NexusOS/management/nexus-cli.sh

#!/usr/bin/env bash

# Load nvm so npm/node resolve to the managed version
export NVM_DIR="$HOME/.nvm"
# shellcheck source=/dev/null
[ -s "$NVM_DIR/nvm.sh" ] && \. "$NVM_DIR/nvm.sh"

NEXUS_ROOT="$HOME/nexus-core"
PID_DIR="$NEXUS_ROOT/runtime/pids"
LOG_DIR="$NEXUS_ROOT/runtime"
PYTHON="$NEXUS_ROOT/Promethean/bin/python3"
FRONTEND_DIR="$NEXUS_ROOT/interface/web"

mkdir -p "$PID_DIR" "$LOG_DIR"

BACKEND_PID="$PID_DIR/backend.pid"
MEMORY_PID="$PID_DIR/memory.pid"
FRONTEND_PID="$PID_DIR/frontend.pid"

BACKEND_LOG="$LOG_DIR/backend.log"
MEMORY_LOG="$LOG_DIR/memory.log"
FRONTEND_LOG="$LOG_DIR/frontend.log"

# -----------------------------
# INTERNAL HELPERS
# -----------------------------

_launch() {
    # _launch <name> <pid_file> <log_file> <work_dir> <cmd> [args...]
    local name="$1" pid_file="$2" log_file="$3" work_dir="$4"
    shift 4

    if [ -f "$pid_file" ] && kill -0 "$(cat "$pid_file")" 2>/dev/null; then
        echo "  $name is already running (PID $(cat "$pid_file"))"
        return 0
    fi

    : > "$log_file"
    ( cd "$work_dir" && exec "$@" ) >> "$log_file" 2>&1 &
    echo $! > "$pid_file"
}

_check() {
    # _check <label> <pid_file> <log_file>
    # label should be the full service name, e.g. "NEXUS BACKEND SERVICE"
    local label="$1" pid_file="$2" log_file="$3"
    local pid
    pid=$(cat "$pid_file" 2>/dev/null)
    if [ -n "$pid" ] && kill -0 "$pid" 2>/dev/null; then
        echo "$label STARTED"
        return 0
    else
        echo "$label FAILED TO START"
        [ -s "$log_file" ] && tail -8 "$log_file" | sed 's/^/  /'
        rm -f "$pid_file"
        return 1
    fi
}

_stop() {
    # _stop <label> <pid_file>
    local label="$1" pid_file="$2"
    if [ -f "$pid_file" ]; then
        local pid
        pid=$(cat "$pid_file")
        if kill -0 "$pid" 2>/dev/null; then
            kill "$pid" 2>/dev/null
            echo "$label STOPPED"
        else
            echo "$label NOT RUNNING"
        fi
        rm -f "$pid_file"
    else
        echo "$label NOT RUNNING"
    fi
}

_status() {
    # _status <name> <pid_file>
    local name="$1" pid_file="$2"
    if [ -f "$pid_file" ] && kill -0 "$(cat "$pid_file")" 2>/dev/null; then
        echo "  $name: RUNNING (PID $(cat "$pid_file"))"
    else
        echo "  $name: STOPPED"
    fi
}

# -----------------------------
# KILL COMMAND
# -----------------------------

kill_services() {
    echo "Force-killing all Nexus processes..."
    local ports=(8000 8001 5173)
    local names=("SYNAPSE" "MEMORY" "INTERFACE")
    for i in "${!ports[@]}"; do
        port="${ports[$i]}"
        name="${names[$i]}"
        if fuser -k "${port}/tcp" 2>/dev/null; then
            echo "  KILLED:      ${name} (:${port})"
        else
            echo "  NOT RUNNING: ${name} (:${port})"
        fi
    done
    pkill -9 -f "uvicorn synapse" 2>/dev/null || true
    pkill -9 -f "npm run dev"     2>/dev/null || true
    rm -f "$PID_DIR"/*.pid
    echo "Done."
}

