fix(server): replaces text logo with proper logo

docs: adds images

README.md

@@ -1,8 +1,10 @@
-# Ngino
+# <img src="docs/images/Ngino_logo_symbol.png" width="25" alt="logo symbol"/> Ngino
 
-Ngino is a small outbound HTTP tunnel for running Ollama (or vLLM, etc.) on GPU workstations while exposing the API from a server that cannot reach those workstations directly.
+Ngino is a small outbound HTTP tunnel for running Ollama, llama.cpp, vLLM, etc. on GPU workstations while exposing the API from a server that cannot reach those workstations directly.
 
-The client opens and maintains a WebSocket connection to the server. The server accepts normal HTTP requests and forwards them through that WebSocket to the client. The client then calls a local upstream such as `http://localhost:11434` and streams the response back.
+<img src="docs/images/Ngino_logo_full.png" alt="logo symbol"/>
+
+The client opens and maintains a WebSocket connection to the server. The server accepts normal HTTP requests from users and forwards them through that WebSocket to the client. The client then calls a local upstream such as `http://localhost:11434` and streams the response back.
 
 <table>
 <tr>
@@ -15,8 +17,8 @@ The server provides
 - An API with
   - Authentication via user keys
   - Authorization (planned)
-- Load balancing (Scale your AI strategy horizontally!)
-- (Ollama-only) Model management (install, remove, load, unload models)
+- Load balancing (Scale your AI inferencing horizontally by adding more nodes!)
+- (Ollama-based) GGUF Model management (install, remove, load, unload models)
 - Client monitoring
   - Who is active
   - What models are running
@@ -129,6 +131,14 @@ Linux:
 ```bash
 sudo bash deploy/install-client.sh --server http://your-server:5050 --token "change-me"
 ```
+or using ollama models with llama.cpp backend using ROCm:
+```bash
+sudo bash deploy/install-client.sh   --server https://ai.domain.tld   --token "change-me"   --use-llama-cpp-via-docker   --use-ollama-models-path /usr/share/ollama/.ollama/models   --llama-cpp-docker-image ghcr.io/ggml-org/llama.cpp:server-rocm   --llama-cpp-base-port 8081   --llama-cpp-parallel 128
+```
+or using ollama models with llama.cpp backend using CUDA:
+```bash
+sudo bash deploy/install-client.sh   --server https://ai.domain.tld   --token "change-me"   --use-llama-cpp-via-docker   --use-ollama-models-path /usr/share/ollama/.ollama/models   --llama-cpp-docker-image ghcr.io/ggml-org/llama.cpp:server-cuda   --llama-cpp-base-port 8081   --llama-cpp-parallel 128
+```
 
 Options: `--server`, `--token` (required); `--client-id`, `--upstream`, `--install-dir`, `--service-name`, `--no-ollama` (optional). Missing required values are prompted interactively.
 
@@ -140,6 +150,15 @@ llama.cpp via Docker options (replaces Ollama for inferencing):
 | `--use-ollama-models-path <dir>` | Path to Ollama models directory (`manifests/blobs`); required with the flag above |
 | `--llama-cpp-docker-image <img>` | Docker image; defaults to auto-detected (rocm/cuda/cpu) |
 | `--llama-cpp-base-port <num>` | Base port for containers; defaults to `8081` |
+| `--llama-cpp-parallel <num>` | llama.cpp parallel slots per container; if unset, llama.cpp's own default is used (which is `1`) |
+| `--llama-cpp-fallback-cooldown <sec>` | Seconds before llama.cpp is retried after a failed container start; defaults to `180` |
+| `--log-dir <dir>` | Directory for log files; defaults to `<app dir>/Logs` |
+
+### llama.cpp fallback to Ollama
+
+Models are served via llama.cpp Docker containers. If a container cannot be started for a model (for example, the model's GGUF blob is incompatible with the llama.cpp build), the client falls back to the Ollama upstream for that model. Transient start failures are remembered for `--llama-cpp-fallback-cooldown` seconds (default 3 minutes) and then retried; a container that starts but exits before becoming ready marks the model as falling back until it is unloaded. `load`/`unload` model commands and on-demand request routing are all covered; a failed container start is detected quickly by watching the container state, and the container log tail is written to the client log to aid debugging.
+
+Note: some hybrid SSM/attention models (e.g. `qwen3.5-coder-next`) are converted by Ollama into a GGUF tensor layout that stock llama.cpp cannot load (`missing tensor 'blk.0.ssm_dt.bias'` and similar). Such models are served via the Ollama fallback above. If you want them to run on llama.cpp instead, use a Hugging Face-converted GGUF (e.g. `unsloth/Qwen3-Coder-Next-GGUF`) rather than the Ollama blob.
 
 The script ensures .NET 10 and Ollama are installed, builds the client self-contained, installs it to `/opt/Ngino-client`, and creates a systemd service (`Ngino-client`). Logs: `journalctl -u Ngino-client -f`.
 

deploy/install-client.sh

@@ -20,6 +20,9 @@ USE_LLAMA_CPP_VIA_DOCKER=false
 USE_OLLAMA_MODELS_PATH=""
 LLAMA_CPP_DOCKER_IMAGE=""
 LLAMA_CPP_BASE_PORT=""
+LLAMA_CPP_PARALLEL=""
+LLAMA_CPP_FALLBACK_COOLDOWN=""
+LOG_DIR=""
 
 # ── Colors ────────────────────────────────────────────────────────────────────
 RED='\033[0;31m'
@@ -57,13 +60,19 @@ Optional:
                        llama.cpp Docker image; defaults to auto-detected (rocm/cuda/cpu)
   --llama-cpp-base-port <num>
                        Base port for llama.cpp containers; defaults to 8081
+  --llama-cpp-parallel <num>
+                       llama.cpp parallel slots per container; if unset, llama.cpp's own default is used
+  --llama-cpp-fallback-cooldown <sec>
+                       Seconds before llama.cpp is retried after a failed container start; defaults to 180
+  --log-dir <dir>      Directory for log files; defaults to <install-dir>/Logs
   -h, --help           Show this help message
 
 Examples:
   $0 --server http://gpu-server:5050 --token "my-secret"
   $0 --server http://gpu-server:5050 --token "my-secret" --no-ollama
   $0 --server http://gpu-server:5050 --token "my-secret" \\
-     --use-llama-cpp-via-docker --use-ollama-models-path /usr/share/ollama/.ollama/models
+     --use-llama-cpp-via-docker --use-ollama-models-path /usr/share/ollama/.ollama/models \\
+     --llama-cpp-parallel 128
 EOF
     exit 0
 }
@@ -82,6 +91,9 @@ while [[ $# -gt 0 ]]; do
         --use-ollama-models-path)   USE_OLLAMA_MODELS_PATH="$2"; shift 2 ;;
         --llama-cpp-docker-image)   LLAMA_CPP_DOCKER_IMAGE="$2"; shift 2 ;;
         --llama-cpp-base-port)      LLAMA_CPP_BASE_PORT="$2"; shift 2 ;;
+        --llama-cpp-parallel)       LLAMA_CPP_PARALLEL="$2"; shift 2 ;;
+        --llama-cpp-fallback-cooldown) LLAMA_CPP_FALLBACK_COOLDOWN="$2"; shift 2 ;;
+        --log-dir)                  LOG_DIR="$2"; shift 2 ;;
         -h|--help)                  usage ;;
         *)                          die "Unknown option: $1" ;;
     esac
@@ -311,6 +323,15 @@ mkdir -p "$ENV_DIR"
         if [[ -n "$LLAMA_CPP_BASE_PORT" ]]; then
             printf 'NGINO_LLAMA_CPP_BASE_PORT=%s\n' "$LLAMA_CPP_BASE_PORT"
         fi
+        if [[ -n "$LLAMA_CPP_PARALLEL" ]]; then
+            printf 'NGINO_LLAMA_CPP_PARALLEL=%s\n' "$LLAMA_CPP_PARALLEL"
+        fi
+        if [[ -n "$LLAMA_CPP_FALLBACK_COOLDOWN" ]]; then
+            printf 'NGINO_LLAMA_CPP_FALLBACK_COOLDOWN_SECONDS=%s\n' "$LLAMA_CPP_FALLBACK_COOLDOWN"
+        fi
+    fi
+    if [[ -n "$LOG_DIR" ]]; then
+        printf 'NGINO_LOG_DIR=%s\n' "$LOG_DIR"
     fi
 } > "$ENV_DIR/env"
 chmod 600 "$ENV_DIR/env"

deploy/uninstall-client.ps1

@@ -0,0 +1,42 @@
+#Requires -Version 5.1
+#Requires -RunAsAdministrator
+
+[CmdletBinding()]
+param(
+    [string]$InstallDir = "$env:ProgramFiles\Ngino Client",
+    [string]$ServiceName = "NginoClient"
+)
+
+$ErrorActionPreference = "Stop"
+
+function Write-Info([string]$Message) { Write-Host "[INFO]  $Message" -ForegroundColor Green }
+
+if ($ServiceName -notmatch '^[A-Za-z0-9_.-]+$') { throw "ServiceName contains unsupported characters." }
+
+# ── Stop and delete the Windows service ───────────────────────────────────────
+$service = Get-Service -Name $ServiceName -ErrorAction SilentlyContinue
+if ($service) {
+    if ($service.Status -ne "Stopped") {
+        Write-Info "Stopping service $ServiceName..."
+        Stop-Service -Name $ServiceName -Force
+        $service.WaitForStatus("Stopped", [TimeSpan]::FromSeconds(30))
+    }
+    Write-Info "Deleting service $ServiceName..."
+    & sc.exe delete $ServiceName
+    if ($LASTEXITCODE -ne 0) { throw "Could not delete Windows service $ServiceName." }
+} else {
+    Write-Info "Service $ServiceName does not exist; skipping."
+}
+
+# ── Remove install directory ──────────────────────────────────────────────────
+if (Test-Path -LiteralPath $InstallDir) {
+    Write-Info "Removing install directory $InstallDir..."
+    Remove-Item -LiteralPath $InstallDir -Recurse -Force
+} else {
+    Write-Info "Install directory $InstallDir does not exist; skipping."
+}
+
+Write-Host ""
+Write-Info "Uninstall complete."
+Write-Host "  Service:     $ServiceName (stopped and deleted)"
+Write-Host "  Install dir: $InstallDir (removed)"

