Upload 2 files

Dockerfile

@@ -0,0 +1,25 @@
+# استخدام الصورة المدمجة الرسمية
+FROM ghcr.io/open-webui/open-webui:ollama
+
+# إعداد وتوجيه كافة متغيرات البيئة إلى المخزن الدائم
+ENV OLLAMA_MODELS=/data/ollama
+ENV DATA_DIR=/data/webui
+ENV HF_HOME=/data/.hf_cache
+ENV PORT=7860
+ENV HOST=0.0.0.0
+ENV OLLAMA_BASE_URL=http://localhost:11434
+
+# السطر الأهم: إجبار Ollama على إبقاء النماذج محملة في الذاكرة دائماً
+ENV OLLAMA_KEEP_ALIVE=-1
+
+# تثبيت curl للفحص الذكي، وتحديث أداة huggingface_hub مع تنظيف المخلفات لتقليص الحجم
+RUN apt-get update && apt-get install -y curl && \
+    pip install --no-cache-dir -U huggingface_hub && \
+    rm -rf /var/lib/apt/lists/*
+
+# نسخ سكريبت التشغيل وضبط الصلاحيات
+COPY run.sh /app/run.sh
+RUN chmod +x /app/run.sh
+
+# نقطة الانطلاق السليمة
+CMD ["/app/run.sh"]