# -----------------------------
# START COMMANDS
# -----------------------------

start_memory() {
    _launch "NEXUS MEMORY SERVICE" "$MEMORY_PID" "$MEMORY_LOG" "$NEXUS_ROOT" \
        "$PYTHON" -m uvicorn synapse.memory.service:app --host 0.0.0.0 --port 8001 --reload
    sleep 3
    _check "NEXUS MEMORY SERVICE" "$MEMORY_PID" "$MEMORY_LOG"
}

start_backend() {
    _launch "NEXUS BACKEND SERVICE" "$BACKEND_PID" "$BACKEND_LOG" "$NEXUS_ROOT" \
        "$PYTHON" -m uvicorn synapse.main:sio_app --host 0.0.0.0 --port 8000 --reload
    sleep 3
    _check "NEXUS BACKEND SERVICE" "$BACKEND_PID" "$BACKEND_LOG"
}

start_frontend() {
    _launch "NEXUS FRONTEND SERVICE" "$FRONTEND_PID" "$FRONTEND_LOG" "$FRONTEND_DIR" \
        npm run dev -- --host 0.0.0.0
    sleep 5
    _check "NEXUS FRONTEND SERVICE" "$FRONTEND_PID" "$FRONTEND_LOG"
}

start_all() {
    # Memory + backend launch in parallel, then single wait for both
    _launch "NEXUS MEMORY SERVICE" "$MEMORY_PID" "$MEMORY_LOG" "$NEXUS_ROOT" \
        "$PYTHON" -m uvicorn synapse.memory.service:app --host 0.0.0.0 --port 8001 --reload
    _launch "NEXUS BACKEND SERVICE" "$BACKEND_PID" "$BACKEND_LOG" "$NEXUS_ROOT" \
        "$PYTHON" -m uvicorn synapse.main:sio_app --host 0.0.0.0 --port 8000 --reload
    sleep 3
    _check "NEXUS MEMORY SERVICE" "$MEMORY_PID" "$MEMORY_LOG"
    _check "NEXUS BACKEND SERVICE" "$BACKEND_PID" "$BACKEND_LOG"
    # Frontend last (Vite takes a moment longer)
    _launch "NEXUS FRONTEND SERVICE" "$FRONTEND_PID" "$FRONTEND_LOG" "$FRONTEND_DIR" \
        npm run dev -- --host 0.0.0.0
    sleep 5
    _check "NEXUS FRONTEND SERVICE" "$FRONTEND_PID" "$FRONTEND_LOG"
}

# -----------------------------
# STOP COMMANDS
# -----------------------------

stop_memory() {
    _stop "NEXUS MEMORY SERVICE" "$MEMORY_PID"
}

stop_backend() {
    _stop "NEXUS BACKEND SERVICE" "$BACKEND_PID"
}

stop_frontend() {
    _stop "NEXUS FRONTEND SERVICE" "$FRONTEND_PID"
}

stop_all() {
    stop_memory
    stop_backend
    stop_frontend
}

# -----------------------------
# STATUS COMMAND
# -----------------------------

status_services() {
    echo "Nexus Service Status:"
    echo
    _status "Memory service" "$MEMORY_PID"
    _status "Backend      " "$BACKEND_PID"
    _status "Frontend     " "$FRONTEND_PID"
}

# -----------------------------
# LOGS COMMAND
# -----------------------------

show_logs() {
    case "$1" in
        --frontend|-f)
            echo "=== FRONTEND LOGS ==="
            tail -n 50 "$FRONTEND_LOG"
            ;;
        --backend|-b)
            echo "=== BACKEND LOGS ==="
            tail -n 50 "$BACKEND_LOG"
            ;;
        --memory|-m)
            echo "=== MEMORY SERVICE LOGS ==="
            tail -n 50 "$MEMORY_LOG"
            ;;
        ""|all)
            echo "=== MEMORY SERVICE LOGS ==="
            tail -n 30 "$MEMORY_LOG"
            echo
            echo "=== BACKEND LOGS ==="
            tail -n 30 "$BACKEND_LOG"
            echo
            echo "=== FRONTEND LOGS ==="
            tail -n 30 "$FRONTEND_LOG"
            ;;
        *)
            echo "Unknown logs target: '$1'"
            echo "Usage: ncp logs [frontend|backend|memory|all]"
            ;;
    esac
}