deploy/uninstall-client.sh

@@ -0,0 +1,95 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+# ── Defaults ──────────────────────────────────────────────────────────────────
+DEFAULT_INSTALL_DIR="/opt/ngino-client"
+DEFAULT_SERVICE_NAME="ngino-client"
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+
+INSTALL_DIR="$DEFAULT_INSTALL_DIR"
+SERVICE_NAME="$DEFAULT_SERVICE_NAME"
+
+# ── Colors ────────────────────────────────────────────────────────────────────
+RED='\033[0;31m'
+GREEN='\033[0;32m'
+YELLOW='\033[0;33m'
+NC='\033[0m'
+
+info()    { echo -e "${GREEN}[INFO]${NC}  $*"; }
+warn()    { echo -e "${YELLOW}[WARN]${NC}  $*"; }
+err()     { echo -e "${RED}[ERROR]${NC} $*" >&2; }
+die()     { err "$@"; exit 1; }
+
+# ── Usage ─────────────────────────────────────────────────────────────────────
+usage() {
+    cat <<EOF
+Usage: $0 [OPTIONS]
+
+Uninstalls the Ngino client systemd service and removes its files.
+Does NOT remove .NET SDK, Ollama, or Docker.
+
+Optional:
+  --install-dir <dir>  Install directory; defaults to $DEFAULT_INSTALL_DIR
+  --service-name <n>   systemd service name; defaults to $DEFAULT_SERVICE_NAME
+  -h, --help           Show this help message
+
+Examples:
+  $0
+  $0 --install-dir /opt/ngino-client --service-name ngino-client
+EOF
+    exit 0
+}
+
+# ── Argument parsing ──────────────────────────────────────────────────────────
+while [[ $# -gt 0 ]]; do
+    case "$1" in
+        --install-dir)              INSTALL_DIR="$2"; shift 2 ;;
+        --service-name)             SERVICE_NAME="$2"; shift 2 ;;
+        -h|--help)                  usage ;;
+        *)                          die "Unknown option: $1" ;;
+    esac
+done
+
+# ── Root check ────────────────────────────────────────────────────────────────
+if [[ $EUID -ne 0 ]]; then
+    die "This script must be run as root (or with sudo)."
+fi
+
+SERVICE_FILE="/etc/systemd/system/${SERVICE_NAME}.service"
+ENV_DIR="/etc/ngino-client"
+
+# ── Stop and disable the service ──────────────────────────────────────────────
+if systemctl list-unit-files "$SERVICE_NAME.service" &>/dev/null 2>&1; then
+    info "Stopping service $SERVICE_NAME..."
+    systemctl stop "$SERVICE_NAME" 2>/dev/null || true
+    info "Disabling service $SERVICE_NAME..."
+    systemctl disable "$SERVICE_NAME" 2>/dev/null || true
+fi
+
+# ── Remove service unit file ──────────────────────────────────────────────────
+if [[ -f "$SERVICE_FILE" ]]; then
+    info "Removing service unit file $SERVICE_FILE..."
+    rm -f "$SERVICE_FILE"
+fi
+
+systemctl daemon-reload
+
+# ── Remove install directory ──────────────────────────────────────────────────
+if [[ -d "$INSTALL_DIR" ]]; then
+    info "Removing install directory $INSTALL_DIR..."
+    rm -rf "$INSTALL_DIR"
+fi
+
+# ── Remove environment file ───────────────────────────────────────────────────
+if [[ -d "$ENV_DIR" ]]; then
+    info "Removing environment directory $ENV_DIR..."
+    rm -rf "$ENV_DIR"
+fi
+
+# ── Done ──────────────────────────────────────────────────────────────────────
+echo
+info "Uninstall complete."
+echo "  Service:     $SERVICE_NAME (stopped, disabled, unit removed)"
+echo "  Install dir: $INSTALL_DIR (removed)"
+echo "  Env dir:     $ENV_DIR (removed)"

docs/images/Ngino_logo.svg

@@ -0,0 +1,332 @@
+<?xml version="1.0" encoding="UTF-8" standalone="no"?>
+<svg
+   viewBox="0 0 1280 720"
+   role="img"
+   aria-labelledby="title desc"
+   version="1.1"
+   id="svg10"
+   sodipodi:docname="Ngino_logo.svg"
+   xml:space="preserve"
+   inkscape:version="1.4.3 (0d15f75042, 2025-12-25)"
+   inkscape:export-filename="Ngino_logo_full_dark.png"
+   inkscape:export-xdpi="125.532"
+   inkscape:export-ydpi="125.532"
+   xmlns:inkscape="http://www.inkscape.org/namespaces/inkscape"
+   xmlns:sodipodi="http://sodipodi.sourceforge.net/DTD/sodipodi-0.dtd"
+   xmlns:xlink="http://www.w3.org/1999/xlink"
+   xmlns="http://www.w3.org/2000/svg"
+   xmlns:svg="http://www.w3.org/2000/svg"
+   xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#"
+   xmlns:cc="http://creativecommons.org/ns#"
+   xmlns:dc="http://purl.org/dc/elements/1.1/"><sodipodi:namedview
+     id="namedview10"
+     pagecolor="#000000"
+     bordercolor="#000000"
+     borderopacity="1"
+     inkscape:showpageshadow="2"
+     inkscape:pageopacity="0.0"
+     inkscape:pagecheckerboard="0"
+     inkscape:deskcolor="#000000"
+     inkscape:zoom="0.7671875"
+     inkscape:cx="391.69043"
+     inkscape:cy="406.02851"
+     inkscape:window-width="2560"
+     inkscape:window-height="1412"
+     inkscape:window-x="0"
+     inkscape:window-y="0"
+     inkscape:window-maximized="1"
+     inkscape:current-layer="layer4"
+     inkscape:export-bgcolor="#00000000" /><title
+     id="title">Ngino logo</title><desc
+     id="desc">A fully vector llama outline with one input connection branching to three AI endpoints, beside the Ngino wordmark.</desc><defs
+     id="defs2"><linearGradient
+       id="swatch181"><stop
+         style="stop-color:#000000;stop-opacity:1;"
+         offset="0"
+         id="stop181" /></linearGradient><clipPath
+       clipPathUnits="userSpaceOnUse"
+       id="clipPath15"><rect
+         style="fill:none;stroke:#000000;stroke-linecap:butt;stroke-linejoin:bevel;paint-order:stroke markers fill;stop-color:#000000"
+         id="rect15-7"
+         width="45.034248"
+         height="38.955639"
+         x="262.48285"
+         y="121.00073" /></clipPath><linearGradient
+       id="swatch7"
+       inkscape:swatch="solid"><stop
+         style="stop-color:#000000;stop-opacity:1;"
+         offset="0"
+         id="stop7" /></linearGradient><linearGradient
+       id="linearGradient1"
+       inkscape:swatch="gradient"><stop
+         style="stop-color:#000000;stop-opacity:0;"
+         offset="0"
+         id="stop3" /><stop
+         style="stop-color:#000000;stop-opacity:0;"
+         offset="1"
+         id="stop4" /></linearGradient><linearGradient
+       id="routeGradient"
+       x1="95"
+       y1="0"
+       x2="430"
+       y2="0"
+       gradientUnits="userSpaceOnUse"><stop
+         offset="0"
+         stop-color="#0697b5"
+         id="stop1" /><stop
+         offset="1"
+         stop-color="#45d05f"
+         id="stop2" /></linearGradient><linearGradient
+       inkscape:collect="always"
+       xlink:href="#routeGradient"
+       id="linearGradient10"
+       gradientUnits="userSpaceOnUse"
+       x1="95"
+       y1="0"
+       x2="430"
+       y2="0"
+       gradientTransform="translate(50.835031)" /><linearGradient
+       inkscape:collect="always"
+       xlink:href="#routeGradient"
+       id="linearGradient24"
+       gradientUnits="userSpaceOnUse"
+       x1="95"
+       y1="0"
+       x2="430"
+       y2="0" /><linearGradient
+       inkscape:collect="always"
+       xlink:href="#routeGradient"
+       id="linearGradient25"
+       gradientUnits="userSpaceOnUse"
+       x1="95"
+       y1="0"
+       x2="430"
+       y2="0" /><linearGradient
+       inkscape:collect="always"
+       xlink:href="#routeGradient"
+       id="linearGradient26"
+       gradientUnits="userSpaceOnUse"
+       x1="95"
+       y1="0"
+       x2="430"
+       y2="0" /><linearGradient
+       inkscape:collect="always"
+       xlink:href="#routeGradient"
+       id="linearGradient27"
+       gradientUnits="userSpaceOnUse"
+       x1="95"
+       y1="0"
+       x2="430"
+       y2="0" /><linearGradient
+       inkscape:collect="always"
+       xlink:href="#routeGradient"
+       id="linearGradient28"
+       gradientUnits="userSpaceOnUse"
+       x1="95"
+       y1="0"
+       x2="430"
+       y2="0"
+       gradientTransform="translate(18.248473,-20.855397)" /><linearGradient
+       inkscape:collect="always"
+       xlink:href="#routeGradient"
+       id="linearGradient29"
+       gradientUnits="userSpaceOnUse"
+       x1="95"
+       y1="0"
+       x2="430"
+       y2="0"
+       gradientTransform="translate(18.248473,-20.855397)" /><linearGradient
+       inkscape:collect="always"
+       xlink:href="#routeGradient"
+       id="linearGradient30"
+       gradientUnits="userSpaceOnUse"
+       x1="95"
+       y1="0"
+       x2="430"
+       y2="0"
+       gradientTransform="translate(18.248473,-20.855397)" /><linearGradient
+       inkscape:collect="always"
+       xlink:href="#routeGradient"
+       id="linearGradient31"
+       gradientUnits="userSpaceOnUse"
+       x1="95"
+       y1="0"
+       x2="430"
+       y2="0"
+       gradientTransform="translate(18.248473,-20.855397)" /><filter
+       inkscape:collect="always"
+       style="color-interpolation-filters:sRGB"
+       id="filter215"
+       x="-0.0021624053"
+       y="-0.0047179751"
+       width="1.0043248"
+       height="1.009436"><feGaussianBlur
+         inkscape:collect="always"
+         stdDeviation="1.0812026"
+         id="feGaussianBlur215" /></filter><filter
+       inkscape:collect="always"
+       style="color-interpolation-filters:sRGB"
+       id="filter215-8"
+       x="-0.003844276"
+       y="-0.0074139609"
+       width="1.0076886"
+       height="1.0148279"><feGaussianBlur
+         inkscape:collect="always"
+         stdDeviation="1.0812026"
+         id="feGaussianBlur215-8" /></filter><filter
+       inkscape:collect="always"
+       style="color-interpolation-filters:sRGB"
+       id="filter215-6"
+       x="-0.0051897726"
+       y="-0.0047179751"
+       width="1.0103795"
+       height="1.009436"><feGaussianBlur
+         inkscape:collect="always"
+         stdDeviation="1.0812026"
+         id="feGaussianBlur215-0" /></filter></defs><!-- Transparent background; no embedded raster image. --><rect
+     style="display:inline;opacity:1;fill:#ffffff;fill-opacity:1;stroke:none;stroke-width:50;stroke-linecap:round;stroke-miterlimit:0.8;stroke-dasharray:none;stroke-opacity:1;filter:url(#filter215)"
+     id="rect1"
+     width="1200"
+     height="550"
+     x="26.069246"
+     y="20.855396"
+     inkscape:label="background_full"
+     rx="100"
+     inkscape:highlight-color="#000000"
+     inkscape:export-filename="Ngino_logo_full.png"
+     inkscape:export-xdpi="153.60001"
+     inkscape:export-ydpi="153.60001" /><g
+     inkscape:groupmode="layer"
+     id="layer1"
+     inkscape:label="text"
+     style="display:inline"
+     inkscape:export-filename="Ngino_logo_text.png"
+     inkscape:export-xdpi="125.532"
+     inkscape:export-ydpi="125.532"><rect
+       style="display:inline;fill:#ffffff;fill-opacity:1;stroke:none;stroke-width:50;stroke-linecap:round;stroke-miterlimit:0.8;stroke-dasharray:none;stroke-opacity:1;filter:url(#filter215-8)"
+       id="rect1-1"
+       width="675"
+       height="350"
+       x="458.96759"
+       y="152.75374"
+       inkscape:label="background_text"
+       rx="100"
+       inkscape:highlight-color="#000000"
+       inkscape:export-filename="Ngino_logo_full.png"
+       inkscape:export-xdpi="153.60001"
+       inkscape:export-ydpi="153.60001" /><text
+       xml:space="preserve"
+       style="font-style:normal;font-variant:normal;font-weight:normal;font-stretch:normal;font-size:37.3333px;line-height:1.25;font-family:Sans;-inkscape-font-specification:'Sans, Normal';font-variant-ligatures:normal;font-variant-caps:normal;font-variant-numeric:normal;font-variant-east-asian:normal;display:inline;fill:#000000;fill-opacity:1"
+       x="503.97342"
+       y="430.72964"
+       id="text2"
+       inkscape:label="sub"><tspan
+         sodipodi:role="line"
+         id="tspan2"
+         x="503.97342"
+         y="430.72964">R E V E R S E   P R O X Y   F O R  A I</tspan></text><text
+       xml:space="preserve"
+       style="font-weight:900;font-size:170.667px;line-height:1.25;font-family:Montserrat;-inkscape-font-specification:'Montserrat, Heavy';display:inline;fill:#000000;fill-opacity:1"
+       x="496.78931"
+       y="357.61456"
+       id="text1"
+       inkscape:label="main"
+       inkscape:export-filename="Ngino_logo_text.png"
+       inkscape:export-xdpi="125.532"
+       inkscape:export-ydpi="125.532"><tspan
+         sodipodi:role="line"
+         id="tspan1"
+         x="496.78931"
+         y="357.61456">Ngino</tspan></text></g><!-- Connection nodes --><!-- Wordmark --><rect
+     style="display:inline;fill:#ffffff;fill-opacity:1;stroke:none;stroke-width:50;stroke-linecap:round;stroke-miterlimit:0.8;stroke-dasharray:none;stroke-opacity:1;filter:url(#filter215-6)"
+     id="rect1-8"
+     width="500"
+     height="550"
+     x="28.825022"
+     y="21.104044"
+     inkscape:label="background_logo"
+     rx="100"
+     inkscape:highlight-color="#000000"
+     inkscape:export-filename="Ngino_logo_full.png"
+     inkscape:export-xdpi="153.60001"
+     inkscape:export-ydpi="153.60001" /><g
+     inkscape:groupmode="layer"
+     id="layer2"
+     inkscape:label="Logo"
+     style="display:inline"
+     inkscape:export-filename="Ngino_Logo_Symbol.png"
+     inkscape:export-xdpi="125.532"
+     inkscape:export-ydpi="125.532"><g
+       inkscape:groupmode="layer"
+       id="layer3"
+       inkscape:label="Body"
+       style="display:inline"><path
+         d="m 115.82078,398.64766 c -22.393081,-46.4277 -14.055,-69.1446 2.64154,-97.02036 C 143.97964,277.26884 147,275.30346 190,275.30346 h 56 c 36,0 53.48269,-18.60692 63.48269,-46.60692 6,-16 4,-43 5,-65 0,-20 6,-37.30346 16,-52.30346 l 0.39308,-26.965381 C 333.17923,71.427699 347,53 356,48 l 2,47 c 6,-19 17,-34 31,-40 l -1.69653,38.268839 c 13,3 22.69653,17.517311 28.69653,28.517311 18,3 29,6.17923 31,22.17923 0,14 -3.78615,18.08961 -16.78615,20.08961 l -37,7.21385 c 12,24 28.91039,76.76578 36.91039,108.76578 14,58 -6.24848,146.8554 -29.24848,202.8554 l -13,29 h -47 l -8,-35 c -4,-22 -23,-38 -48,-38 h -53.50305 c -27,0 -45,16 -49,40 l -6,33 h -47.12424 z"
+         id="path2"
+         sodipodi:nodetypes="ccsscccccccccccccccccsscccc"
+         style="display:inline;fill:none;stroke:url(#linearGradient27)"
+         stroke-width="13"
+         inkscape:label="body" /></g><g
+       inkscape:groupmode="layer"
+       id="layer4"
+       inkscape:label="Symbol"
+       style="display:inline"><g
+         fill="url(#routeGradient)"
+         id="g9"
+         style="display:inline;fill:url(#linearGradient27)"
+         transform="translate(18.248473,-20.855397)"
+         inkscape:label="Symbol_circles"><circle
+           cx="354"
+           cy="398"
+           r="15"
+           id="circle9"
+           style="fill:url(#linearGradient26)"
+           inkscape:label="ball_fork_down" /><circle
+           cx="354"
+           cy="356"
+           r="15"
+           id="circle8"
+           style="display:inline;fill:url(#linearGradient25)"
+           inkscape:label="ball_fork_mid" /><circle
+           cx="354"
+           cy="314"
+           r="15"
+           id="circle7"
+           style="display:inline;fill:url(#linearGradient24)"
+           inkscape:label="ball_fork_up" /><circle
+           cx="130.83504"
+           cy="356"
+           r="15"
+           id="circle6"
+           style="display:inline;fill:url(#linearGradient10)"
+           inkscape:label="ball_connector" /></g><g
+         fill="none"
+         stroke="url(#routeGradient)"
+         stroke-width="13"
+         stroke-linecap="round"
+         stroke-linejoin="round"
+         id="g6"
+         style="display:inline"
+         inkscape:label="Symbol_paths"><!-- Llama silhouette --><!-- One input, three routed endpoints --><path
+           d="m 254.24847,335.1446 c 29,0 42,42 78,42 h 40"
+           id="path6"
+           style="display:inline;stroke:url(#linearGradient31)"
+           inkscape:label="fork_down" /><path
+           d="m 254.24847,335.1446 h 118"
+           id="path5"
+           style="display:inline;stroke:url(#linearGradient28)"
+           inkscape:label="fork_mid" /><path
+           d="m 254.24847,335.1446 c 29,0 42,-42 78,-42 h 40"
+           id="path4"
+           style="display:inline;stroke:url(#linearGradient30)"
+           inkscape:label="fork_up" /><path
+           d="M 150.38696,335.1446 H 254.24847"
+           id="path3"
+           sodipodi:nodetypes="cc"
+           style="display:inline;stroke:url(#linearGradient29)"
+           inkscape:label="connector"
+           inkscape:export-filename="Ngino_logo_symbol.png"
+           inkscape:export-xdpi="153.60001"
+           inkscape:export-ydpi="153.60001" /></g></g></g><metadata
+     id="metadata215"><rdf:RDF><cc:Work
+         rdf:about=""><dc:title>Ngino logo</dc:title></cc:Work></rdf:RDF></metadata></svg>

