techjarves · ViperNetwork · Jun 5, 2026
diff --git a/install-core.ps1 b/install-core.ps1
@@ -119,13 +119,28 @@ Write-Host "   PORTABLE AI USB - Multi-Model Setup                    " -Foregro
 Write-Host "==========================================================" -ForegroundColor Cyan
 Write-Host ""
 
-# Show USB free space
+# Show USB free space and system RAM
 $freeGB = Get-USBFreeSpaceGB
 if ($freeGB -gt 0) {
     Write-Host "  USB Free Space: $freeGB GB" -ForegroundColor DarkGray
-    Write-Host ""
 }
 
+$ramGB = -1
+try {
+    $ramGB = [math]::Round((Get-CimInstance -ClassName Win32_ComputerSystem -ErrorAction Stop).TotalPhysicalMemory / 1GB, 0)
+} catch {}
+if ($ramGB -gt 0) {
+    Write-Host "  System RAM    : $ramGB GB" -ForegroundColor DarkGray
+    if ($ramGB -lt 4) {
+        Write-Host "  WARNING: $ramGB GB RAM is insufficient. AI models require at least 4 GB." -ForegroundColor Red
+    } elseif ($ramGB -lt 6) {
+        Write-Host "  NOTE: Only 3B lightweight models recommended on $ramGB GB RAM." -ForegroundColor Yellow
+    } elseif ($ramGB -lt 8) {
+        Write-Host "  NOTE: $ramGB GB RAM is enough for 7B models. NemoMix 12B needs 8 GB." -ForegroundColor Yellow
+    }
+}
+Write-Host ""
+
 # =================================================================
 # STEP 1: MODEL SELECTION MENU
 # =================================================================
@@ -375,7 +390,7 @@ foreach ($m in $SelectedModels) {
             Write-Host "      Retry attempt $attempt..." -ForegroundColor Yellow
         }
 
-        curl.exe -L --ssl-no-revoke --progress-bar $m.URL -o $dest
+        curl.exe -L --progress-bar $m.URL -o $dest
 
         if (Test-DownloadedFile -Path $dest -MinSize $m.MinBytes) {
             $success = $true
@@ -441,7 +456,7 @@ $OllamaDest = "$USB_Drive\ollama\ollama-windows-amd64.zip"
 if (Test-Path "$USB_Drive\ollama\ollama.exe") {
     Write-Host "      Ollama already installed! Skipping..." -ForegroundColor Green
 } else {
-    curl.exe -L --ssl-no-revoke --progress-bar $OllamaURL -o $OllamaDest
+    curl.exe -L --progress-bar $OllamaURL -o $OllamaDest
 
     if (Test-Path $OllamaDest) {
         Write-Host "      Extracting Ollama..." -ForegroundColor Yellow
@@ -477,7 +492,7 @@ if (Test-Path $ExistingApp -PathType Leaf) {
     # Download the installer
     if (-Not (Test-Path $InstallerDest) -or (Get-Item $InstallerDest).Length -lt 10000000) {
         Write-Host "      Downloading installer..." -ForegroundColor Magenta
-        curl.exe -L --ssl-no-revoke --progress-bar $AnythingLLMURL -o $InstallerDest
+        curl.exe -L --progress-bar $AnythingLLMURL -o $InstallerDest
     }
 
     if (Test-Path $InstallerDest) {
@@ -550,7 +565,25 @@ if (-Not (Test-Path "$USB_Drive\ollama\ollama.exe")) {
         $ServerProcess = $null
         try {
             $ServerProcess = Start-Process -FilePath "$USB_Drive\ollama\ollama.exe" -ArgumentList "serve" -WindowStyle Hidden -PassThru
-            Start-Sleep -Seconds 5
+
+            # Poll for readiness (up to 30 s) so import doesn't start against a cold server.
+            Write-Host -NoNewline "      Waiting for engine to initialise"
+            $ollamaReady = $false
+            for ($wi = 0; $wi -lt 30; $wi++) {
+                try {
+                    Invoke-WebRequest -Uri "http://127.0.0.1:11434/api/tags" -UseBasicParsing -TimeoutSec 1 -ErrorAction Stop | Out-Null
+                    $ollamaReady = $true
+                    Write-Host " ready." -ForegroundColor Green
+                    break
+                } catch {
+                    Write-Host -NoNewline "."
+                    Start-Sleep -Seconds 1
+                }
+            }
+            if (-not $ollamaReady) {
+                Write-Host " timeout." -ForegroundColor Yellow
+                Write-Host "      Warning: Ollama did not respond within 30 seconds." -ForegroundColor Yellow
+            }
 