# -----------------------------
# DOCTOR COMMAND
# -----------------------------

doctor() {
    echo "Running Nexus Diagnostics..."
    echo

    echo "Checking directories..."
    [ -d "$NEXUS_ROOT" ] && echo "  ✔ Nexus root found" || echo "  ✘ Missing Nexus root"
    [ -d "$FRONTEND_DIR" ] && echo "  ✔ Frontend directory found" || echo "  ✘ Missing frontend directory"

    echo
    echo "Checking Python venv..."
    [ -x "$PYTHON" ] && echo "  ✔ Promethean venv found ($($PYTHON --version 2>&1))" || echo "  ✘ Promethean venv missing at $PYTHON"

    echo
    echo "Checking Node & npm..."
    command -v node >/dev/null && echo "  ✔ Node installed ($(node -v))" || echo "  ✘ Node missing"
    command -v npm >/dev/null && echo "  ✔ npm installed ($(npm -v))" || echo "  ✘ npm missing"

    echo
    echo "Checking Uvicorn..."
    "$PYTHON" -m uvicorn --version >/dev/null 2>&1 && echo "  ✔ Uvicorn installed" || echo "  ✘ Uvicorn missing"

    echo
    status_services
}

# -----------------------------
# HELP COMMAND
# -----------------------------

show_help() {
    echo "Nexus Command Tree"
    echo
    echo "Usage: ncp <command> [options]"
    echo
    echo "Commands:"
    echo "  chat                  Open the interactive chat TUI"
    echo "  playbook              Browse and edit playbooks"
    echo "  panel                 Launch the Nexus Control Panel"
    echo
    echo "  start                 Start ALL Nexus services (memory + backend + frontend)"
    echo "  start --memory,  -m   Start only the memory service"
    echo "  start --frontend,-f   Start only the frontend"
    echo "  start --backend, -b   Start only the backend"
    echo
    echo "  stop                  Stop ALL Nexus services"
    echo "  stop --memory,  -m    Stop only the memory service"
    echo "  stop --frontend,-f    Stop only the frontend"
    echo "  stop --backend, -b    Stop only the backend"
    echo
    echo "  kill                  Force-kill all Nexus processes by port (nuclear option)
  refresh               Restart all services"
    echo
    echo "  status                Show service status"
    echo "  logs                  Show logs for all services"
    echo "  logs --memory,  -m    Memory service logs"
    echo "  logs --frontend,-f    Frontend logs"
    echo "  logs --backend, -b    Backend logs"
    echo
    echo "  doctor                Run Nexus diagnostics"
    echo "  update                Update Nexus dependencies"
    echo "  clean                 Remove runtime files and caches"
    echo
    echo "  models"
    echo "    list                List installed models"
    echo "    available           Show models available to install"
    echo "    install <name>      Pull a model into Nexus"
    echo
    echo "  backup                Backup vital Nexus files to router (smart/incremental)"
    echo "  backup -f, full       Full backup of all project files"
    echo "  restore               Restore vital Nexus files from router backup"
    echo "  restore -f, full      Full restore of all project files"
    echo
    echo "  help, -h              Show this help message"
}

# -----------------------------
# UPDATE COMMAND
# -----------------------------

update_nexus() {
    echo "Updating Project Nexus..."
    echo

    cd "$NEXUS_ROOT" || exit 1

    if [ -d ".git" ]; then
        echo "Pulling latest changes from Git..."
        git pull
    else
        echo "No Git repository found — skipping git pull."
    fi

    echo
    echo "Updating backend Python dependencies..."
    if [ -f "amd_requirements.txt" ]; then
        "$PYTHON" -m pip install -r amd_requirements.txt
    else
        echo "No amd_requirements.txt found."
    fi

    echo
    echo "Updating frontend dependencies..."
    if [ -d "$FRONTEND_DIR" ]; then
        cd "$FRONTEND_DIR"
        npm install
    else
        echo "Frontend directory missing — skipping npm install."
    fi

    echo
    echo "Running post-update diagnostics..."
    doctor
}

# -----------------------------
# BACKUP / RESTORE COMMANDS
# -----------------------------

ensure_router_reachable() {
    echo "Checking router connectivity..."
    if ssh -q -o BatchMode=yes -o ConnectTimeout=5 router exit 2>/dev/null; then
        echo "Router is reachable."
        return 0
    fi