src/Ngino.Client/ClientOptions.cs

@@ -28,6 +28,12 @@ internal sealed class ClientOptions
 
     public int LlamaCppBasePort { get; init; } = 8081;
 
+    public int? LlamaCppParallel { get; init; }
+
+    public TimeSpan LlamaCppFallbackCooldown { get; init; } = TimeSpan.FromMinutes(3);
+
+    public string? LogDirectory { get; init; }
+
     public Uri TunnelUri
     {
         get
@@ -68,7 +74,10 @@ internal sealed class ClientOptions
             UseLlamaCppViaDocker = ReadBool(values, false, "use-llama-cpp-via-docker", "NGINO_USE_LLAMA_CPP_VIA_DOCKER"),
             UseOllamaModelsPath = NormalizeDirectoryPath(Read(values, "use-ollama-models-path", "NGINO_USE_OLLAMA_MODELS_PATH")),
             LlamaCppDockerImage = Read(values, "llama-cpp-docker-image", "NGINO_LLAMA_CPP_DOCKER_IMAGE"),
-            LlamaCppBasePort = ReadInt(values, 8081, "llama-cpp-base-port", "NGINO_LLAMA_CPP_BASE_PORT")
+            LlamaCppBasePort = ReadInt(values, 8081, "llama-cpp-base-port", "NGINO_LLAMA_CPP_BASE_PORT"),
+            LlamaCppParallel = ReadOptionalInt(values, "llama-cpp-parallel", "NGINO_LLAMA_CPP_PARALLEL"),
+            LlamaCppFallbackCooldown = TimeSpan.FromSeconds(ReadInt(values, 180, "llama-cpp-fallback-cooldown", "NGINO_LLAMA_CPP_FALLBACK_COOLDOWN_SECONDS")),
+            LogDirectory = NormalizeDirectoryPath(Read(values, "log-dir", "NGINO_LOG_DIR"))
         };
     }
 
@@ -87,6 +96,10 @@ internal sealed class ClientOptions
           --use-ollama-models-path <dir> Path to Ollama models directory (manifests/blobs), required with --use-llama-cpp-via-docker
           --llama-cpp-docker-image <img> llama.cpp Docker image; defaults to auto-detected (rocm/cuda/cpu)
           --llama-cpp-base-port <num>    Base port for llama.cpp containers; defaults to 8081
+          --llama-cpp-parallel <num>     llama.cpp parallel slots per container; if unset, llama.cpp's own default is used
+          --llama-cpp-fallback-cooldown <sec>
+                                         Seconds before llama.cpp is retried after a failed container start; defaults to 180
+          --log-dir <dir>                Directory for log files; defaults to <app dir>/Logs
         """;
 
     private static Dictionary<string, string> ParseArgs(string[] args)
@@ -160,6 +173,12 @@ internal sealed class ClientOptions
         return int.TryParse(value, out var parsed) && parsed > 0 ? parsed : fallback;
     }
 