             foreach ($m in $modelsToImport) {
                 Write-Host "      Importing $($m.Name)..." -ForegroundColor Yellow
@@ -602,10 +635,16 @@ if (-Not (Test-Path $envFilePath)) {
     Set-Content -Path $envFilePath -Value $envContent -Force -Encoding UTF8
     Write-Host "      AnythingLLM configured to use: $firstModelLocal" -ForegroundColor Green
 } else {
-    # Update just the model preference if .env already exists
     $existing = Get-Content $envFilePath -Raw
     if ($existing -match 'LLM_PROVIDER=ollama') {
-        Write-Host "      AnythingLLM already configured for Ollama." -ForegroundColor Green
+        # Update OLLAMA_MODEL_PREF without rewriting the rest of the user's config.
+        if ($existing -match '(?m)^OLLAMA_MODEL_PREF=') {
+            $existing = [regex]::Replace($existing, '(?m)^OLLAMA_MODEL_PREF=.*', "OLLAMA_MODEL_PREF=$firstModelLocal")
+        } else {
+            $existing = $existing.TrimEnd() + "`r`nOLLAMA_MODEL_PREF=$firstModelLocal"
+        }
+        Set-Content -Path $envFilePath -Value $existing.TrimEnd() -Force -Encoding UTF8
+        Write-Host "      Updated OLLAMA_MODEL_PREF to: $firstModelLocal" -ForegroundColor Green
     } else {
         # Overwrite with correct config (user was using built-in ollama)
         Set-Content -Path $envFilePath -Value $envContent -Force -Encoding UTF8

diff --git a/linux/install-core.sh b/linux/install-core.sh
@@ -585,7 +585,13 @@ VECTOR_DB=lancedb
 EOF
     echo -e "${GREEN}      AnythingLLM configured to use: ${FIRST_LOCAL}${NC}"
 elif grep -q "LLM_PROVIDER=ollama" "$ENV_FILE"; then
-    echo -e "${GREEN}      AnythingLLM already configured for Ollama.${NC}"
+    # Update OLLAMA_MODEL_PREF without rewriting the rest of the user's config.
+    if grep -q "^OLLAMA_MODEL_PREF=" "$ENV_FILE"; then
+        sed -i "s|^OLLAMA_MODEL_PREF=.*|OLLAMA_MODEL_PREF=${FIRST_LOCAL}|" "$ENV_FILE"
+    else
+        echo "OLLAMA_MODEL_PREF=${FIRST_LOCAL}" >> "$ENV_FILE"
+    fi
+    echo -e "${GREEN}      Updated OLLAMA_MODEL_PREF to: ${FIRST_LOCAL}${NC}"
 else
     cat > "$ENV_FILE" <<EOF
 LLM_PROVIDER=ollama

diff --git a/linux/preflight-check.sh b/linux/preflight-check.sh
@@ -30,6 +30,8 @@ REC_WRITE_MBPS=25
 MIN_READ_MBPS=20
 REC_READ_MBPS=50
 BENCH_SIZE_MB=128
+MIN_RAM_GB=4
+REC_RAM_GB=8
 
 # ── Paths ─────────────────────────────────────────────────────
 SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
@@ -360,13 +362,13 @@ echo -e "${CYAN}  ║       Verifying drive before installation...        ║${N
 echo -e "${CYAN}  ╚══════════════════════════════════════════════════════╝${NC}"
 echo ""
 echo -e "  ${DGRAY}Script Dir   : $SCRIPT_DIR${NC}"
-echo -e "  ${DGRAY}Minimum      : ${MIN_SPACE_GB} GB free  |  Write ≥ ${MIN_WRITE_MBPS} MB/s  |  Read ≥ ${MIN_READ_MBPS} MB/s${NC}"
-echo -e "  ${DGRAY}Recommended  : ${REC_SPACE_GB} GB free  |  Write ≥ ${REC_WRITE_MBPS} MB/s  |  Read ≥ ${REC_READ_MBPS} MB/s${NC}"
+echo -e "  ${DGRAY}Minimum      : ${MIN_SPACE_GB} GB free  |  Write ≥ ${MIN_WRITE_MBPS} MB/s  |  Read ≥ ${MIN_READ_MBPS} MB/s  |  RAM ≥ ${MIN_RAM_GB} GB${NC}"
+echo -e "  ${DGRAY}Recommended  : ${REC_SPACE_GB} GB free  |  Write ≥ ${REC_WRITE_MBPS} MB/s  |  Read ≥ ${REC_READ_MBPS} MB/s  |  RAM ≥ ${REC_RAM_GB} GB${NC}"
 
