(date +%H:%M:%S)" "$*"; }

check_opencode_running() {
  if curl -sf "${BASE}/global/health" >/dev/null 2>&1; then
    return 0
  else
    return 1
  fi
}

cleanup() {
  if [[ -n "$IDLE_EVENTS_FILE" ]]; then
    rm -f "$IDLE_EVENTS_FILE" 2>/dev/null || true
  fi
  if $OWN_SERVER && [[ -n "$SERVER_PID" ]] && kill -0 "$SERVER_PID" 2>/dev/null; then
    log "Stopping opencode server (PID $SERVER_PID)..."
    kill "$SERVER_PID" 2>/dev/null
    wait "$SERVER_PID" 2>/dev/null || true
  elif ! $OWN_SERVER; then
    log "Leaving existing OpenCode server running (not managed by this script)"
  fi
}

handle_sigint() {
  local now
  now=$(date +%s)

  # Second Ctrl+C within 2 seconds → force-quit
  if $STOP_LOOP && (( now - LAST_SIGINT <= 2 )); then
    printf "\nForce quit!\n"
    [[ -n "$CURL_PID" ]] && kill "$CURL_PID" 2>/dev/null || true
    cleanup
    exit 130
  fi

  # First Ctrl+C → graceful stop after current request
  STOP_LOOP=true
  LAST_SIGINT=$now
  printf "\nWill stop after the current request completes. Ctrl+C again within 2s to force-quit.\n"
}

trap handle_sigint INT
trap cleanup EXIT

# ── Idle-event tracking ───────────────────────────────────────────────────
init_idle_tracking() {
  IDLE_EVENTS_FILE="/tmp/opencode-builder-idle-events-$$.txt"
  : > "$IDLE_EVENTS_FILE" 2>/dev/null || true
}

record_idle_event() {
  local now
  now=$(date +%s)
  echo "$now" >> "$IDLE_EVENTS_FILE"
}

count_idle_events_last_minute() {
  local now cutoff
  now=$(date +%s)
  cutoff=$((now - 60))
  if [[ -f "$IDLE_EVENTS_FILE" ]]; then
    awk -v c="$cutoff" '$1 > c {count++} END {print count+0}' "$IDLE_EVENTS_FILE" 2>/dev/null
  else
    echo 0
  fi
}

# ── Session lifecycle helpers ─────────────────────────────────────────────
query_all_sessions() {
  $SESSION_LIST --server "${BASE}" 2>/dev/null
}

find_sessions_by_prefix() {
  local prefix="$1"
  local status_filter="${2:-all}"
  $SESSION_FIND_PREFIX --server "${BASE}" --prefix "$prefix" --exclude-supervisor --status "$status_filter" 2>/dev/null
}

create_session() {
  local title="$1"
  curl -sf -X POST "${BASE}/session" \
    -H "Content-Type: application/json" \
    -d "{\"title\":\"${title}\"}" 2>/dev/null
}

delete_session() {
  local sid="$1"
  $SESSION_DELETE --server "${BASE}" --session-id "$sid" --force >/dev/null 2>&1 || true
}

stop_session() {
  local sid="$1"
  $SESSION_STOP --server "${BASE}" --session-id "$sid" >/dev/null 2>&1 || true
}

# Send a message to a session and block until the LLM responds.
# Captures the assistant's response to a temp file.
send_message_capture() {
  local sid="$1"
  local body="$2"
  local out_file="$3"

  ( trap '' INT
    exec curl -sf -X POST "${BASE}/session/${sid}/message" \
      -H "Content-Type: application/json" \
      -d "$body" \
      -w "\n"
  ) > "$out_file" 2>/dev/null &
  CURL_PID=$!

  while kill -0 "$CURL_PID" 2>/dev/null; do
    wait "$CURL_PID" 2>/dev/null || true
  done
  CURL_PID=""
}