+    private static int? ReadOptionalInt(Dictionary<string, string> values, params string[] keys)
+    {
+        var value = Read(values, keys);
+        return int.TryParse(value, out var parsed) && parsed > 0 ? parsed : null;
+    }
+
     private static bool ReadBool(Dictionary<string, string> values, bool fallback, params string[] keys)
     {
         var value = Read(values, keys);

src/Ngino.Client/FileLogger.cs

@@ -0,0 +1,113 @@
+using System.Text;
+using Microsoft.Extensions.Logging;
+
+namespace Ngino.Client;
+
+internal sealed class FileLoggerProvider : ILoggerProvider
+{
+    private const long DefaultMaxFileSizeBytes = 5L * 1024 * 1024;
+    private const string LogFileName = "ngino-client.log";
+    private const string RotatedLogFileName = "ngino-client.log.1";
+
+    private readonly string _directory;
+    private readonly long _maxFileSizeBytes;
+    private readonly object _lock = new();
+    private StreamWriter _writer = null!;
+    private string _currentFile = null!;
+
+    public FileLoggerProvider(string directory, long maxFileSizeBytes = DefaultMaxFileSizeBytes)
+    {
+        _directory = directory;
+        _maxFileSizeBytes = maxFileSizeBytes;
+        Directory.CreateDirectory(directory);
+        OpenFile();
+    }
+
+    public string LogDirectory => _directory;
+
+    public ILogger CreateLogger(string categoryName) => new FileLogger(this, categoryName);
+
+    public void WriteLog(DateTime timestamp, LogLevel level, string category, string message)
+    {
+        lock (_lock)
+        {
+            var line = $"{timestamp:yyyy-MM-dd HH:mm:ss.fff} [{level}] {category}: {message}";
+            if (_writer.BaseStream.Length + line.Length + 2 > _maxFileSizeBytes)
+            {
+                RotateFile();
+            }
+
+            _writer.WriteLine(line);
+        }
+    }
+
+    public void Dispose()
+    {
+        lock (_lock)
+        {
+            _writer.Dispose();
+        }
+    }
+
+    private void OpenFile()
+    {
+        _currentFile = Path.Combine(_directory, LogFileName);
+        _writer = new StreamWriter(
+            new FileStream(_currentFile, FileMode.Append, FileAccess.Write, FileShare.ReadWrite),
+            new UTF8Encoding(encoderShouldEmitUTF8Identifier: false))
+        {
+            AutoFlush = true
+        };
+    }
+
+    private void RotateFile()
+    {
+        _writer.Dispose();
+
+        var rotatedFile = Path.Combine(_directory, RotatedLogFileName);
+        try
+        {
+            File.Delete(rotatedFile);
+            if (File.Exists(_currentFile))
+            {
+                File.Move(_currentFile, rotatedFile);
+            }
+        }
+        catch (IOException)
+        {
+        }
+        catch (UnauthorizedAccessException)
+        {
+        }
+
+        OpenFile();
+    }
+}
+
+internal sealed class FileLogger(FileLoggerProvider provider, string category) : ILogger
+{
+    public IDisposable? BeginScope<TState>(TState state) where TState : notnull => null;
+
+    public bool IsEnabled(LogLevel logLevel) => logLevel >= LogLevel.Trace;
+
+    public void Log<TState>(
+        LogLevel logLevel,
+        EventId eventId,
+        TState state,
+        Exception? exception,
+        Func<TState, Exception?, string> formatter)
+    {
+        if (!IsEnabled(logLevel))
+        {
+            return;
+        }
+
+        var message = formatter(state, exception);
+        if (exception is not null)
+        {
+            message += Environment.NewLine + exception;
+        }
+
+        provider.WriteLog(DateTime.Now, logLevel, category, message);
+    }
+}

src/Ngino.Client/LlamaCppManager.cs

@@ -1,5 +1,7 @@
 using System.Collections.Concurrent;
 using System.Diagnostics;
+using System.Net.Http;
+using System.Net.Sockets;
 using System.Text.Json;
 using System.Text.RegularExpressions;
 using Microsoft.Extensions.Logging;
@@ -13,24 +15,36 @@ internal sealed partial class LlamaCppManager : IAsyncDisposable
     private const int DefaultBasePort = 8081;
     private const string NginoContainerLabel = "ngino-llamacpp";
     private static readonly TimeSpan DockerTimeout = TimeSpan.FromSeconds(60);
+    private static readonly TimeSpan ContainerStartTimeout = TimeSpan.FromMinutes(5);
+    private static readonly TimeSpan DefaultFallbackCooldown = TimeSpan.FromMinutes(3);
 
     private readonly string _blobsPath;
     private readonly string _manifestsPath;
     private readonly string _dockerImage;
     private readonly int _basePort;
+    private readonly int? _parallel;
+    private readonly TimeSpan _fallbackCooldown;
     private readonly ILogger _logger;
     private readonly ConcurrentDictionary<string, int> _modelPorts = new(StringComparer.OrdinalIgnoreCase);
+    private readonly ConcurrentDictionary<int, byte> _reservedPorts = new();
+    private readonly ConcurrentDictionary<string, DateTime> _fallbackModels = new(StringComparer.OrdinalIgnoreCase);
+    private readonly ConcurrentDictionary<string, SemaphoreSlim> _modelStartLocks = new(StringComparer.OrdinalIgnoreCase);
+    private readonly object _portAllocationLock = new();
 
     public LlamaCppManager(
         string ollamaModelsPath,
         string? dockerImage,
         int? basePort,
-        ILogger? logger = null)
+        ILogger? logger = null,
+        TimeSpan fallbackCooldown = default,
+        int? parallel = null)
     {
         _manifestsPath = Path.Combine(ollamaModelsPath, "manifests");
         _blobsPath = Path.Combine(ollamaModelsPath, "blobs");
         _dockerImage = dockerImage ?? GetDefaultDockerImage();
         _basePort = basePort ?? DefaultBasePort;
+        _fallbackCooldown = fallbackCooldown > TimeSpan.Zero ? fallbackCooldown : DefaultFallbackCooldown;
+        _parallel = parallel is > 0 ? parallel : null;
         _logger = logger ?? NullLogger<LlamaCppManager>.Instance;
     }
 
@@ -75,6 +89,82 @@ internal sealed partial class LlamaCppManager : IAsyncDisposable
         return _modelPorts.ContainsKey(ollamaModelName);
     }
 
+    public void MarkModelAsFallback(string ollamaModelName)
+    {
+        _fallbackModels[ollamaModelName] = DateTime.UtcNow.Add(_fallbackCooldown);
+        _logger.LogWarning(
+            "Model {Model} will fall back to the Ollama upstream for {Cooldown} before llama.cpp is retried.",
+            ollamaModelName, _fallbackCooldown);
+    }
+
+    public void MarkModelAsPermanentFallback(string ollamaModelName)
+    {
+        _fallbackModels[ollamaModelName] = DateTime.MaxValue;
+        _logger.LogError(
+            "Model {Model} exited its llama.cpp container before becoming ready. It will fall back to the Ollama upstream until it is unloaded.",
+            ollamaModelName);
+    }
+
+    public bool IsModelOnFallback(string ollamaModelName)
+    {
+        if (_fallbackModels.TryGetValue(ollamaModelName, out var expiresAt))
+        {
+            if (expiresAt > DateTime.UtcNow)
+            {
+                return true;
+            }
+
+            _fallbackModels.TryRemove(ollamaModelName, out _);
+        }
+
+        return false;
+    }
+
+    public void ClearModelFallback(string ollamaModelName)
+    {
+        if (_fallbackModels.TryRemove(ollamaModelName, out _))
+        {
+            _logger.LogInformation("Cleared llama.cpp fallback marker for model {Model}.", ollamaModelName);
+        }
+    }
+
+    public async Task<bool> IsContainerRunningAsync(string ollamaModelName)
+    {
+        if (!_modelPorts.ContainsKey(ollamaModelName))
+        {
+            return false;
+        }
+
+        var containerName = SanitizeContainerName($"ngino-llamacpp-{ollamaModelName}");
+        var (exitCode, output) = await RunDockerWithOutputAsync(
+            ["ps", "--filter", $"name=^{containerName}$", "--format", "{{.ID}}"],
+            CancellationToken.None);
+
+        var running = exitCode == 0 && !string.IsNullOrWhiteSpace(output);
+        if (!running)
+        {
+            _logger.LogWarning(
+                "llama.cpp container {ContainerName} is no longer running. Invalidating cached port for {Model}.",
+                containerName, ollamaModelName);
+            RemoveModelPort(ollamaModelName);
+        }
+
+        return running;
+    }
+
+    public bool RemoveModelMapping(string ollamaModelName)
+    {
+        if (RemoveModelPort(ollamaModelName))
+        {
+            _logger.LogWarning(
+                "Removed stale llama.cpp port mapping for {Model}.",
+                ollamaModelName);
+            return true;
+        }
+
+        return false;
+    }
+
     public Uri? GetUpstream(string ollamaModelName)
     {
         if (_modelPorts.TryGetValue(ollamaModelName, out var port))
@@ -94,62 +184,130 @@ internal sealed partial class LlamaCppManager : IAsyncDisposable
             return false;
         }
 
+        var startLock = _modelStartLocks.GetOrAdd(ollamaName, static _ => new SemaphoreSlim(1, 1));
+        await startLock.WaitAsync(cancellationToken);
+        try
+        {
+            return await StartModelContainerCoreAsync(model, cancellationToken);
+        }
+        finally
+        {
+            startLock.Release();
+        }
+    }
+
+    private async Task<bool> StartModelContainerCoreAsync(LlamaCppModel model, CancellationToken cancellationToken)
+    {
+        var ollamaName = model.OllamaName;
+        if (string.IsNullOrWhiteSpace(ollamaName))
+        {
+            return false;
+        }
+
         if (_modelPorts.ContainsKey(ollamaName))
         {
             _logger.LogInformation("Model {Model} already has a running container", ollamaName);
             return true;
         }
 
+        if (IsModelOnFallback(ollamaName))
+        {
+            _logger.LogInformation(
+                "Model {Model} previously failed to load via llama.cpp. Skipping container start.",
+                ollamaName);
+            return false;
+        }
+
         if (!File.Exists(model.BlobPath))
         {
             _logger.LogError("Model blob not found: {BlobPath}", model.BlobPath);
+            MarkModelAsFallback(ollamaName);
             return false;
         }
 
         var port = FindAvailablePort();
         var containerName = SanitizeContainerName($"ngino-llamacpp-{ollamaName}");
 
-        var existingPort = await FindExistingContainerPortAsync(containerName);
-        if (existingPort.HasValue)
-        {
-            _logger.LogInformation(
-                "Reusing existing container for {Model} on port {Port}", ollamaName, existingPort.Value);
-            _modelPorts[ollamaName] = existingPort.Value;
-            return true;
-        }
-
-        await RunDockerAsync(["rm", "-f", containerName], CancellationToken.None);
-
-        var args = BuildDockerRunArgs(containerName, model, port);
-        _logger.LogInformation(
-            "Starting llama.cpp container for {Model} on port {Port}: docker {Args}",
-            ollamaName, port, string.Join(" ", args));
-
         try
         {
+            var existingPort = await FindExistingContainerPortAsync(containerName);
+            if (existingPort.HasValue)
+            {
+                _logger.LogInformation(
+                    "Reusing existing container for {Model} on port {Port}", ollamaName, existingPort.Value);
+                _modelPorts[ollamaName] = existingPort.Value;
+                return true;
+            }
+
+            await RunDockerAsync(["rm", "-f", containerName], CancellationToken.None);
+
+            var args = BuildDockerRunArgs(containerName, model, port);
+            _logger.LogInformation(
+                "Starting llama.cpp container for {Model} on port {Port}: docker {Args}",
+                ollamaName, port, string.Join(" ", args));
+
             var (exitCode, output) = await RunDockerWithOutputAsync(args, cancellationToken);
             if (exitCode != 0)
             {
                 _logger.LogError(
                     "Failed to start llama.cpp container for {Model}, exit code: {ExitCode}, output: {Output}",
                     ollamaName, exitCode, output);
+                MarkModelAsFallback(ollamaName);
+                return false;
+            }
+
+            _logger.LogInformation(
+                "llama.cpp container for {Model} started on port {Port}. Waiting for it to become ready...",
+                ollamaName, port);
+
+            var result = await WaitForServerReadyAsync("localhost", port, containerName, cancellationToken);
+            if (result != ContainerStartResult.Ready)
+            {
+                if (result == ContainerStartResult.ContainerExited)
+                {
+                    MarkModelAsPermanentFallback(ollamaName);
+                }
+                else
+                {
+                    MarkModelAsFallback(ollamaName);
+                }
+
+                _logger.LogError(
+                    "llama.cpp container for {Model} did not become ready on port {Port} within {Timeout} ({Result}). Stopping it.",
+                    ollamaName, port, ContainerStartTimeout, result);
+
+                try
+                {
+                    await RunDockerAsync(["stop", "--time", "10", containerName], CancellationToken.None);
+                    await RunDockerAsync(["rm", "-f", containerName], CancellationToken.None);
+                }
+                catch (Exception cleanupException)
+                {
+                    _logger.LogWarning(cleanupException, "Failed to clean up container {ContainerName}", containerName);
+                }
+
                 return false;
             }
 