run.sh

@@ -0,0 +1,171 @@
+#!/bin/bash
+set -euo pipefail
+
+# ============================================================
+# المتغيرات الأساسية
+# ============================================================
+MODELS_DIR="/data/models"
+MODEL_FILE="gemma-4-E2B-it-UD-Q5_K_XL.gguf"
+OLLAMA_MODEL_NAME="my-gemma"
+OLLAMA_URL="http://localhost:11434"
+HEARTBEAT_INTERVAL=120   # إرسال نبضة كل 120 ثانية للإبقاء على النموذج محملاً
+MAX_WAIT=60              # أقصى انتظار لـ Ollama بالثواني
+
+# ============================================================
+# دالة التسجيل الموحّد مع الطابع الزمني
+# ============================================================
+log() {
+    echo "[$(date '+%Y-%m-%d %H:%M:%S')] $*"
+}
+
+# ============================================================
+# إنشاء المجلدات المطلوبة
+# ============================================================
+mkdir -p "$MODELS_DIR" /data/ollama /data/webui
+
+# ============================================================
+# تشغيل Ollama في الخلفية مع إعادة التشغيل التلقائي عند الفشل
+# ============================================================
+start_ollama() {
+    log "Starting Ollama server..."
+    OLLAMA_MODELS=/data/ollama ollama serve >> /data/ollama.log 2>&1 &
+    echo $!
+}
+
+wait_for_ollama() {
+    local waited=0
+    log "Waiting for Ollama to be ready..."
+    until curl -sf "$OLLAMA_URL/api/tags" > /dev/null 2>&1; do
+        if [ $waited -ge $MAX_WAIT ]; then
+            log "ERROR: Ollama did not start within ${MAX_WAIT}s. Check /data/ollama.log"
+            return 1
+        fi
+        sleep 2
+        waited=$((waited + 2))
+    done
+    log "Ollama is ready after ${waited}s."
+}
+
+# ============================================================
+# تنزيل النموذج إن لم يكن موجوداً
+# ============================================================
+download_model() {
+    if [ ! -f "$MODELS_DIR/$MODEL_FILE" ]; then
+        log "Model not found. Downloading from HuggingFace..."
+        huggingface-cli download gijl/gemma-4-E2B-it-GGUF "$MODEL_FILE" \
+            --local-dir "$MODELS_DIR" \
+            --local-dir-use-symlinks False
+
+        huggingface-cli download gijl/gemma-4-E2B-it-GGUF mmproj-BF16.gguf \
+            --local-dir "$MODELS_DIR" \
+            --local-dir-use-symlinks False
+
+        log "Download complete."
+    else
+        log "Model already exists. Skipping download."
+    fi
+}
+
+# ============================================================
+# إنشاء Modelfile وتسجيل النموذج في Ollama
+# ============================================================
+register_model() {
+    # إعادة التسجيل دائماً للتأكد من تطابق الإعدادات
+    log "Creating Modelfile..."
+    cat > "$MODELS_DIR/Modelfile" <<EOF
+FROM $MODELS_DIR/$MODEL_FILE
+
+PARAMETER num_ctx 8192
+PARAMETER num_thread 2
+PARAMETER num_keep -1
+EOF
+
+    log "Registering model with Ollama..."
+    if ollama create "$OLLAMA_MODEL_NAME" -f "$MODELS_DIR/Modelfile"; then
+        log "Model registered successfully."
+    else
+        log "ERROR: Failed to register model."
+        return 1
+    fi
+}
+
+# ============================================================
+# تحميل النموذج في الذاكرة مع keep_alive = -1
+# ============================================================
+preload_model() {
+    log "Preloading model into memory (keep_alive=-1)..."
+    local response
+    response=$(curl -sf -X POST "$OLLAMA_URL/api/generate" \
+        -H "Content-Type: application/json" \
+        -d "{\"model\": \"$OLLAMA_MODEL_NAME\", \"prompt\": \"\", \"keep_alive\": -1}" 2>&1) || true
+
+    if echo "$response" | grep -q "error"; then
+        log "WARNING: Preload returned an error: $response"
+    else
+        log "Model preloaded successfully."
+    fi
+}
+
+# ============================================================
+# دالة النبضة القلبية: تُشغَّل في الخلفية لإبقاء النموذج حياً
+# وتعيد تشغيل Ollama تلقائياً إن سقط
+# ============================================================
+heartbeat_loop() {
+    log "Heartbeat loop started (interval: ${HEARTBEAT_INTERVAL}s)."
+    while true; do
+        sleep "$HEARTBEAT_INTERVAL"
+
+        # التحقق من أن Ollama لا يزال يعمل
+        if ! curl -sf "$OLLAMA_URL/api/tags" > /dev/null 2>&1; then
+            log "WARNING: Ollama is not responding. Restarting..."
+            start_ollama
+            sleep 5
+            wait_for_ollama || continue
+            preload_model
+            continue
+        fi
+
+        # إرسال طلب فارغ للإبقاء على النموذج محملاً في الذاكرة
+        local status
+        status=$(curl -sf -o /dev/null -w "%{http_code}" \
+            -X POST "$OLLAMA_URL/api/generate" \
+            -H "Content-Type: application/json" \
+            -d "{\"model\": \"$OLLAMA_MODEL_NAME\", \"prompt\": \"\", \"keep_alive\": -1}" 2>/dev/null) || status="000"
+
+        if [ "$status" = "200" ]; then
+            log "Heartbeat OK — model is alive in memory."
+        else
+            log "WARNING: Heartbeat failed (HTTP $status). Attempting reload..."
+            preload_model
+        fi
+    done
+}
+
+# ============================================================
+# التنظيف عند إيقاف الحاوية (SIGTERM / SIGINT)
+# ============================================================
+cleanup() {
+    log "Shutdown signal received. Stopping services gracefully..."
+    kill "$HEARTBEAT_PID" 2>/dev/null || true
+    kill "$OLLAMA_PID"    2>/dev/null || true
+    log "Cleanup done. Goodbye."
+    exit 0
+}
+trap cleanup SIGTERM SIGINT
+
+# ============================================================
+# التسلسل الرئيسي للتشغيل
+# ============================================================
+OLLAMA_PID=$(start_ollama)
+wait_for_ollama
+download_model
+register_model
+preload_model
+
+# تشغيل حلقة النبضة في الخلفية
+heartbeat_loop &
+HEARTBEAT_PID=$!
+
+log "All services up. Starting Open WebUI..."
+cd /app/backend
+exec bash start.sh