Expose DuckLM runtime status

scripts/duck-mtp.sh

@@ -53,6 +53,7 @@ Commands:
   stop        Stop DuckLM API and managed MTP llama-server
   restart     Stop and start the whole local DuckLM stack
   status      Print process and HTTP health status
+              Use "status --probe" to include live model/vector checks
   logs        Show DuckLM API and llama-server logs; use --follow/-f and --lines N
   help        Show this help
 
@@ -169,6 +170,20 @@ stop_stack() {
 }
 
 status_stack() {
+  local probe=0
+  while [[ $# -gt 0 ]]; do
+    case "$1" in
+      --probe)
+        probe=1
+        shift
+        ;;
+      *)
+        echo "Unknown status argument: $1" >&2
+        return 2
+        ;;
+    esac
+  done
+
   local rc=0
   if api_is_running; then
     local pid
@@ -176,6 +191,7 @@ status_stack() {
     echo "DuckLM API running: pid=${pid}"
     if command -v curl >/dev/null 2>&1 && curl --noproxy "*" -fsS "${API_URL}/health" >/dev/null 2>&1; then
       echo "DuckLM API health: ok (${API_URL})"
+      print_runtime_status "${probe}"
     else
       echo "DuckLM API health: not ready (${API_URL})"
     fi
@@ -197,6 +213,55 @@ status_stack() {
   return "${rc}"
 }
 
+print_runtime_status() {
+  local probe="${1:-0}"
+  local status_url="${API_URL}/v1/status"
+  if [[ "${probe}" == "1" ]]; then
+    status_url="${status_url}?probe=true"
+  fi
+  local payload
+  if ! payload="$(curl --noproxy "*" -fsS "${status_url}" 2>/dev/null)"; then
+    echo "DuckLM runtime status: unavailable (${status_url})"
+    return 0
+  fi
+
+  local python_bin="${DUCK_PYTHON_BIN:-${ROOT_DIR}/.venv/bin/python}"
+  if [[ ! -x "${python_bin}" ]]; then
+    python_bin="python3"
+  fi
+  if ! command -v "${python_bin}" >/dev/null 2>&1 && [[ ! -x "${python_bin}" ]]; then
+    echo "DuckLM runtime status: ${status_url}"
+    return 0
+  fi
+
+  printf '%s' "${payload}" | "${python_bin}" -c '
+import json
+import sys
+
+data = json.load(sys.stdin)
+models = data.get("models", {})
+roles = sorted((models.get("roles") or {}).keys())
+services = data.get("services", {})
+llama = services.get("llama", {})
+vector = services.get("vector_memory", {})
+
+def service_line(name, service):
+    if not service.get("probed"):
+        return f"{name}: not probed"
+    ok = "ok" if service.get("ok") else "failed"
+    error = service.get("error")
+    return f"{name}: {ok}" + (f" ({error})" if error else "")
+
+print("DuckLM runtime:")
+print("  workspace: {}".format(data.get("workspace")))
+print("  db: {}".format(data.get("db_path")))
+print("  model endpoints: {}".format(len(models.get("endpoints") or [])))
+print("  roles: {}".format(", ".join(roles)))
+print("  {}".format(service_line("llama", llama)))
+print("  {}".format(service_line("vector memory", vector)))
+'
+}
+
 logs_stack() {
   local follow=0
   local lines=100
@@ -246,7 +311,8 @@ case "${ACTION}" in
     start_stack
     ;;
   status)
-    status_stack
+    shift || true
+    status_stack "$@"
     ;;
   logs)
     logs_stack "$@"

scripts/duck.sh

@@ -53,6 +53,7 @@ Commands:
   stop        Stop DuckLM API and managed llama-server
   restart     Stop and start the whole local DuckLM stack
   status      Print process and HTTP health status
+              Use "status --probe" to include live model/vector checks
   logs        Show DuckLM API and llama-server logs; use --follow/-f and --lines N
   help        Show this help
 
@@ -169,6 +170,20 @@ stop_stack() {
 }
 
 status_stack() {
+  local probe=0
+  while [[ $# -gt 0 ]]; do
+    case "$1" in
+      --probe)
+        probe=1
+        shift
+        ;;
+      *)
+        echo "Unknown status argument: $1" >&2
+        return 2
+        ;;
+    esac
+  done
+
   local rc=0
   if api_is_running; then
     local pid
@@ -176,6 +191,7 @@ status_stack() {
     echo "DuckLM API running: pid=${pid}"
     if command -v curl >/dev/null 2>&1 && curl --noproxy "*" -fsS "${API_URL}/health" >/dev/null 2>&1; then
       echo "DuckLM API health: ok (${API_URL})"
+      print_runtime_status "${probe}"
     else
       echo "DuckLM API health: not ready (${API_URL})"
     fi
@@ -197,6 +213,55 @@ status_stack() {
   return "${rc}"
 }
 
+print_runtime_status() {
+  local probe="${1:-0}"
+  local status_url="${API_URL}/v1/status"
+  if [[ "${probe}" == "1" ]]; then
+    status_url="${status_url}?probe=true"
+  fi
+  local payload
+  if ! payload="$(curl --noproxy "*" -fsS "${status_url}" 2>/dev/null)"; then
+    echo "DuckLM runtime status: unavailable (${status_url})"
+    return 0
+  fi
+
+  local python_bin="${DUCK_PYTHON_BIN:-${ROOT_DIR}/.venv/bin/python}"
+  if [[ ! -x "${python_bin}" ]]; then
+    python_bin="python3"
+  fi
+  if ! command -v "${python_bin}" >/dev/null 2>&1 && [[ ! -x "${python_bin}" ]]; then
+    echo "DuckLM runtime status: ${status_url}"
+    return 0
+  fi
+
+  printf '%s' "${payload}" | "${python_bin}" -c '
+import json
+import sys
+
+data = json.load(sys.stdin)
+models = data.get("models", {})
+roles = sorted((models.get("roles") or {}).keys())
+services = data.get("services", {})
+llama = services.get("llama", {})
+vector = services.get("vector_memory", {})
+
+def service_line(name, service):
+    if not service.get("probed"):
+        return f"{name}: not probed"
+    ok = "ok" if service.get("ok") else "failed"
+    error = service.get("error")
+    return f"{name}: {ok}" + (f" ({error})" if error else "")
+
+print("DuckLM runtime:")
+print("  workspace: {}".format(data.get("workspace")))
+print("  db: {}".format(data.get("db_path")))
+print("  model endpoints: {}".format(len(models.get("endpoints") or [])))
+print("  roles: {}".format(", ".join(roles)))
+print("  {}".format(service_line("llama", llama)))
+print("  {}".format(service_line("vector memory", vector)))
+'
+}
+
 logs_stack() {
   local follow=0
   local lines=100
@@ -246,7 +311,8 @@ case "${ACTION}" in
     start_stack
     ;;
   status)
-    status_stack
+    shift || true
+    status_stack "$@"
     ;;
   logs)
     logs_stack "$@"