             _modelPorts[ollamaName] = port;
-            _logger.LogInformation("llama.cpp container for {Model} started on port {Port}", ollamaName, port);
+            _logger.LogInformation("llama.cpp container for {Model} is ready on port {Port}.", ollamaName, port);
             return true;
         }
         catch (Exception ex)
         {
             _logger.LogError(ex, "Failed to start llama.cpp container for {Model}", ollamaName);
+            MarkModelAsFallback(ollamaName);
             return false;
         }
+        finally
+        {
+            ReleaseReservedPort(port);
+        }
     }
 
     public async Task<bool> StopModelContainerAsync(string ollamaModelName, CancellationToken cancellationToken)
     {
-        if (!_modelPorts.TryRemove(ollamaModelName, out _))
+        if (!RemoveModelPort(ollamaModelName))
         {
             _logger.LogWarning("No running container found for model {Model}", ollamaModelName);
             return false;
@@ -163,6 +321,7 @@ internal sealed partial class LlamaCppManager : IAsyncDisposable
         {
             await RunDockerAsync(["stop", "--time", "10", containerName], cancellationToken);
             await RunDockerAsync(["rm", "-f", containerName], cancellationToken);
+            ClearModelFallback(ollamaModelName);
             return true;
         }
         catch (Exception ex)
@@ -198,6 +357,8 @@ internal sealed partial class LlamaCppManager : IAsyncDisposable
         }
 
         _modelPorts.Clear();
+        _reservedPorts.Clear();
+        _fallbackModels.Clear();
     }
 
     public async Task<bool> TestDockerAsync()
@@ -325,7 +486,6 @@ internal sealed partial class LlamaCppManager : IAsyncDisposable
         {
             "run",
             "-d",
-            "--rm",
             "--label", $"{NginoContainerLabel}=true",
             "--name", containerName,
             "-p", $"{port}:{port}",
@@ -349,9 +509,12 @@ internal sealed partial class LlamaCppManager : IAsyncDisposable
         args.Add("-m");
         args.Add($"/models/blobs/{blobFile}");
         args.Add("-ngl");
-        args.Add("999");
-        args.Add("--parallel");
-        args.Add("4");
+        args.Add("auto");
+        if (_parallel.HasValue)
+        {
+            args.Add("--parallel");
+            args.Add(_parallel.Value.ToString());
+        }
         args.Add("--host");
         args.Add("0.0.0.0");
         args.Add("--port");
@@ -412,15 +575,170 @@ internal sealed partial class LlamaCppManager : IAsyncDisposable
 
     private int FindAvailablePort()
     {
-        var usedPorts = new HashSet<int>(_modelPorts.Values);
-        var port = _basePort;
-
-        while (usedPorts.Contains(port))
+        lock (_portAllocationLock)
         {
-            port++;
+            var usedPorts = new HashSet<int>(_modelPorts.Values);
+            foreach (var reservedPort in _reservedPorts.Keys)
+            {
+                usedPorts.Add(reservedPort);
+            }
+
+            var port = _basePort;
+
+            while (usedPorts.Contains(port))
+            {
+                port++;
+            }
+
+            _reservedPorts[port] = 0;
+            return port;
+        }
+    }
+
+    private void ReleaseReservedPort(int port)
+    {
+        _reservedPorts.TryRemove(port, out _);
+    }
+
+    private bool RemoveModelPort(string ollamaModelName)
+    {
+        if (_modelPorts.TryRemove(ollamaModelName, out var port))
+        {
+            ReleaseReservedPort(port);
+            return true;
         }
 
-        return port;
+        return false;
+    }
+
+    private async Task<ContainerStartResult> WaitForServerReadyAsync(
+        string host, int port, string containerName, CancellationToken cancellationToken)
+    {
+        var deadline = DateTime.UtcNow + ContainerStartTimeout;
+
+        var tcpResult = await WaitForTcpPortAsync(host, port, containerName, deadline, cancellationToken);
+        if (tcpResult != ContainerStartResult.Ready)
+        {
+            return tcpResult;
+        }
+
+        using var handler = new SocketsHttpHandler
+        {
+            ConnectTimeout = TimeSpan.FromSeconds(3),
+            UseProxy = false
+        };
+        using var httpClient = new HttpClient(handler) { Timeout = TimeSpan.FromSeconds(10) };
+
+        while (DateTime.UtcNow < deadline)
+        {
+            cancellationToken.ThrowIfCancellationRequested();
+
+            if (!await IsDockerContainerRunningAsync(containerName))
+            {
+                await LogContainerOutputAsync(containerName);
+                return ContainerStartResult.ContainerExited;
+            }
+
+            try
+            {
+                using var response = await httpClient.GetAsync(
+                    $"http://{host}:{port}/health", cancellationToken);
+                if (response.IsSuccessStatusCode)
+                {
+                    return ContainerStartResult.Ready;
+                }
+            }
+            catch (OperationCanceledException) when (!cancellationToken.IsCancellationRequested)
+            {
+            }
+            catch (HttpRequestException ex)
+            {
+                _logger.LogDebug(ex, "Health probe of llama.cpp container {ContainerName} failed; retrying.", containerName);
+            }
+            catch (IOException ex)
+            {
+                _logger.LogDebug(ex, "Health probe of llama.cpp container {ContainerName} failed; retrying.", containerName);
+            }
+
+            await Task.Delay(TimeSpan.FromSeconds(2), cancellationToken);
+        }
+
+        return ContainerStartResult.TimedOut;
+    }
+
+    private async Task<ContainerStartResult> WaitForTcpPortAsync(
+        string host, int port, string containerName, DateTime deadline, CancellationToken cancellationToken)
+    {
+        while (DateTime.UtcNow < deadline)
+        {
+            cancellationToken.ThrowIfCancellationRequested();
+
+            if (!await IsDockerContainerRunningAsync(containerName))
+            {
+                await LogContainerOutputAsync(containerName);
+                return ContainerStartResult.ContainerExited;
+            }
+
+            try
+            {
+                using var client = new TcpClient();
+                await client.ConnectAsync(host, port, cancellationToken);
+                return ContainerStartResult.Ready;
+            }
+            catch (OperationCanceledException)
+            {
+                throw;
+            }
+            catch (SocketException)
+            {
+            }
+
+            await Task.Delay(TimeSpan.FromSeconds(2), cancellationToken);
+        }
+
+        return ContainerStartResult.TimedOut;
+    }
+
+    private async Task<bool> IsDockerContainerRunningAsync(string containerName)
+    {
+        try
+        {
+            var (exitCode, output) = await RunDockerWithOutputAsync(
+                ["inspect", "-f", "{{.State.Running}}", containerName],
+                CancellationToken.None);
+            return exitCode == 0 && string.Equals(output.Trim(), "true", StringComparison.OrdinalIgnoreCase);
+        }
+        catch
+        {
+            return true;
+        }
+    }
+
+    private async Task LogContainerOutputAsync(string containerName)
+    {
+        try
+        {
+            var (_, output) = await RunDockerWithOutputAsync(
+                ["logs", "--tail", "100", containerName],
+                CancellationToken.None);
+
+            if (!string.IsNullOrWhiteSpace(output))
+            {
+                _logger.LogError(
+                    "llama.cpp container {ContainerName} exited before becoming ready. Last output:\n{Output}",
+                    containerName, output);
+            }
+            else
+            {
+                _logger.LogError(
+                    "llama.cpp container {ContainerName} exited before becoming ready, but produced no output.",
+                    containerName);
+            }
+        }
+        catch (Exception ex)
+        {
+            _logger.LogWarning(ex, "Failed to read logs of container {ContainerName}", containerName);
+        }
     }
 
     private static string SanitizeContainerName(string name)
@@ -520,6 +838,13 @@ internal sealed partial class LlamaCppManager : IAsyncDisposable
 
     [GeneratedRegex(@"[^a-zA-Z0-9_.-]")]
     private static partial Regex InvalidContainerNameChars();
+
+    private enum ContainerStartResult
+    {
+        Ready,
+        ContainerExited,
+        TimedOut
+    }
 }
 
 internal sealed record LlamaCppModel

src/Ngino.Client/Program.cs

@@ -22,8 +22,15 @@ try
         Console.WriteLine($"  ollama models path: {options.UseOllamaModelsPath ?? "(not set)"}");
         Console.WriteLine($"  llama.cpp docker image: {options.LlamaCppDockerImage ?? "(auto)"}");
         Console.WriteLine($"  llama.cpp base port: {options.LlamaCppBasePort}");
+        Console.WriteLine(
+            options.LlamaCppParallel.HasValue
+                ? $"  llama.cpp parallel slots: {options.LlamaCppParallel.Value}"
+                : "  llama.cpp parallel slots: (llama.cpp default)");
     }
 
+    var logDirectory = options.LogDirectory ?? Path.Combine(AppContext.BaseDirectory, "Logs");
+    Console.WriteLine($"  log directory: {logDirectory}");
+
     // Args are parsed by ClientOptions; keep them away from the host configuration.
     var builder = Host.CreateApplicationBuilder(new HostApplicationBuilderSettings { Args = [] });
     builder.Services.AddSingleton(options);
@@ -32,6 +39,7 @@ try
     builder.Services.AddWindowsService(service => service.ServiceName = "NginoClient");
     // The EventLog provider defaults to Warning; connection state is worth seeing there.
     builder.Logging.AddFilter<Microsoft.Extensions.Logging.EventLog.EventLogLoggerProvider>("Ngino.Client", LogLevel.Information);
+    builder.Logging.AddProvider(new FileLoggerProvider(logDirectory));
 
     await builder.Build().RunAsync();
     return 0;

src/Ngino.Client/TunnelClient.cs

@@ -47,7 +47,9 @@ internal sealed class TunnelClient
                 _options.UseOllamaModelsPath,
                 _options.LlamaCppDockerImage,
                 _options.LlamaCppBasePort,
-                _logger);
+                _logger,
+                _options.LlamaCppFallbackCooldown,
+                _options.LlamaCppParallel);
         }
     }
 
