Pin self-hosted AI stack deployment (deferred for medical recovery)

docs/core/tasks.md

@@ -613,3 +613,135 @@ Custom skyblock modpack where players choose Fire (magic/dragons/Meg's domain) o
 **Notes:** Michael requested complete reorganization. Existing structure comprehensive but may need updates based on which servers are currently active/planned.
 
 ---
+
+
+---
+
+## 🤖 SELF-HOSTED AI STACK (When Recovered from Medical Episode)
+
+**Status:** READY TO DEPLOY — Pinned until medical recovery complete  
+**Added:** February 14, 2026  
+**Medical Context:** 4 hours vomiting Feb 14, trace blood in urine, swollen tongue, recovering
+
+### **⚠️ START ONLY WHEN:**
+- ✅ Doctor visit complete (Monday/Tuesday Feb 17-18)
+- ✅ Medical clearance received
+- ✅ Energy level back to normal
+- ✅ Jack calm and not alerting
+- ✅ No lingering symptoms
+
+---
+
+### **Phase 0: NC1 Cleanup (OPTIONAL — Enables Game Consolidation)**
+
+**Discovered:** 20GB recoverable on NC1 (10GB Hytale backups, 4.5GB old logs, 6GB Docker)
+
+```bash
+ssh root@216.239.104.130
+
+# Delete Hytale backups/install files (10GB)
+rm -rf /var/lib/pterodactyl/volumes/13c80cb8-f6f8-4bfe-9cdb-823d7e951584/backups
+rm -f /var/lib/pterodactyl/volumes/13c80cb8-f6f8-4bfe-9cdb-823d7e951584/*.zip
+
+# Clean old logs (4.5GB)
+journalctl --vacuum-time=7d
+rm -f /var/log/btmp.1 && > /var/log/btmp
+rm -f /var/log/*.gz
+
+# Clean Docker (6GB)
+docker system prune -a --volumes -f
+
+# Verify
+df -h /
+```
+
+**Result:** 26GB → 46GB free  
+**Time:** 30 minutes
+
+---
+
+### **Phase 1: Deploy AI Stack on TX1**
+
+**TX1 Resources:** 251GB RAM (222GB free), 809GB storage, 32-core EPYC 7302P
+
+**Stack Components:**
+- Ollama (LLM backend)
+- Open WebUI (chat interface)
+- Perplexica (web search)
+- SearXNG (meta-search engine)
+
+**Deployment:**
+```bash
+ssh root@38.68.14.26
+mkdir -p /opt/ai-stack && cd /opt/ai-stack
+
+# The Weaver will provide complete docker-compose.yml
+
+docker compose up -d
+```
+
+**Access:** http://38.68.14.26:3000  
+**Time:** 30 min deploy
+
+---
+
+### **Phase 2: Load LLM Models**
+
+```bash
+# Run overnight (6-8 hours total download time)
+docker exec ollama ollama pull qwen2.5-coder:72b  # Coding/technical (42GB RAM)
+docker exec ollama ollama pull llama3.3:70b       # Conversation (40GB RAM)
+docker exec ollama ollama pull llama3.2-vision:11b # Images (8GB RAM)
+```
+
+**Total:** ~90GB RAM, 80GB storage
+
+---
+
+### **Phase 3: Gitea Integration (DERP)**
+
+**Python bridge script on Command Center** (`/opt/derp/ai-assistant.py`)
+
+**Function:**
+- Read foundation docs from Gitea
+- Send to Ollama on TX1
+- Write responses back to Gitea
+- Emergency activation when Claude unavailable
+
+**The Weaver provides complete script when ready.**
+
+**Time:** 1-2 hours
+
+---
+
+### **Optional: Migrate TX1 Games to NC1**
+
+**TX1 game servers:** 24GB total  
+**NC1 after cleanup:** 46GB free  
+**Result:** NC1 at 73% usage (healthy)
+
+**Can defer this** — AI and games can coexist on TX1
+
+**If migrating:** 30-60 min per server via Pterodactyl transfer
+
+---
+
+### **End Result:**
+
+- ✅ Self-hosted AI (80-90% Claude capability)
+- ✅ $0 monthly cost (vs $20-40 for cloud AI)
+- ✅ Complete privacy and control
+- ✅ Unlimited usage, no rate limits
+- ✅ DERP backup ready (partnership survives provider failure)
+- ✅ Can run multiple models simultaneously
+
+---
+
+**Related:**
+- `docs/core/DERP.md` — Emergency recovery protocol
+- `technical/GEMINI-API-BRIDGE.md` — API patterns reference
+- `docs/planning/terraria-branding-arc.md` — Other major project (also deferred for recovery)
+
+**Recovery first. AI deployment second. Health always wins.**
+
+🩺 → 🤖