 # ================================================================
 # STEP 1 — DEPENDENCY CHECK
 # ================================================================
-section "Step 1 / 4 — Required Tools"
+section "Step 1 / 5 — Required Tools"
 echo ""
 
 MISSING_DEPS=()
@@ -388,7 +390,7 @@ fi
 # ================================================================
 # STEP 2 — USB DRIVE DETECTION
 # ================================================================
-section "Step 2 / 4 — USB Drive Detection"
+section "Step 2 / 5 — USB Drive Detection"
 echo ""
 
 if choose_from_mounted_usbs; then
@@ -497,7 +499,7 @@ fi
 # ================================================================
 # STEP 3 — DISK SPACE CHECK
 # ================================================================
-section "Step 3 / 4 — Available Disk Space"
+section "Step 3 / 5 — Available Disk Space"
 echo ""
 
 if [[ -z "$TARGET_MOUNT" || ! -d "$TARGET_MOUNT" ]]; then
@@ -542,9 +544,47 @@ echo -e "  ${DGRAY}│${NC}  ${DGRAY}Ollama engine + AppImage   ~1.0 GB   REQUIR
 echo -e "  ${DGRAY}└───────────────────────────────────────────────────────┘${NC}"
 
 # ================================================================
-# STEP 4 — DRIVE SPEED BENCHMARK
+# STEP 4 — SYSTEM RAM
 # ================================================================
-section "Step 4 / 4 — Drive Speed Benchmark"
+section "Step 4 / 5 — System RAM"
+echo ""
+
+TOTAL_RAM_KB=$(awk '/^MemTotal:/{print $2}' /proc/meminfo 2>/dev/null || echo 0)
+AVAIL_RAM_KB=$(awk '/^MemAvailable:/{print $2}' /proc/meminfo 2>/dev/null || echo 0)
+TOTAL_RAM_GB=$(( TOTAL_RAM_KB / 1024 / 1024 ))
+AVAIL_RAM_GB=$(( AVAIL_RAM_KB / 1024 / 1024 ))
+
+echo -e "  ${DGRAY}┌─ Memory ──────────────────────────────────────────────┐${NC}"
+printf  "  ${DGRAY}│${NC}  %-16s ${BOLD}%d GB${NC}\n" "Total RAM:"  "$TOTAL_RAM_GB"
+printf  "  ${DGRAY}│${NC}  %-16s ${BOLD}%d GB${NC}\n" "Available:"  "$AVAIL_RAM_GB"
+echo -e "  ${DGRAY}└───────────────────────────────────────────────────────┘${NC}"
+echo ""
+
+if (( TOTAL_RAM_GB == 0 )); then
+  result_warn "Could not read system RAM from /proc/meminfo"
+elif (( TOTAL_RAM_GB >= REC_RAM_GB )); then
+  result_pass "RAM: ${TOTAL_RAM_GB} GB — sufficient for all preset models"
+elif (( TOTAL_RAM_GB >= 6 )); then
+  result_warn "RAM: ${TOTAL_RAM_GB} GB — enough for 7B models; NemoMix 12B requires ${REC_RAM_GB} GB"
+elif (( TOTAL_RAM_GB >= MIN_RAM_GB )); then
+  result_warn "RAM: ${TOTAL_RAM_GB} GB — only 3B lightweight models recommended"
+  result_info "7B models need ≥ 6 GB; NemoMix 12B needs ≥ ${REC_RAM_GB} GB"
+else
+  result_fail "RAM: ${TOTAL_RAM_GB} GB — insufficient (minimum ${MIN_RAM_GB} GB required)"
+  result_info "This system may not have enough RAM to run any AI model reliably"
+fi
+
+echo ""
+echo -e "  ${DGRAY}┌─ Model RAM Guide ─────────────────────────────────────┐${NC}"
+echo -e "  ${DGRAY}│${NC}  ${GREEN}Llama 3.2 3B / Phi-3.5 Mini${NC}      ≥ 4 GB RAM"
+echo -e "  ${DGRAY}│${NC}  ${CYAN}Mistral / Qwen / Dolphin 7-8B${NC}     ≥ 6 GB RAM"
+echo -e "  ${DGRAY}│${NC}  ${MAGENTA}NemoMix Unleashed 12B${NC}             ≥ 8 GB RAM"
+echo -e "  ${DGRAY}└───────────────────────────────────────────────────────┘${NC}"
+
+# ================================================================
+# STEP 5 — DRIVE SPEED BENCHMARK
+# ================================================================
+section "Step 5 / 5 — Drive Speed Benchmark"
 echo ""
 echo -e "  ${DGRAY}Using a ${BENCH_SIZE_MB} MB temporary test file on:${NC}"
 echo -e "  ${DGRAY}${TARGET_MOUNT}${NC}"