@@ -477,13 +479,16 @@ internal sealed class TunnelClient
                 var started = await _llamaCppManager.StartModelContainerAsync(model, cancellationToken);
                 if (!started)
                 {
-                    return new TunnelMessage
-                    {
-                        Type = TunnelMessageTypes.ModelCommandResult,
-                        RequestId = message.RequestId,
-                        StatusCode = 500,
-                        Error = $"Failed to start llama.cpp container for model '{modelName}'."
-                    };
+                    _logger.LogWarning(
+                        "Unable to load model '{Model}' via llama.cpp. Falling back to Ollama upstream.",
+                        modelName);
+
+                    using var fallbackRequest = BuildModelCommandRequest(_options.Upstream, "load", modelName);
+                    using var fallbackResponse = await _httpClient.SendAsync(
+                        fallbackRequest, HttpCompletionOption.ResponseHeadersRead, cancellationToken);
+                    var fallbackBody = await fallbackResponse.Content.ReadAsByteArrayAsync(cancellationToken);
+
+                    return BuildModelCommandResult(message.RequestId, fallbackResponse, fallbackBody);
                 }
 
                 return BuildModelCommandResult(message.RequestId, 200, "OK", []);
@@ -494,13 +499,18 @@ internal sealed class TunnelClient
                 var stopped = await _llamaCppManager!.StopModelContainerAsync(modelName!, cancellationToken);
                 if (!stopped)
                 {
-                    return new TunnelMessage
-                    {
-                        Type = TunnelMessageTypes.ModelCommandResult,
-                        RequestId = message.RequestId,
-                        StatusCode = 404,
-                        Error = $"No running llama.cpp container for model '{modelName}'."
-                    };
+                    _logger.LogWarning(
+                        "No running llama.cpp container for model '{Model}'. Falling back to Ollama upstream to unload it.",
+                        modelName);
+
+                    _llamaCppManager.ClearModelFallback(modelName!);
+
+                    using var fallbackRequest = BuildModelCommandRequest(_options.Upstream, "unload", modelName);
+                    using var fallbackResponse = await _httpClient.SendAsync(
+                        fallbackRequest, HttpCompletionOption.ResponseHeadersRead, cancellationToken);
+                    var fallbackBody = await fallbackResponse.Content.ReadAsByteArrayAsync(cancellationToken);
+
+                    return BuildModelCommandResult(message.RequestId, fallbackResponse, fallbackBody);
                 }
 
                 return BuildModelCommandResult(message.RequestId, 200, "OK", []);
@@ -694,28 +704,54 @@ internal sealed class TunnelClient
             if (modelName is not null)
             {
                 effectiveUpstream = _llamaCppManager.GetUpstream(modelName);
+                if (effectiveUpstream is not null)
+                {
+                    var running = await _llamaCppManager.IsContainerRunningAsync(modelName);
+                    if (!running)
+                    {
+                        _logger.LogWarning(
+                            "Cached llama.cpp container for model '{Model}' is not running anymore. Starting a fresh one on demand...",
+                            modelName);
+                        effectiveUpstream = null;
+                    }
+                }
+
                 if (effectiveUpstream is null)
                 {
-                    _logger.LogInformation(
-                        "Request for model '{Model}' but no llama.cpp container is running. Starting one on demand...",
-                        modelName);
-
                     var model = _llamaCppManager.DiscoverModelsWithBlob()
                         .FirstOrDefault(m => string.Equals(m.OllamaName, modelName, StringComparison.OrdinalIgnoreCase));
 
                     if (model is not null)
                     {
-                        var started = await _llamaCppManager.StartModelContainerAsync(model, cancellationToken);
-                        if (started)
+                        if (_llamaCppManager.IsModelOnFallback(modelName))
                         {
-                            effectiveUpstream = _llamaCppManager.GetUpstream(modelName);
+                            _logger.LogInformation(
+                                "Model '{Model}' previously failed to load via llama.cpp. Routing directly to the Ollama upstream.",
+                                modelName);
+                        }
+                        else
+                        {
+                            _logger.LogInformation(
+                                "Request for model '{Model}' but no llama.cpp container is running. Starting one on demand...",
+                                modelName);
+
+                            var started = await _llamaCppManager.StartModelContainerAsync(model, cancellationToken);
+                            if (started)
+                            {
+                                effectiveUpstream = _llamaCppManager.GetUpstream(modelName);
+                            }
+                            else
+                            {
+                                _logger.LogWarning(
+                                    "Unable to load model '{Model}' via llama.cpp. Falling back to the Ollama upstream.",
+                                    modelName);
+                            }
                         }
                     }
-
-                    if (effectiveUpstream is null)
+                    else
                     {
                         _logger.LogWarning(
-                            "Failed to start llama.cpp container for model '{Model}'. Falling back to default upstream.",
+                            "Model '{Model}' was not found in the Ollama models path. Falling back to default upstream.",
                             modelName);
                     }
                 }
@@ -766,7 +802,14 @@ internal sealed class TunnelClient
             effectiveUpstream: effectiveUpstream,
             responseHandler: responseHandler,
             pathTransform: pathTransform,
-            bodyTransform: bodyTransform);
+            bodyTransform: bodyTransform,
+            onConnectionRefused: () =>
+            {
+                if (_llamaCppManager is not null && modelName is not null)
+                {
+                    _llamaCppManager.RemoveModelMapping(modelName);
+                }
+            });
 
         if (!_activeRequests.TryAdd(message.RequestId, request))
         {

src/Ngino.Client/UpstreamRequest.cs

@@ -1,4 +1,5 @@
 using System.Net.Http.Headers;
+using System.Net.Sockets;
 using System.Threading.Channels;
 using Ngino.Protocol;
 
@@ -39,6 +40,7 @@ internal sealed class UpstreamRequest
     private readonly Func<HttpResponseMessage, CancellationToken, Task>? _responseHandler;
     private readonly Func<string, string?>? _pathTransform;
     private readonly Func<byte[], byte[]>? _bodyTransform;
+    private readonly Action? _onConnectionRefused;
     private readonly List<byte[]> _bufferedBody = [];
 
     public UpstreamRequest(
@@ -51,7 +53,8 @@ internal sealed class UpstreamRequest
         Uri? effectiveUpstream = null,
         Func<HttpResponseMessage, CancellationToken, Task>? responseHandler = null,
         Func<string, string?>? pathTransform = null,
-        Func<byte[], byte[]>? bodyTransform = null)
+        Func<byte[], byte[]>? bodyTransform = null,
+        Action? onConnectionRefused = null)
     {
         _httpClient = httpClient;
         _initialMessage = initialMessage;
@@ -61,6 +64,7 @@ internal sealed class UpstreamRequest
         _responseHandler = responseHandler;
         _pathTransform = pathTransform;
         _bodyTransform = bodyTransform;
+        _onConnectionRefused = onConnectionRefused;
         _cancellationTokenSource = CancellationTokenSource.CreateLinkedTokenSource(cancellationToken);
 
         if (!initialMessage.HasBody)
@@ -143,6 +147,17 @@ internal sealed class UpstreamRequest
         }
         catch (Exception exception)
         {
+            if (IsConnectionRefused(exception))
+            {
+                try
+                {
+                    _onConnectionRefused?.Invoke();
+                }
+                catch
+                {
+                }
+            }
+
             await SendErrorAsync(exception);
         }
         finally
@@ -296,6 +311,25 @@ internal sealed class UpstreamRequest
         }
     }
 
+    internal static bool IsConnectionRefused(Exception exception)
+    {
+        for (var current = exception; current is not null; current = current.InnerException)
+        {
+            if (current is SocketException socketException
+                && socketException.SocketErrorCode == SocketError.ConnectionRefused)
+            {
+                return true;
+            }
+
+            if (current.Message.Contains("Connection refused", StringComparison.OrdinalIgnoreCase))
+            {
+                return true;
+            }
+        }
+
+        return false;
+    }
+
     private static List<HeaderPair> CollectResponseHeaders(HttpResponseMessage response)
     {
         var headers = new List<HeaderPair>();

src/Ngino.Server/AdminEndpoints.cs

@@ -167,11 +167,14 @@ internal static class AdminEndpoints
             try
             {
                 var manual = string.Equals(request.Mode, "manual", StringComparison.OrdinalIgnoreCase);
+
                 TimeSpan? duration = manual
                     ? null
-                    : TimeSpan.FromMinutes(Math.Clamp(request.DurationMinutes ?? 60, 1, 24 * 60));
+                    : request.DurationMinutes is { } minutes
+                        ? TimeSpan.FromMinutes(Math.Clamp(minutes, 1, 24 * 60))
+                        : null;
 
-                store.DisableClient(clientId, duration, manual, request.Reason);
+                store.DisableClient(clientId, duration, manual, request.Reason, request.StartAtUtc, request.UntilUtc);
                 return Results.Ok(new { clientId, disabled = true });
             }
             catch (Exception exception)
@@ -673,6 +676,7 @@ internal static class AdminEndpoints
                 snapshot?.ActiveModels ?? [],
                 snapshot?.ModelsUpdatedAt,
                 access.IsDisabled,
+                access.DisabledFromUtc,
                 access.DisabledUntilUtc,
                 access.DisabledManually,
                 access.DisabledReason,
@@ -880,7 +884,9 @@ internal static class AdminEndpoints
 internal sealed record DisableClientRequest(
     string? Mode,
     int? DurationMinutes,
-    string? Reason);
+    string? Reason,
+    DateTimeOffset? StartAtUtc,
+    DateTimeOffset? UntilUtc);
 
 internal sealed record ModelActionRequest(
     string ClientId,
@@ -929,6 +935,7 @@ internal sealed record ClientSummary(
     IReadOnlyList<string> ActiveModels,
     DateTimeOffset? ModelsUpdatedAt,
     bool Disabled,
+    DateTimeOffset? DisabledFromUtc,
     DateTimeOffset? DisabledUntilUtc,
     bool DisabledManually,
     string? DisabledReason,

src/Ngino.Server/ManagementStore.cs

@@ -378,7 +378,7 @@ internal sealed class ManagementStore
             using var connection = OpenConnection();
             using var command = connection.CreateCommand();
             command.CommandText = """
-                SELECT disabled_until_utc, disabled_manually, disabled_reason
+                SELECT disabled_until_utc, disabled_manually, disabled_reason, disabled_from_utc
                 FROM client_controls
                 WHERE client_id = $client_id
                 """;
@@ -393,18 +393,22 @@ internal sealed class ManagementStore
             var disabledUntil = ReadNullableDateTimeOffset(reader, 0);
             var disabledManually = reader.GetInt32(1) != 0;
             var reason = reader.IsDBNull(2) ? null : reader.GetString(2);
+            var disabledFrom = ReadNullableDateTimeOffset(reader, 3);
 
-            if (disabledManually)
+            var now = DateTimeOffset.UtcNow;
+            var isScheduled = disabledFrom > now;
+
+            if (disabledManually && !isScheduled)
             {
-                return new ClientAccess(true, null, true, reason);
+                return new ClientAccess(true, disabledFrom, null, true, reason);
             }
 