    echo "Router unreachable. Activating WireGuard (wgs_client)..."
    nmcli connection up wgs_client

    echo "Waiting for VPN to establish..."
    sleep 5

    if ssh -q -o BatchMode=yes -o ConnectTimeout=10 router exit 2>/dev/null; then
        echo "Router reachable via VPN."
        return 0
    fi

    echo "Router still unreachable after VPN activation."
    return 1
}

backup_nexus() {
    ensure_router_reachable || return 1
    echo "Backing up Nexus to router..."
    bash "$NEXUS_ROOT/bin/backup-smart.sh"
}

backup_nexus_full() {
    ensure_router_reachable || return 1
    echo "Full backup of Nexus to router..."
    bash "$NEXUS_ROOT/bin/backup.sh"
}

restore_nexus() {
    local script="$1"
    ensure_router_reachable || return 1
    echo "WARNING: This will overwrite local files with the remote backup."
    printf "Continue? [y/N] "
    read -r confirm
    if [[ "$confirm" != "y" && "$confirm" != "Y" ]]; then
        echo "Restore cancelled."
        return
    fi
    bash "$NEXUS_ROOT/bin/$script"
}

# -----------------------------
# MODELS COMMANDS
# -----------------------------

OLLAMA_BIN="$NEXUS_ROOT/ollama/bin/ollama"
OLLAMA_MODELS_DIR="$NEXUS_ROOT/models"

models_list() {
    if ! curl -sf http://localhost:11434/api/tags > /dev/null 2>&1; then
        echo "Ollama is not running. Start the backend first with: ncp start -b"
        return 1
    fi
    echo "Installed models:"
    echo
    curl -s http://localhost:11434/api/tags | "$PYTHON" -c "
import sys, json
data = json.load(sys.stdin)
models = data.get('models', [])
if not models:
    print('  No models installed.')
else:
    for m in models:
        size_mb = m['size'] // 1024 // 1024
        size = f'{size_mb / 1024:.1f} GB' if size_mb >= 1024 else f'{size_mb} MB'
        print(f'  {m[\"name\"]:<35} {size}')
"
}

models_available() {
    echo "Available models (via Ollama library):"
    echo
    printf "  %-30s %-10s %s\n" "MODEL" "SIZE" "DESCRIPTION"
    printf "  %-30s %-10s %s\n" "-----" "----" "-----------"
    printf "  %-30s %-10s %s\n" "gemma3:1b"       "~815 MB"  "Google Gemma 3 — fast, lightweight"
    printf "  %-30s %-10s %s\n" "gemma3:4b"       "~3.3 GB"  "Google Gemma 3 — balanced"
    printf "  %-30s %-10s %s\n" "gemma3:12b"      "~8.1 GB"  "Google Gemma 3 — capable"
    printf "  %-30s %-10s %s\n" "llama3.2:1b"     "~1.3 GB"  "Meta Llama 3.2 — fast, lightweight"
    printf "  %-30s %-10s %s\n" "llama3.2:3b"     "~2.0 GB"  "Meta Llama 3.2 — compact, capable"
    printf "  %-30s %-10s %s\n" "llama3.1:8b"     "~4.7 GB"  "Meta Llama 3.1 — strong general use"
    printf "  %-30s %-10s %s\n" "mistral:latest"  "~4.1 GB"  "Mistral 7B — solid all-rounder"
    printf "  %-30s %-10s %s\n" "mistral-nemo"    "~7.1 GB"  "Mistral Nemo 12B — strong reasoning"
    printf "  %-30s %-10s %s\n" "qwen2.5:3b"      "~2.0 GB"  "Alibaba Qwen 2.5 — great at code"
    printf "  %-30s %-10s %s\n" "qwen2.5:7b"      "~4.7 GB"  "Alibaba Qwen 2.5 — strong coder"
    printf "  %-30s %-10s %s\n" "phi4-mini"       "~2.5 GB"  "Microsoft Phi-4 Mini — efficient"
    printf "  %-30s %-10s %s\n" "phi4:14b"        "~8.9 GB"  "Microsoft Phi-4 — strong reasoning"
    printf "  %-30s %-10s %s\n" "deepseek-r1:7b"  "~4.7 GB"  "DeepSeek R1 — reasoning model"
    printf "  %-30s %-10s %s\n" "deepseek-r1:14b" "~9.0 GB"  "DeepSeek R1 — strong reasoning"
    printf "  %-30s %-10s %s\n" "codellama:7b"    "~3.8 GB"  "Meta Code Llama — code focused"
    printf "  %-30s %-10s %s\n" "nomic-embed-text" "~274 MB" "Text embeddings model"
    echo
    echo "Install any model with: ncp models install <model>"
    echo "Browse more at: https://ollama.com/library"
}

