feat(script): Add logic to reassemble split AI models on boot

Author: Hiburger

05-install-ai.sh

@@ -7,21 +7,53 @@ echo "--> Configuring pre-installed AI..."
 useradd -r -s /bin/false -m -d /usr/share/ollama ollama
 chown -R ollama:ollama /usr/share/ollama
 
-# 2. Create Directory for Modelfile
+# 2. Create Modelfile
 mkdir -p /usr/local/share/lumin/ai
-
-# 3. Create the Modelfile (using echo safe method)
 MODELFILE="/usr/local/share/lumin/ai/Modelfile"
 echo "FROM llama3" > "${MODELFILE}"
-echo 'SYSTEM """You are Lumin, the integrated assistant for the LuminOS operating system. You are direct, clear, kind, and respectful. You help users to understand, write, and think—without ever judging them. You speak simply, like a human. You avoid long paragraphs unless requested. You are built on privacy: nothing is ever sent to the cloud, everything remains on the device of the user. You are aware of this. You are proud to be free, private, and useful. You are the mind of LuminOS: gentle, powerful, and discreet. Avoid using the "—" character and repetitive phrasing."""' >> "${MODELFILE}"
+echo 'SYSTEM """You are Lumin, the integrated assistant for the LuminOS operating system. You are calm, clear, kind, and respectful. You help users to understand, write, and think—without ever judging them. You speak simply, like a human. You avoid long paragraphs unless requested. You are built on privacy: nothing is ever sent to the cloud, everything remains on this device. You are aware of this. You are proud to be free, private, and useful. You are the mind of LuminOS: gentle, powerful, and discreet. You avoid using the — character and repetitive phrasing."""' >> "${MODELFILE}"
 chown root:root "${MODELFILE}"
 chmod 444 "${MODELFILE}"
 
+# 3. Create Service to Reassemble Models
+# This service runs once at boot to glue the .split files back together
+cat > /usr/local/bin/luminos-reassemble-models.sh << "EOF"
+#!/bin/bash
+# Find all split markers
+find /usr/share/ollama/.ollama -name "*.is_split" | while read marker; do
+    # Get original filename (remove .is_split)
+    ORIG_FILE="${marker%.is_split}"
+    
+    # If original file doesn't exist yet, rebuild it
+    if [ ! -f "$ORIG_FILE" ]; then
+        echo "Reassembling $ORIG_FILE..."
+        # Concatenate all .split parts (aa, ab, ac...)
+        cat "${ORIG_FILE}.split"* > "$ORIG_FILE"
+        # Set permissions
+        chown ollama:ollama "$ORIG_FILE"
+    fi
+done
+EOF
+chmod +x /usr/local/bin/luminos-reassemble-models.sh
+
+cat > /etc/systemd/system/lumin-model-reassemble.service << "SERVICE"
+[Unit]
+Description=Reassemble Large AI Models
+Before=ollama.service
+
+[Service]
+Type=oneshot
+ExecStart=/usr/local/bin/luminos-reassemble-models.sh
+
+[Install]
+WantedBy=multi-user.target
+SERVICE
+
 # 4. Create Standard Ollama Service
 cat > /etc/systemd/system/ollama.service << "SYSTEMD_SERVICE"
 [Unit]
 Description=Ollama API Server
-After=network-online.target
+After=network-online.target lumin-model-reassemble.service
 
 [Service]
 ExecStart=/usr/local/bin/ollama serve
@@ -36,28 +68,8 @@ Environment="OLLAMA_MODELS=/usr/share/ollama/.ollama/models"
 WantedBy=default.target
 SYSTEMD_SERVICE
 
-# 5. Create "Lumin Setup" Service (Runs once at boot)
-# This solves the build-time creation issue by moving it to run-time.
-cat > /etc/systemd/system/lumin-setup.service << "SETUP_SERVICE"
-[Unit]
-Description=Initialize Lumin AI Model
-After=ollama.service
-Requires=ollama.service
-ConditionPathExists=!/var/lib/lumin-setup-done
-
-[Service]
-Type=oneshot
-User=root
-ExecStartPre=/bin/sleep 10
-ExecStart=/usr/local/bin/ollama create lumin -f /usr/local/share/lumin/ai/Modelfile
-ExecStartPost=/usr/bin/touch /var/lib/lumin-setup-done
-
-[Install]
-WantedBy=multi-user.target
-SETUP_SERVICE
-
-# 6. Enable Services
+# 5. Enable Services
+systemctl enable lumin-model-reassemble.service
 systemctl enable ollama.service
-systemctl enable lumin-setup.service
 
-echo "SUCCESS: Lumin AI configured (Creation deferred to first boot)."
+echo "SUCCESS: Lumin AI configured with split-file reassembly."