-            if (disabledUntil is { } until && until > DateTimeOffset.UtcNow)
+            if (!isScheduled && disabledUntil is { } until && until > now)
             {
-                return new ClientAccess(true, until, false, reason);
+                return new ClientAccess(true, disabledFrom, until, false, reason);
             }
 
-            return ClientAccess.Enabled;
+            return new ClientAccess(false, disabledFrom, disabledUntil, false, reason);
         }
     }
 
@@ -421,7 +425,7 @@ internal sealed class ManagementStore
         {
             using var connection = OpenConnection();
             using var command = connection.CreateCommand();
-            command.CommandText = "SELECT client_id, disabled_until_utc, disabled_manually, disabled_reason FROM client_controls";
+            command.CommandText = "SELECT client_id, disabled_until_utc, disabled_manually, disabled_reason, disabled_from_utc FROM client_controls";
 
             using var reader = command.ExecuteReader();
             while (reader.Read())
@@ -430,19 +434,23 @@ internal sealed class ManagementStore
                 var disabledUntil = ReadNullableDateTimeOffset(reader, 1);
                 var disabledManually = reader.GetInt32(2) != 0;
                 var reason = reader.IsDBNull(3) ? null : reader.GetString(3);
+                var disabledFrom = ReadNullableDateTimeOffset(reader, 4);
 
-                result[clientId] = disabledManually
-                    ? new ClientAccess(true, null, true, reason)
-                    : disabledUntil is { } until && until > DateTimeOffset.UtcNow
-                        ? new ClientAccess(true, until, false, reason)
-                        : ClientAccess.Enabled;
+                var now = DateTimeOffset.UtcNow;
+                var isScheduled = disabledFrom > now;
+
+                result[clientId] = disabledManually && !isScheduled
+                    ? new ClientAccess(true, disabledFrom, null, true, reason)
+                    : !isScheduled && disabledUntil is { } until && until > now
+                        ? new ClientAccess(true, disabledFrom, until, false, reason)
+                        : new ClientAccess(false, disabledFrom, disabledUntil, disabledManually, reason);
             }
         }
 
         return result;
     }
 
-    public void DisableClient(string clientId, TimeSpan? duration, bool manually, string? reason)
+    public void DisableClient(string clientId, TimeSpan? duration, bool manually, string? reason, DateTimeOffset? startAtUtc = null, DateTimeOffset? untilUtc = null)
     {
         EnsureAvailable();
 
@@ -452,7 +460,9 @@ internal sealed class ManagementStore
         }
 
         var now = DateTimeOffset.UtcNow;
-        var disabledUntil = manually ? null : now.Add(duration ?? TimeSpan.FromHours(1)).ToString("O");
+        var disabledFrom = startAtUtc?.ToString("O");
+        var disabledUntil = untilUtc?.ToString("O")
+            ?? (manually ? null : now.Add(duration ?? TimeSpan.FromHours(1)).ToString("O"));
 
         lock (_lock)
         {
@@ -462,23 +472,27 @@ internal sealed class ManagementStore
                 INSERT INTO client_controls (
                     client_id,
                     disabled_until_utc,
+                    disabled_from_utc,
                     disabled_manually,
                     disabled_reason,
                     updated_at_utc)
                 VALUES (
                     $client_id,
                     $disabled_until_utc,
+                    $disabled_from_utc,
                     $disabled_manually,
                     $disabled_reason,
                     $updated_at_utc)
                 ON CONFLICT(client_id) DO UPDATE SET
                     disabled_until_utc = excluded.disabled_until_utc,
+                    disabled_from_utc = excluded.disabled_from_utc,
                     disabled_manually = excluded.disabled_manually,
                     disabled_reason = excluded.disabled_reason,
                     updated_at_utc = excluded.updated_at_utc
                 """;
             command.Parameters.AddWithValue("$client_id", clientId);
             command.Parameters.AddWithValue("$disabled_until_utc", (object?)disabledUntil ?? DBNull.Value);
+            command.Parameters.AddWithValue("$disabled_from_utc", (object?)disabledFrom ?? DBNull.Value);
             command.Parameters.AddWithValue("$disabled_manually", manually ? 1 : 0);
             command.Parameters.AddWithValue("$disabled_reason", string.IsNullOrWhiteSpace(reason) ? DBNull.Value : reason.Trim());
             command.Parameters.AddWithValue("$updated_at_utc", now.ToString("O"));
@@ -503,17 +517,20 @@ internal sealed class ManagementStore
                 INSERT INTO client_controls (
                     client_id,
                     disabled_until_utc,
+                    disabled_from_utc,
                     disabled_manually,
                     disabled_reason,
                     updated_at_utc)
                 VALUES (
                     $client_id,
                     NULL,
+                    NULL,
                     0,
                     NULL,
                     $updated_at_utc)
                 ON CONFLICT(client_id) DO UPDATE SET
                     disabled_until_utc = NULL,
+                    disabled_from_utc = NULL,
                     disabled_manually = 0,
                     disabled_reason = NULL,
                     updated_at_utc = excluded.updated_at_utc
@@ -1997,6 +2014,7 @@ internal sealed class ManagementStore
                 CREATE TABLE IF NOT EXISTS client_controls (
                     client_id TEXT NOT NULL PRIMARY KEY,
                     disabled_until_utc TEXT NULL,
+                    disabled_from_utc TEXT NULL,
                     disabled_manually INTEGER NOT NULL DEFAULT 0,
                     disabled_reason TEXT NULL,
                     updated_at_utc TEXT NOT NULL
@@ -2306,6 +2324,21 @@ internal sealed class ManagementStore
             }
         }
 
+        using (var migrate = connection.CreateCommand())
+        {
+            migrate.CommandText = """
+                SELECT COUNT(*) FROM pragma_table_info('client_controls') WHERE name = 'disabled_from_utc'
+                """;
+            var hasFromColumn = (long)migrate.ExecuteScalar()! > 0;
+
+            if (!hasFromColumn)
+            {
+                using var addFromCol = connection.CreateCommand();
+                addFromCol.CommandText = "ALTER TABLE client_controls ADD COLUMN disabled_from_utc TEXT NULL";
+                addFromCol.ExecuteNonQuery();
+            }
+        }
+
         _logger.LogInformation(
             "Loaded {ClientKeyCount} client key(s) from {DatabasePath}.",
             _clientKeysByHash.Count,
@@ -2406,11 +2439,12 @@ internal sealed class ManagementStore
 
 internal sealed record ClientAccess(
     bool IsDisabled,
+    DateTimeOffset? DisabledFromUtc,
     DateTimeOffset? DisabledUntilUtc,
     bool DisabledManually,
     string? DisabledReason)
 {
-    public static ClientAccess Enabled { get; } = new(false, null, false, null);
+    public static ClientAccess Enabled { get; } = new(false, null, null, false, null);
 }
 
 internal sealed record UserKeyInfo(

src/Ngino.Server/Program.cs

@@ -269,6 +269,11 @@ app.UseStaticFiles();
 
 app.MapAdminEndpoints(settings);
 
+app.MapGet("/favicon.ico", () =>
+    Results.File(
+        Path.Combine(app.Environment.WebRootPath, "favicon.ico"),
+        "image/x-icon"));
+
 app.MapGet("/", () => Results.Redirect("/admin"));
 
 app.MapGet(settings.StatusPath, (HttpContext context, TunnelHub hub, ServerSettings serverSettings, EmbeddingCache embeddingCache, ManagementStore managementStore) =>

src/Ngino.Server/ReverseProxyEndpoint.cs

@@ -10,6 +10,8 @@ internal static class ReverseProxyEndpoint
 {
     private const string UnauthorizedMessage = "Missing or invalid Ngino token.";
 
+    private const string OllamaVersion = "0.32.5";
+
     private static readonly HashSet<string> HopByHopHeaders = new(StringComparer.OrdinalIgnoreCase)
     {
         "Connection",
@@ -72,6 +74,12 @@ internal static class ReverseProxyEndpoint
             return;
         }
 
+        if (IsVersionRequest(context.Request, proxyPath))
+        {
+            await WriteVersionResponseAsync(context);
+            return;
+        }
+
         if (TryGetClientAddress(proxyPath, out var pathClientId, out var clientPath))
         {
             if (!groupAccess.IsClientAllowed(pathClientId))
@@ -317,6 +325,15 @@ internal static class ReverseProxyEndpoint
         HttpMethods.IsGet(request.Method)
         && string.Equals(proxyPath.Value, "/api/tags", StringComparison.OrdinalIgnoreCase);
 
+    private static bool IsVersionRequest(HttpRequest request, PathString proxyPath) =>
+        HttpMethods.IsGet(request.Method)
+        && string.Equals(proxyPath.Value, "/api/version", StringComparison.OrdinalIgnoreCase);
+
+    private static Task WriteVersionResponseAsync(HttpContext context) =>
+        context.Response.WriteAsJsonAsync(
+            new { version = OllamaVersion },
+            context.RequestAborted);
+
     private static async Task HandleTagsAsync(
         HttpContext context,
         TunnelHub hub,
@@ -373,14 +390,26 @@ internal static class ReverseProxyEndpoint
 
     private static string GetClientDisabledMessage(string clientId, ClientAccess access)
     {
+        var reason = string.IsNullOrWhiteSpace(access.DisabledReason)
+            ? ""
+            : $" Reason: {access.DisabledReason.Trim()}.";
+
         if (access.DisabledManually)
         {
-            return $"Tunnel client '{clientId}' is disabled until it is enabled manually.";
+            return $"Tunnel client '{clientId}' is disabled until it is enabled manually.{reason}";
         }
 
-        return access.DisabledUntilUtc is { } disabledUntil
-            ? $"Tunnel client '{clientId}' is disabled until {disabledUntil:O}."
-            : $"Tunnel client '{clientId}' is disabled.";
+        if (access.DisabledUntilUtc is { } disabledUntil)
+        {
+            var from = access.DisabledFromUtc is { } fromUtc
+                ? $" (scheduled from {fromUtc:O})"
+                : "";
+            return $"Tunnel client '{clientId}' is disabled until {disabledUntil:O}.{from}{reason}";
+        }
+
+        return access.DisabledFromUtc is { } fromUtc2
+            ? $"Tunnel client '{clientId}' is disabled (scheduled from {fromUtc2:O}).{reason}"
+            : $"Tunnel client '{clientId}' is disabled.{reason}";
     }
 
     private static async Task<string?> GetRequestedModelAsync(HttpRequest request, PathString proxyPath)

src/Ngino.Server/wwwroot/admin/app.css

@@ -58,16 +58,12 @@ textarea {
 }
 
 .brand-mark {
-  display: grid;
-  place-items: center;
   width: 34px;
   height: 34px;
   border: 1px solid #4c5d74;
   border-radius: 6px;
   background: #223044;
-  color: #c8f2df;
-  font-weight: 800;
-  letter-spacing: 0;
+  object-fit: cover;
 }
 
