feat: implement server-wide management actions and enhance TUI dashboard

dev-scripts/_common.sh

@@ -50,9 +50,64 @@ read_pid() {
   [[ -f "$f" ]] && cat "$f" || echo 0
 }
 
+# Map agent ID to its config path by scanning agent directories.
+config_path_for() {
+  local target_id="$1"
+  for cfg in agents/*/config.yaml; do
+    [[ -f "$cfg" ]] || continue
+    local id
+    id=$(grep -m1 '^  id:' "$cfg" | awk '{print $2}')
+    if [[ "$id" == "$target_id" ]]; then
+      echo "$cfg"
+      return
+    fi
+  done
+}
+
+# Find all PIDs of launcher processes for a given agent ID.
+# Searches for the actual config path in the process command line.
+# Returns newline-separated PIDs (may be empty).
+find_agent_pids() {
+  local id="$1"
+  local cfg; cfg="$(config_path_for "$id")"
+  if [[ -z "$cfg" ]]; then
+    return
+  fi
+  pgrep -f "launcher.*-c.*${cfg}" 2>/dev/null || true
+}
+
 is_running() {
-  local pid; pid="$(read_pid "$1")"
-  [[ "$pid" -gt 0 ]] && kill -0 "$pid" 2>/dev/null
+  local id="$1"
+
+  # First check PID file
+  local pid; pid="$(read_pid "$id")"
+  if [[ "$pid" -gt 0 ]] && kill -0 "$pid" 2>/dev/null; then
+    return 0
+  fi
+
+  # PID file is stale or missing — search for actual processes
+  local pids; pids="$(find_agent_pids "$id")"
+  if [[ -n "$pids" ]]; then
+    # Update PID file with the first found process
+    local first_pid; first_pid="$(echo "$pids" | head -1)"
+    echo "$first_pid" > "$(pid_file "$id")"
+    return 0
+  fi
+
+  # Truly not running — clean up stale PID file
+  [[ "$pid" -gt 0 ]] && rm -f "$(pid_file "$id")"
+  return 1
+}
+
+# Count how many instances of an agent are running.
+count_instances() {
+  local id="$1"
+  local pids; pids="$(find_agent_pids "$id")"
+  if [[ -z "$pids" ]]; then
+    echo 0
+  else
+    echo "$pids" | wc -l
+  fi
 }
 
 agent_status() {

dev-scripts/ps.sh

@@ -26,6 +26,7 @@ while IFS='|' read -r id _version _enabled _desc _cfg; do
   fi
 
   pid="$(read_pid "$id")"
+  instance_count="$(count_instances "$id")"
   ((found++)) || true
 
   # Uptime: calcular desde el inicio del proceso
@@ -78,6 +79,12 @@ while IFS='|' read -r id _version _enabled _desc _cfg; do
   printf "%-22s  ${GRN}%-8s${RST}  %-12s  %-10s  %-8s  %s\n" \
     "$id" "$pid" "$uptime" "$mem" "${cpu_pct}%" "$log_size"
 
+  # Warn about duplicate instances
+  if [[ "$instance_count" -gt 1 ]]; then
+    printf "  ${RED}⚠ WARNING: %d instances running!${RST} PIDs: %s\n" \
+      "$instance_count" "$(find_agent_pids "$id" | tr '\n' ' ')"
+  fi
+
 done < <(list_agents_raw)
 
 if [[ "$found" -eq 0 ]]; then

dev-scripts/server.sh

@@ -55,11 +55,14 @@ case "$CMD" in
 
     killed=0
     for id in "${agents[@]}"; do
-      pid="$(read_pid "$id")"
-      if [[ "$pid" -gt 0 ]] && kill -0 "$pid" 2>/dev/null; then
-        kill -9 "$pid" 2>/dev/null || true
+      all_pids="$(find_agent_pids "$id")"
+      if [[ -n "$all_pids" ]]; then
+        cnt="$(echo "$all_pids" | wc -l)"
+        for p in $all_pids; do
+          kill -9 "$p" 2>/dev/null || true
+        done
         rm -f "$(pid_file "$id")"
-        ok "$id  killed  (PID $pid)"
+        ok "$id  killed  ($cnt instance(s), PIDs: $(echo $all_pids | tr '\n' ' '))"
         ((killed++)) || true
       else
         dim "  $id  (no estaba corriendo)"