diff --git a/linux/start-linux.sh b/linux/start-linux.sh
@@ -113,13 +113,49 @@ rm -rf "$ANYTHINGLLM_CACHE/Cache"                 2>/dev/null || true
 rm -rf "$ANYTHINGLLM_CACHE/Code Cache"            2>/dev/null || true
 rm -rf "$ANYTHINGLLM_CACHE/GPUCache"              2>/dev/null || true
 
+# -------------------------------------------------------
+# RAM ADVISORY
+# -------------------------------------------------------
+_RAM_KB=$(awk '/^MemTotal:/{print $2}' /proc/meminfo 2>/dev/null || echo 0)
+_RAM_GB=$(( _RAM_KB / 1024 / 1024 ))
+if (( _RAM_GB > 0 && _RAM_GB < 4 )); then
+    echo -e "${RED}WARNING: Only ${_RAM_GB} GB RAM detected. AI models require at least 4 GB — expect crashes or very slow responses.${NC}"
+elif (( _RAM_GB > 0 && _RAM_GB < 6 )); then
+    echo -e "${YELLOW}NOTE: ${_RAM_GB} GB RAM. 7B+ models need 6 GB; NemoMix 12B needs 8 GB.${NC}"
+fi
+unset _RAM_KB _RAM_GB
+
 # -------------------------------------------------------
 # START OLLAMA ENGINE (background)
 # -------------------------------------------------------
 echo "Starting Ollama Engine..."
 OLLAMA_HOST="127.0.0.1:11434" "$OLLAMA_BIN" serve &>/dev/null &
 OLLAMA_PID=$!
-sleep 3   # give it time to bind
+
+# Poll until the API responds (up to 30 s) instead of a fixed sleep.
+# If Ollama OOMs on start-up it exits before the timeout and we can explain why.
+printf "Waiting for Ollama to be ready"
+_ollama_ready=false
+for (( _i=1; _i<=30; _i++ )); do
+    if curl -sf --max-time 1 "http://127.0.0.1:11434/api/tags" &>/dev/null; then
+        echo " ready."
+        _ollama_ready=true
+        break
+    fi
+    printf "."
+    sleep 1
+done
+if ! $_ollama_ready; then
+    echo " timeout."
+    if ! kill -0 "$OLLAMA_PID" 2>/dev/null; then
+        echo -e "${RED}ERROR: Ollama exited before becoming ready.${NC}"
+        echo "This usually means insufficient RAM for the selected model."
+        echo "NemoMix 12B requires at least 8 GB RAM; 7B models need at least 6 GB."
+        exit 1
+    fi
+    echo -e "${YELLOW}Warning: Ollama did not respond yet — it may still be initialising.${NC}"
+fi
+unset _ollama_ready _i
 
 # -------------------------------------------------------
 # START ANYTHINGLLM (foreground-launched, detached)

diff --git a/start-mac.command b/start-mac.command
@@ -46,9 +46,16 @@ if [ ! -d "$USB_DIR/anythingllm_mac/AnythingLLM.app" ]; then
     echo "First time setup: Downloading AnythingLLM directly to USB..."
     echo "NO installation on the Mac! Everything stays on the drive."
     mkdir -p "$USB_DIR/anythingllm_mac"
-
+
+    # Select the correct build for this Mac's CPU — arm64 = Apple Silicon, x86_64 = Intel
+    if [ "$(uname -m)" = "arm64" ]; then
+        ANYTHINGLLM_DMG_URL="https://cdn.anythingllm.com/latest/AnythingLLMDesktop-Silicon.dmg"
+    else
+        ANYTHINGLLM_DMG_URL="https://cdn.anythingllm.com/latest/AnythingLLMDesktop.dmg"
+    fi
+
     # Download the DMG
-    curl -L --progress-bar "https://cdn.anythingllm.com/latest/AnythingLLMDesktop-Silicon.dmg" -o "$USB_DIR/anythingllm_mac/AnythingLLM_Installer.dmg"
+    curl -L --progress-bar "$ANYTHINGLLM_DMG_URL" -o "$USB_DIR/anythingllm_mac/AnythingLLM_Installer.dmg"
 