 .brand strong,
@@ -592,6 +588,71 @@ tr:last-child td {
   }
 }
 
+.modal-overlay {
+  position: fixed;
+  inset: 0;
+  z-index: 100;
+  display: grid;
+  place-items: center;
+  background: rgba(0, 0, 0, 0.35);
+  opacity: 0;
+  transition: opacity 0.15s;
+}
+
+.modal-overlay.visible {
+  opacity: 1;
+}
+
+.modal-dialog {
+  width: min(480px, calc(100vw - 32px));
+  max-height: calc(100vh - 32px);
+  overflow-y: auto;
+  display: grid;
+  gap: 14px;
+  padding: 20px;
+  border-radius: 10px;
+  background: #ffffff;
+  box-shadow: 0 20px 60px rgba(0, 0, 0, 0.2);
+}
+
+.modal-dialog h3 {
+  margin: 0;
+  font-size: 16px;
+  overflow-wrap: anywhere;
+}
+
+.field-label {
+  color: var(--muted);
+  font-size: 12px;
+  font-weight: 600;
+}
+
+.radio-row {
+  display: flex;
+  gap: 14px;
+  align-items: center;
+}
+
+.radio-row label {
+  display: flex;
+  align-items: center;
+  gap: 5px;
+  font-size: 14px;
+  cursor: pointer;
+}
+
+.field-row {
+  display: flex;
+  gap: 8px;
+  align-items: center;
+}
+
+.modal-actions {
+  display: flex;
+  gap: 8px;
+  justify-content: flex-end;
+}
+
 @media (max-width: 520px) {
   .nav {
     grid-template-columns: 1fr;

src/Ngino.Server/wwwroot/admin/app.js

@@ -94,13 +94,21 @@ content.addEventListener("click", async (event) => {
   try {
     setBusy(button, true);
 
-    if (action === "disable-hour") {
-      await api(`/clients/${encodeURIComponent(clientId)}/disable`, {
-        method: "POST",
-        body: { mode: "duration", durationMinutes: 60 }
-      });
-      setNotice(`Disabled ${clientId} for one hour.`);
-      await refresh();
+    if (action === "disable-temporary") {
+      setBusy(button, false);
+      try {
+        const body = await showDisableModal(clientId);
+        setBusy(button, true);
+        await api(`/clients/${encodeURIComponent(clientId)}/disable`, {
+          method: "POST",
+          body
+        });
+        setNotice(`Disabled ${clientId}.`);
+        await refresh();
+      } catch {
+        // cancelled
+      }
+      return;
     }
 
     if (action === "disable-manual") {
@@ -498,6 +506,7 @@ function clientsTable(clients) {
           ${client.disabled ? badge(client.disabledManually ? "Disabled manual" : "Disabled timed", "bad") : badge("Enabled", "good")}
         </div>
         ${client.disabled ? `<div class="cell-sub">${escapeHtml(disabledText(client))}</div>` : ""}
+        ${isScheduled(client) ? `<div class="cell-sub">${escapeHtml(disabledText(client))}</div>` : ""}
       </td>
       <td>${number(client.pendingRequests)}</td>
       <td>
@@ -513,7 +522,7 @@ function clientsTable(clients) {
       </td>
       <td>
         <div class="actions">
-          <button class="button warning" data-action="disable-hour" data-client-id="${escapeAttr(client.id)}" ${client.disabled ? "disabled" : ""}>Disable 1h</button>
+          <button class="button warning" data-action="disable-temporary" data-client-id="${escapeAttr(client.id)}" ${client.disabled ? "disabled" : ""}>Disable temporary</button>
           <button class="button warning" data-action="disable-manual" data-client-id="${escapeAttr(client.id)}" ${client.disabled ? "disabled" : ""}>Disable</button>
           <button class="button secondary" data-action="enable-client" data-client-id="${escapeAttr(client.id)}" ${client.disabled ? "" : "disabled"}>Enable</button>
         </div>
@@ -1657,11 +1666,24 @@ function emptyState(text) {
 }
 
 function disabledText(client) {
+  var text = "";
   if (client.disabledManually) {
-    return "Until enabled manually";
+    text = "Until enabled manually";
+  } else if (client.disabledUntilUtc) {
+    text = `Until ${formatDate(client.disabledUntilUtc)}`;
+  } else if (isScheduled(client)) {
+    text = `Scheduled from ${formatDate(client.disabledFromUtc)}`;
+  } else {
+    text = "Disabled";
   }
+  if (client.disabledReason) {
+    text += ` — ${escapeHtml(client.disabledReason)}`;
+  }
+  return text;
+}
 
-  return client.disabledUntilUtc ? `Until ${formatDate(client.disabledUntilUtc)}` : "Disabled";
+function isScheduled(client) {
+  return client.disabledFromUtc && new Date(client.disabledFromUtc) > new Date();
 }
 
 function formatDate(value) {
@@ -1723,4 +1745,127 @@ function escapeAttr(value) {
   return escapeHtml(value);
 }
 
+function showDisableModal(clientId) {
+  return new Promise((resolve, reject) => {
+    const overlay = document.createElement("div");
+    overlay.className = "modal-overlay";
+    overlay.innerHTML = `
+      <div class="modal-dialog">
+        <h3>Disable client &quot;${escapeHtml(clientId)}&quot;</h3>
+        <div class="field">
+          <span class="field-label">When to disable</span>
+          <div class="radio-row">
+            <label><input type="radio" name="d-when" value="now" checked> Now</label>
+            <label><input type="radio" name="d-when" value="later"> Later</label>
+          </div>
+          <input type="datetime-local" id="d-from" class="input" disabled>
+        </div>
+        <div class="field">
+          <span class="field-label">For how long</span>
+          <div class="radio-row">
+            <label><input type="radio" name="d-for" value="timespan" checked> Timespan</label>
+            <label><input type="radio" name="d-for" value="until"> Until</label>
+          </div>
+          <div id="d-timespan-group" class="field-row">
+            <input type="number" id="d-duration" class="input" value="1" min="1" style="width:100px">
+            <select id="d-unit" class="select" style="width:auto">
+              <option value="1">minute(s)</option>
+              <option value="60" selected>hour(s)</option>
+              <option value="1440">day(s)</option>
+            </select>
+          </div>
+          <input type="datetime-local" id="d-until" class="input" style="display:none" disabled>
+        </div>
+        <div class="field">
+          <label for="d-reason">Reason (optional)</label>
+          <input type="text" id="d-reason" class="input" placeholder="e.g. maintenance">
+        </div>
+        <div class="form-row modal-actions">
+          <button class="button secondary" id="d-cancel" type="button">Cancel</button>
+          <button class="button warning" id="d-confirm" type="button">Disable</button>
+        </div>
+      </div>
+    `;
+
+    document.body.appendChild(overlay);
+    requestAnimationFrame(() => overlay.classList.add("visible"));
+
+    const whenRadios = overlay.querySelectorAll('input[name="d-when"]');
+    const forRadios = overlay.querySelectorAll('input[name="d-for"]');
+    const fromInput = overlay.querySelector("#d-from");
+    const timespanGroup = overlay.querySelector("#d-timespan-group");
+    const untilInput = overlay.querySelector("#d-until");
+
+    function updateWhen() {
+      const val = overlay.querySelector('input[name="d-when"]:checked').value;
+      fromInput.style.display = val === "later" ? "" : "none";
+      fromInput.disabled = val !== "later";
+      if (val === "now") fromInput.value = "";
+    }
+
+    function updateFor() {
+      const val = overlay.querySelector('input[name="d-for"]:checked').value;
+      timespanGroup.style.display = val === "timespan" ? "flex" : "none";
+      untilInput.style.display = val === "until" ? "" : "none";
+      untilInput.disabled = val !== "until";
+      if (val === "timespan") untilInput.value = "";
+    }
+
+    whenRadios.forEach(r => r.addEventListener("change", updateWhen));
+    forRadios.forEach(r => r.addEventListener("change", updateFor));
+    updateFor();
+
+    overlay.querySelector("#d-confirm").addEventListener("click", () => {
+      const when = overlay.querySelector('input[name="d-when"]:checked').value;
+      const forVal = overlay.querySelector('input[name="d-for"]:checked').value;
+      const reason = overlay.querySelector("#d-reason").value.trim() || null;
+
+      if (when === "later" && !fromInput.value) {
+        setNotice("Please select a date and time for the disable.", true);
+        return;
+      }
+      if (forVal === "until" && !untilInput.value) {
+        setNotice("Please select a date and time for the end.", true);
+        return;
+      }
+      if (forVal === "timespan") {
+        const val = parseInt(overlay.querySelector("#d-duration").value);
+        if (!val || val < 1) {
+          setNotice("Please enter a valid duration.", true);
+          return;
+        }
+      }
+
+      const body = { reason };
+
+      if (when === "later") {
+        body.startAtUtc = new Date(fromInput.value).toISOString();
+      }
+
+      if (forVal === "timespan") {
+        const val = parseInt(overlay.querySelector("#d-duration").value) || 1;
+        const unit = parseInt(overlay.querySelector("#d-unit").value);
+        body.durationMinutes = val * unit;
+      } else {
+        body.untilUtc = new Date(untilInput.value).toISOString();
+      }
+
+      overlay.remove();
+      resolve(body);
+    });
+
+    overlay.querySelector("#d-cancel").addEventListener("click", () => {
+      overlay.remove();
+      reject(new Error("Cancelled"));
+    });
+
+    overlay.addEventListener("click", (e) => {
+      if (e.target === overlay) {
+        overlay.remove();
+        reject(new Error("Cancelled"));
+      }
+    });
+  });
+}
+
 boot();

src/Ngino.Server/wwwroot/admin/index.html

@@ -10,7 +10,7 @@
   <div class="app-shell">
     <aside class="sidebar">
       <div class="brand">
-        <span class="brand-mark">RL</span>
+        <img class="brand-mark" src="/admin/Ngino_logo_symbol.png" alt="Ngino">
         <span>
           <strong>Ngino</strong>
           <small>Admin</small>

tests/Ngino.Client.Tests/UpstreamRequestTests.cs

@@ -1,3 +1,4 @@
+using System.Net.Sockets;
 using Ngino.Client;
 using Xunit;
 
@@ -32,4 +33,29 @@ public sealed class UpstreamRequestTests
 
         Assert.Contains("origin-form path", exception.Message);
     }
+
+    [Fact]
+    public void IsConnectionRefused_DetectsSocketConnectionRefused()
+    {
+        var socketException = new SocketException((int)SocketError.ConnectionRefused);
+        var exception = new HttpRequestException("Connection refused", socketException);
+
+        Assert.True(UpstreamRequest.IsConnectionRefused(exception));
+    }
+
+    [Fact]
+    public void IsConnectionRefused_DetectsConnectionRefusedByMessage()
+    {
+        var exception = new HttpRequestException("Connection refused (localhost:8081)");
+
+        Assert.True(UpstreamRequest.IsConnectionRefused(exception));
+    }
+
+    [Fact]
+    public void IsConnectionRefused_IgnoresUnrelatedFailures()
+    {
+        var exception = new HttpRequestException("Connection reset by peer");
+
+        Assert.False(UpstreamRequest.IsConnectionRefused(exception));
+    }
 }