# Poll GET /message until we see an assistant response or timeout
get_messages_poll() {
  local sid="$1"
  local limit="${2:-10}"
  local timeout_secs="${3:-60}"
  local deadline
  deadline=$(($(date +%s) + timeout_secs))

  while true; do
    local msgs
    msgs=$($SESSION_MESSAGES --server "${BASE}" --session-id "$sid" --limit "$limit" 2>/dev/null)
    if [[ -n "$msgs" ]]; then
      # Check if there's an assistant message
      local has_assistant
      has_assistant=$(echo "$msgs" | jq -r 'map(select(.info.role == "assistant")) | length')
      if [[ "$has_assistant" -gt 0 ]]; then
        echo "$msgs"
        return 0
      fi
    fi
    if [[ $(date +%s) -ge $deadline ]]; then
      echo "[]"
      return 1
    fi
    sleep 2
  done
}

# Extract the last assistant text from a messages JSON array
extract_last_assistant_text() {
  local msgs="$1"
  echo "$msgs" | jq -r '[.[] | select(.info.role == "assistant")] | last | if . == null then "" else (.parts // [] | map(select(.type == "text")) | .[-1].text // "") end'
}

# Extract a JSON array from text (look for [ ... ] block)
extract_json_array() {
  local text="$1"
  # Try to find a JSON array between [ and ]
  echo "$text" | grep -oP '\[\s*[^\]]*\]' | tail -n1
}

# ── Question handling ─────────────────────────────────────────────────────
get_pending_questions() {
  curl -sf "${BASE}/question" 2>/dev/null
}

answer_question() {
  local qid="$1"
  local text="$2"
  curl -sf -X POST "${BASE}/question/${qid}/answer" \
    -H "Content-Type: application/json" \
    -d "{\"body\":\"${text}\"}" >/dev/null 2>&1
}

reject_question() {
  local qid="$1"
  curl -sf -X POST "${BASE}/question/${qid}/reject" >/dev/null 2>&1
}

# ── Supervisor launcher ───────────────────────────────────────────────────
# Creates a temporary session, sends the supervisor prompt, blocks for the
# response, extracts the JSON array of launched work numbers, and deletes
# the temporary session.
launch_supervisor() {
  local agent_name="$1"
  local tag="$2"
  local prompt_text="$3"
  local session_title="[${tag}] ${agent_name}"
  local tmp_prompt="/tmp/builder-prompt-${tag}-$$.txt"
  local tmp_response="/tmp/builder-response-${tag}-$$.txt"

  log "Creating supervisor session: ${session_title}"
  local session_json
  session_json=$(create_session "$session_title")
  local sid
  sid=$(echo "$session_json" | jq -r '.id')
  if [[ -z "$sid" || "$sid" == "null" ]]; then
    log "Failed to create session for ${tag}"
    return 1
  fi

  # Write prompt to temp file
  printf '%s' "$prompt_text" > "$tmp_prompt"

  # Launch the agent asynchronously via prompt_async first
  log "Launching ${agent_name} in session ${sid}..."
  local start_body
  start_body=$(jq -nc --arg a "$agent_name" --rawfile t "$tmp_prompt" '{agent: $a, parts: [{type: "text", text: $t}]}')
  curl -sf -X POST "${BASE}/session/${sid}/prompt_async" \
    -H "Content-Type: application/json" \
    -d "$start_body" >/dev/null 2>&1

  # Wait for the session to become idle (supervisor finished)
  local wait_deadline
  wait_deadline=$(($(date +%s) + SUPERVISOR_TIMEOUT))
  local is_idle=false
  while [[ $(date +%s) -lt $wait_deadline ]]; do
    local status_json
    status_json=$(curl -sf "${BASE}/session/status" 2>/dev/null | jq -r --arg sid "$sid" '.[$sid].type // "idle"')
    if [[ "$status_json" == "idle" ]]; then
      is_idle=true
      break
    fi
    sleep 3
  done

  if ! $is_idle; then
    log "Supervisor ${tag} timed out after ${SUPERVISOR_TIMEOUT}s"
    delete_session "$sid"
    rm -f "$tmp_prompt" "$tmp_response"
    return 1
  fi