     echo "Extracting AnythingLLM to USB (please wait)..."
     # Mount the DMG silently and extract
@@ -73,6 +80,16 @@ fi
 echo ""
 echo "Starting AI Engine from USB..."
 
+# Brief RAM advisory — Ollama OOM exits silently, so warn early
+_RAM_BYTES=$(sysctl -n hw.memsize 2>/dev/null || echo 0)
+_RAM_GB=$(( _RAM_BYTES / 1073741824 ))
+if (( _RAM_GB > 0 && _RAM_GB < 4 )); then
+    echo "WARNING: Only ${_RAM_GB} GB RAM detected. AI models need at least 4 GB."
+elif (( _RAM_GB > 0 && _RAM_GB < 6 )); then
+    echo "NOTE: ${_RAM_GB} GB RAM. 7B+ models need 6 GB; NemoMix 12B needs 8 GB."
+fi
+unset _RAM_BYTES _RAM_GB
+
 # Lock all data paths to the USB drive
 export OLLAMA_MODELS="$DATA_DIR"
 export STORAGE_DIR="$USB_DIR/anythingllm_data"
@@ -124,16 +141,38 @@ if [ -f "$USB_DIR/models/installed-models.txt" ]; then
 fi
 
 # Start Ollama in background
+OLLAMA_PID=""
 if [ -f "$MAC_OLLAMA_DIR/Ollama.app/Contents/MacOS/Ollama" ]; then
     "$MAC_OLLAMA_DIR/Ollama.app/Contents/MacOS/Ollama" serve > /dev/null 2>&1 &
+    OLLAMA_PID=$!
 elif [ -f "$MAC_OLLAMA_DIR/ollama" ]; then
     "$MAC_OLLAMA_DIR/ollama" serve > /dev/null 2>&1 &
+    OLLAMA_PID=$!
 else
     echo "Error: Could not find the Ollama binary on the USB drive!"
 fi
-OLLAMA_PID=$!
 
-sleep 3
+# Poll until the API responds (up to 30 s) instead of a fixed sleep.
+# If Ollama OOMs on start-up it exits before the timeout and we can explain why.
+printf "Waiting for Ollama to be ready"
+for _i in $(seq 1 30); do
+    if curl -sf --max-time 1 "http://127.0.0.1:11434/api/tags" &>/dev/null; then
+        echo " ready."
+        break
+    fi
+    printf "."
+    sleep 1
+    if [ "$_i" -eq 30 ]; then
+        echo " timeout."
+        if [ -n "$OLLAMA_PID" ] && ! kill -0 "$OLLAMA_PID" 2>/dev/null; then
+            echo ""
+            echo "ERROR: Ollama process exited before becoming ready."
+            echo "This usually means insufficient RAM for the selected model."
+            echo "NemoMix 12B requires at least 8 GB RAM; 7B models need at least 6 GB."
+        fi
+    fi
+done
+unset _i
 
 echo ""
 echo "==================================================="

diff --git a/start-windows.bat b/start-windows.bat
@@ -91,8 +91,14 @@ if exist "%~dp0models\installed-models.txt" (
 echo Starting Ollama Engine...
 start "" /B "%~dp0ollama\ollama.exe" serve
 
-:: Give it a few seconds to boot up
-timeout /t 3 >nul
+:: Poll until Ollama accepts connections (up to 30 s) instead of a fixed sleep.
+echo Waiting for Ollama to be ready...
+for /l %%i in (1,1,30) do (
+    curl.exe -sf --max-time 1 http://127.0.0.1:11434/api/tags >nul 2>&1
+    if not errorlevel 1 goto :ollama_ready
+    timeout /t 1 >nul
+)
+:ollama_ready
 
 :: Find and launch AnythingLLM
 echo Starting AnythingLLM Interface...