models_install() {
    local model="$1"
    if [ -z "$model" ]; then
        echo "Usage: ncp models install <model>"
        echo "Run 'ncp models available' to see options."
        return 1
    fi
    if [ ! -x "$OLLAMA_BIN" ]; then
        echo "Ollama binary not found at $OLLAMA_BIN"
        return 1
    fi
    echo "Pulling '$model' into $OLLAMA_MODELS_DIR ..."
    echo
    OLLAMA_MODELS="$OLLAMA_MODELS_DIR" "$OLLAMA_BIN" pull "$model"
    echo
    echo "Done. Run 'ncp models list' to verify."
}

# -----------------------------
# CLEAN COMMAND
# -----------------------------

clean_nexus() {
    echo "Cleaning Nexus runtime files..."
    echo

    echo "Removing PID files..."
    rm -f "$PID_DIR"/*.pid

    echo "Removing logs..."
    rm -f "$LOG_DIR"/*.log

    echo "Removing Python cache..."
    find "$NEXUS_ROOT" -type d -name "__pycache__" -exec rm -rf {} + 2>/dev/null

    echo "Removing Node/Vite cache..."
    find "$FRONTEND_DIR" -type d -name ".vite" -exec rm -rf {} + 2>/dev/null

    echo
    echo "Cleanup complete."
}

# -----------------------------
# COMMAND TREE ROUTER
# -----------------------------

subcommand="$1"
shift

case "$subcommand" in

    panel)    python3 "$NEXUS_ROOT/management/controlpanel.py" ;;
    chat)     "$PYTHON" "$NEXUS_ROOT/management/nexus-chat.py" ;;
    playbook) "$PYTHON" "$NEXUS_ROOT/management/nexus-playbook.py" ;;

    start)
        case "$1" in
            --memory|-m)   start_memory ;;
            --frontend|-f) start_frontend ;;
            --backend|-b)  start_backend ;;
            ""|all)        start_all ;;
            *)             show_help ;;
        esac
        ;;

    stop)
        case "$1" in
            --memory|-m)   stop_memory ;;
            --frontend|-f) stop_frontend ;;
            --backend|-b)  stop_backend ;;
            ""|all)        stop_all ;;
            *)             show_help ;;
        esac
        ;;

    refresh) stop_all && start_all ;;
    kill)    kill_services ;;
    status)  status_services ;;
    logs)    show_logs "$1" ;;
    doctor)  doctor ;;
    update)  update_nexus ;;
    clean)   clean_nexus ;;
    help|-h|"") show_help ;;

    nvidia-reqs) "$PYTHON" "$NEXUS_ROOT/bin/gen-nvidia-reqs.py" ;;

    backup)
        case "$1" in
            -f|full) backup_nexus_full ;;
            "")      backup_nexus ;;
            *)       echo "Usage: ncp backup [-f|full]" ;;
        esac
        ;;

    restore)
        case "$1" in
            -f|full) restore_nexus restore-full.sh ;;
            "")      restore_nexus restore.sh ;;
            *)       echo "Usage: ncp restore [-f|full]" ;;
        esac
        ;;

    models)
        case "$1" in
            list)             models_list ;;
            available|search) models_available ;;
            install)          models_install "$2" ;;
            *)                echo "Usage: ncp models <list|available|install <model>>" ;;
        esac
        ;;

    *)
        echo "Unknown Nexus command: '$subcommand'"
        echo "Use 'ncp help' for available commands."
        ;;
esac