  # Fetch messages and extract the supervisor's response
  local msgs
  msgs=$($SESSION_MESSAGES --server "${BASE}" --session-id "$sid" --limit 20 2>/dev/null)
  local response_text
  response_text=$(extract_last_assistant_text "$msgs")

  # Try to extract a JSON array from the response
  local launched_array
  launched_array=$(extract_json_array "$response_text")

  log "Supervisor ${tag} finished. Launched: ${launched_array:-none}"

  # Clean up temp session
  delete_session "$sid"
  rm -f "$tmp_prompt" "$tmp_response"

  # Return the extracted array
  echo "$launched_array"
}

# ── Fleet analysis ────────────────────────────────────────────────────────
# Returns the list of busy worker tags for a given prefix, one per line.
get_busy_worker_tags() {
  local prefix="$1"
  find_sessions_by_prefix "$prefix" "busy" | jq -r '.[].tag // empty'
}

# Extract the PR/Issue number from a worker tag.
# Tags look like: AUTO-IMP-PR-42, AUTO-IMP-ISSUE-42, AUTO-MRG-PR-42, AUTO-REV-PR-42
extract_work_number() {
  local tag="$1"
  local num
  num=$(echo "$tag" | grep -oP '(?<=(-PR-|-ISSUE-))\d+$')
  echo "$num"
}

# ── Duplicate worker guard ────────────────────────────────────────────────
deduplicate_workers() {
  local prefix="$1"
  local workers_json
  workers_json=$(find_sessions_by_prefix "$prefix" "busy")
  # Group by tag, keep only the most recently active for each tag
  local duplicates
  duplicates=$(echo "$workers_json" | jq -r '
    group_by(.tag)
    | map(select(length > 1))
    | map(sort_by(.last_active) | reverse | .[1:])
    | flatten
    | .[].id
  ')
  if [[ -n "$duplicates" ]]; then
    while IFS= read -r dup_id; do
      [[ -n "$dup_id" ]] || continue
      log "Duplicate worker guard: deleting older session ${dup_id}"
      delete_session "$dup_id"
    done <<< "$duplicates"
  fi
}

# ── Disk cleanup ──────────────────────────────────────────────────────────
cleanup_orphan_tmp_dirs() {
  local prefix="$1"
  local active_tags
  active_tags=$(find_sessions_by_prefix "$prefix" "all" | jq -r '.[].tag // empty' | sort -u)

  # Look for /tmp directories that match the agent pattern for this prefix
  # Expected pattern: /tmp/<agent_name>-<identifier>-<timestamp>/
  # We can't know the exact agent name, so we look for directories with numbers
  # that might be identifiers, and cross-check against active tags.
  local tag_pattern="${prefix}"
  while IFS= read -r -d '' dir; do
    [[ -d "$dir" ]] || continue
    # Extract any numeric identifier from the directory basename
    local dir_id
    dir_id=$(basename "$dir" | grep -oP '\d+' | tail -n1)
    if [[ -n "$dir_id" ]]; then
      # Check if any active tag contains this number
      local is_active=false
      while IFS= read -r atag; do
        [[ -n "$atag" ]] || continue
        if [[ "$atag" == *"${dir_id}"* ]]; then
          is_active=true
          break
        fi
      done <<< "$active_tags"
      if ! $is_active; then
        log "Disk cleanup: removing orphan ${dir}"
        rm -rf "$dir"
      fi
    fi
  done < <(find /tmp -maxdepth 1 -type d -name "*${tag_pattern}*" -print0 2>/dev/null)
}

# ── Health checks ─────────────────────────────────────────────────────────
run_health_cycle() {
  local now
  now=$(date +%s)
  local did_something=false