dev-scripts/start.sh

@@ -14,11 +14,30 @@ start_agent() {
   local id="$1" cfg="$2"
   local log; log="$(log_file "$id")"
   local pid_f; pid_f="$(pid_file "$id")"
+  local bin="$REPO_ROOT/bin/launcher"
+
+  # Check for duplicate instances already running
+  local existing; existing="$(count_instances "$id")"
+  if [[ "$existing" -gt 0 ]]; then
+    warn "$id  already has $existing instance(s) running (orphan processes?)"
+    warn "  Run ./dev-scripts/stop.sh $id first to clean up"
+    return 1
+  fi
 
   info "Iniciando $id..."
 
-  # Lanza el launcher en background, desacoplado del terminal
-  nohup "$GO" run -tags goolm ./cmd/launcher -c "$cfg" --log-level "${LOG_LEVEL:-info}" \
+  # Build the binary first to avoid go run wrapper PID issues
+  if [[ ! -x "$bin" ]] || [[ "$(find ./cmd/launcher -newer "$bin" 2>/dev/null | head -1)" ]]; then
+    info "Compilando launcher..."
+    mkdir -p "$(dirname "$bin")"
+    "$GO" build -tags goolm -o "$bin" ./cmd/launcher || {
+      fail "$id  error de compilación — revisa el código"
+      return 1
+    }
+  fi
+
+  # Launch the compiled binary directly (no go run wrapper)
+  nohup "$bin" -c "$cfg" --log-level "${LOG_LEVEL:-info}" \
     >> "$log" 2>&1 &
 
   local pid=$!

dev-scripts/stop.sh

@@ -18,23 +18,37 @@ while IFS='|' read -r id _version _enabled _desc _cfg; do
     continue
   fi
 
-  local_pid="$(read_pid "$id")"
-  kill -TERM "$local_pid" 2>/dev/null || true
+  # Kill ALL instances, not just the one in the PID file
+  all_pids="$(find_agent_pids "$id")"
+  instance_count="$(echo "$all_pids" | grep -c . 2>/dev/null || echo 0)"
 
-  # Espera hasta 5s a que muera limpiamente
+  if [[ "$instance_count" -gt 1 ]]; then
+    warn "$id  has $instance_count instances running — stopping all"
+  fi
+
+  # Send SIGTERM to all instances
+  for p in $all_pids; do
+    kill -TERM "$p" 2>/dev/null || true
+  done
+
+  # Wait up to 5s for graceful shutdown
   for _ in {1..10}; do
-    kill -0 "$local_pid" 2>/dev/null || break
+    remaining="$(find_agent_pids "$id")"
+    [[ -z "$remaining" ]] && break
     sleep 0.5
   done
 
-  # SIGKILL si todavía sigue vivo
-  if kill -0 "$local_pid" 2>/dev/null; then
+  # SIGKILL any survivors
+  survivors="$(find_agent_pids "$id")"
+  if [[ -n "$survivors" ]]; then
     warn "$id  no respondió a SIGTERM, enviando SIGKILL..."
-    kill -9 "$local_pid" 2>/dev/null || true
+    for p in $survivors; do
+      kill -9 "$p" 2>/dev/null || true
+    done
   fi
 
   rm -f "$(pid_file "$id")"
-  ok "$id  detenido  (PID $local_pid)"
+  ok "$id  detenido  ($instance_count instance(s) stopped)"
   ((stopped++)) || true
 
 done < <(list_agents_raw)