  # 1) Question handling
  local questions_json
  questions_json=$(get_pending_questions)
  if [[ -n "$questions_json" && "$questions_json" != "[]" && "$questions_json" != "null" ]]; then
    local qids sids
    qids=$(echo "$questions_json" | jq -r '.[].id')
    sids=$(echo "$questions_json" | jq -r '.[].sessionID')
    while IFS= read -r qid && IFS= read -r sid <&3; do
      [[ -n "$qid" ]] || continue
      [[ -n "$sid" ]] || continue
      local cnt="${QUESTION_COUNT[$sid]:-0}"
      if [[ "$cnt" -eq 0 ]]; then
        log "Question from ${sid} (first time) — answering with anti-question text"
        answer_question "$qid" "you are never under any circumstances to ask questions or pass along questions from subagents, use your best judgement and do not ask questions again"
        QUESTION_COUNT[$sid]=1
      else
        log "Question from ${sid} (repeat) — deleting session"
        delete_session "$sid"
        unset QUESTION_COUNT[$sid]
      fi
      did_something=true
    done <<< "$qids" 3<<< "$sids"
  fi

  # 2) Idle worker check
  local all_prefixes=("AUTO-IMP" "AUTO-MRG" "AUTO-REV")
  for prefix in "${all_prefixes[@]}"; do
    local idle_workers
    idle_workers=$(find_sessions_by_prefix "$prefix" "idle" | jq -r '.[].id // empty')
    while IFS= read -r sid; do
      [[ -n "$sid" ]] || continue
      local msgs
      msgs=$($SESSION_MESSAGES --server "${BASE}" --session-id "$sid" --limit 10 2>/dev/null)
      if [[ -z "$msgs" || "$msgs" == "[]" ]]; then
        log "Idle worker ${sid} has no messages — deleting"
        delete_session "$sid"
        did_something=true
        continue
      fi
      # Check if it appears partially done (last assistant message suggests more work)
      local last_text
      last_text=$(extract_last_assistant_text "$msgs")
      # Heuristic: if the last text contains completion signals, delete it
      if echo "$last_text" | grep -qiE '(done|finished|complete|exiting|all tests pass|committed|merged|review submitted)'; then
        log "Idle worker ${sid} appears finished — deleting"
        delete_session "$sid"
        did_something=true
        continue
      fi
      # Otherwise, send continue and track
      local now_ms
      now_ms=$(date +%s)
      local times="${CONTINUE_TIMES[$sid]:-}"
      # Keep only timestamps within the burst window
      local cutoff=$((now_ms - CONTINUE_BURST_WINDOW))
      local new_times=""
      local count=0
      if [[ -n "$times" ]]; then
        while IFS= read -r ts; do
          [[ -n "$ts" ]] || continue
          if [[ "$ts" -gt "$cutoff" ]]; then
            new_times="${new_times}${ts}\n"
            ((count++))
          fi
        done <<< "$times"
      fi
      new_times="${new_times}${now_ms}\n"
      ((count++))
      CONTINUE_TIMES[$sid]="$new_times"
      if [[ "$count" -gt $CONTINUE_BURST_LIMIT ]]; then
        log "Idle worker ${sid} received ${count} continues in ${CONTINUE_BURST_WINDOW}s — deleting"
        delete_session "$sid"
        unset CONTINUE_TIMES[$sid]
        did_something=true
      else
        log "Idle worker ${sid} — sending continue"
        local cont_body
        cont_body=$(jq -nc '{agent: "", parts: [{type: "text", text: "continue"}]}')
        send_message_capture "$sid" "$cont_body" "/tmp/continue-${sid}.txt"
        did_something=true
      fi
    done <<< "$idle_workers"
  done

  # 3) Doom-loop check (every 15 minutes)
  if [[ $((now - LAST_DOOM_CHECK)) -ge $DOOM_CHECK_INTERVAL ]]; then
    LAST_DOOM_CHECK=$now
    for prefix in "${all_prefixes[@]}"; do
      local busy_workers
      busy_workers=$(find_sessions_by_prefix "$prefix" "busy" | jq -r '.[] | select(.last_active < ('"$now"' * 1000 - '"$((DOOM_AGE_THRESHOLD * 1000))"')) | .id // empty')
      while IFS= read -r sid; do
        [[ -n "$sid" ]] || continue
        log "Doom-loop check for ${sid} (busy >30 min)"
        local eval_tag="BUILDER-HEALTH-${sid: -6}"
        local eval_title="[${eval_tag}] worker-health-evaluator"
        local eval_prompt="session_id: ${sid}

Evaluate the health of this session."
        local eval_session_json
        eval_session_json=$(create_session "$eval_title")
        local eval_sid
        eval_sid=$(echo "$eval_session_json" | jq -r '.id')
        if [[ -n "$eval_sid" && "$eval_sid" != "null" ]]; then
          local eval_tmp="/tmp/builder-prompt-${eval_tag}-$$.txt"
          printf '%s' "$eval_prompt" > "$eval_tmp"
          local eval_body
          eval_body=$(jq -nc --arg a "worker-health-evaluator" --rawfile t "$eval_tmp" '{agent: $a, parts: [{type: "text", text: $t}]}')
          curl -sf -X POST "${BASE}/session/${eval_sid}/prompt_async" -H "Content-Type: application/json" -d "$eval_body" >/dev/null 2>&1
          local eval_deadline
          eval_deadline=$(($(date +%s) + 120))
          while [[ $(date +%s) -lt $eval_deadline ]]; do
            local eval_status
            eval_status=$(curl -sf "${BASE}/session/status" 2>/dev/null | jq -r --arg sid "$eval_sid" '.[$sid].type // "idle"')
            [[ "$eval_status" == "idle" ]] && break
            sleep 2
          done
          local eval_msgs
          eval_msgs=$($SESSION_MESSAGES --server "${BASE}" --session-id "$eval_sid" --limit 10 2>/dev/null)
          local eval_text
          eval_text=$(extract_last_assistant_text "$eval_msgs")
          delete_session "$eval_sid"
          rm -f "$eval_tmp"
          local is_unrecoverable
          is_unrecoverable=$(echo "$eval_text" | jq -r '.unrecoverable // false' 2>/dev/null || echo "false")
          if [[ "$is_unrecoverable" == "true" ]]; then
            log "Session ${sid} is unrecoverable — deleting"
            delete_session "$sid"
            did_something=true
          fi
        fi
      done <<< "$busy_workers"
    done
  fi

  if $did_something; then
    return 0
  else
    return 1
  fi
}

# ── Preflight checks ─────────────────────────────────────────────────────
for cmd in curl jq opencode npx; do
  command -v "$cmd" &>/dev/null || die "'$cmd' is required but not found in PATH"
done

# ── Start the opencode server ─────────────────────────────────────────────
if check_opencode_running; then
  echo "┌─────────────────────────────────────────────────────────────────┐"
  echo "│ ⚠️  WARNING: OpenCode server already running on ${HOST}:${PORT}"
  echo "│ 🔗 Connecting to existing server instead of starting new one"
  echo "│ 💡 The script will NOT manage this server's lifecycle"
  echo "└─────────────────────────────────────────────────────────────────┘"
  echo ""
  log "Using existing OpenCode server at ${HOST}:${PORT}"
  OWN_SERVER=false
else
  log "Starting opencode server on ${HOST}:${PORT}..."
  opencode serve --port "$PORT" --hostname "$HOST" &
  SERVER_PID=$!
  OWN_SERVER=true

  log "Waiting for server to become healthy (up to ${HEALTH_TIMEOUT}s)..."
  healthy=false
  for (( i = 1; i <= HEALTH_TIMEOUT; i++ )); do
    if curl -sf "${BASE}/global/health" >/dev/null 2>&1; then
      healthy=true
      break
    fi
    kill -0 "$SERVER_PID" 2>/dev/null || die "Server process exited unexpectedly"
    sleep 1
  done
  $healthy || die "Server did not become healthy within ${HEALTH_TIMEOUT}s"
  log "Server is ready."
fi

# ── Main orchestration loop ───────────────────────────────────────────────
log "Starting orchestration loop."
log "Worker pools: merge=${merge_max_workers} imp=${imp_max_workers} rev=${rev_max_workers}"

while ! $STOP_LOOP; do
  N=$((N + 1))

  # Check idle-event rate in the last minute
  idle_count=$(count_idle_events_last_minute)
  if (( idle_count > MAX_IDLE_PER_MINUTE )); then
    log "── iteration #${N} ── TOO MANY IDLE EVENTS (${idle_count}) — skipping cycle"
    sleep 5
    continue
  fi

  log "── iteration #${N} ──"
  local did_anything=false

  # ── Step 1: Query fleet ───────────────────────────────────────────────
  local all_sessions_json
  all_sessions_json=$(query_all_sessions)
  if [[ -z "$all_sessions_json" ]]; then
    log "No sessions found."
  fi

  # ── Step 2: Per-type analysis ─────────────────────────────────────────
  # Type definitions: prefix, supervisor agent name, max workers
  local types=(
    "AUTO-IMP:implementation-supervisor:${imp_max_workers}"
    "AUTO-MRG:pr-merge-supervisor:${merge_max_workers}"
    "AUTO-REV:pr-review-supervisor:${rev_max_workers}"
  )

  for type_spec in "${types[@]}"; do
    IFS=':' read -r prefix supervisor_name max <<< "$type_spec"

    # Duplicate guard
    deduplicate_workers "$prefix"

    # Disk cleanup
    cleanup_orphan_tmp_dirs "$prefix"

    # Count busy workers and collect active work numbers
    local busy_tags
    busy_tags=$(get_busy_worker_tags "$prefix")
    local busy_count=0
    local exclude_numbers=""
    while IFS= read -r tag; do
      [[ -n "$tag" ]] || continue
      ((busy_count++))
      local num
      num=$(extract_work_number "$tag")
      if [[ -n "$num" ]]; then
        exclude_numbers="${exclude_numbers}${num},"
      fi
    done <<< "$busy_tags"
    # Remove trailing comma
    exclude_numbers="${exclude_numbers%,}"

    local available_slots=$((max - busy_count))
    if [[ "$available_slots" -le 0 ]]; then
      log "Pool ${prefix} is full (${busy_count}/${max})"
      continue
    fi

    log "Pool ${prefix}: ${busy_count}/${max} busy, spawning up to ${available_slots} workers"
    did_anything=true

    # Build supervisor prompt
    local prompt_text
    prompt_text=$(cat <<EOF
spawn_count: ${available_slots}
exclude_work_numbers: [${exclude_numbers}]

tag prefix: "${prefix}"
Max parallel workers: ${max}

Start processing and return the list of PR/Issue numbers you launched workers for.
EOF
)

    # Launch supervisor sequentially, blocking
    local launched_json
    launched_json=$(launch_supervisor "$supervisor_name" "BUILDER-${prefix}" "$prompt_text")
    if [[ -n "$launched_json" && "$launched_json" != "[]" ]]; then
      log "Launched workers for: ${launched_json}"
    fi
  done

  # ── Step 3: Health & maintenance cycle ────────────────────────────────
  if run_health_cycle; then
    did_anything=true
  fi

  # ── Step 4: Idle sleep if nothing happened ────────────────────────────
  if ! $did_anything; then
    log "Nothing to do — sleeping ${SLEEP_WHEN_IDLE}s"
    record_idle_event
    sleep "$SLEEP_WHEN_IDLE"
  fi
done

log "Loop stopped. Shutting down."
exit 0