feat: switch to persistent SQLite database and improve GPU booking/release handling

- Change database from in-memory to file-based SQLite at aitbc_coordinator.db
- Add status="active" to GPU booking creation
- Allow GPU release even when not properly booked (cleanup case)
- Add error handling for missing booking attributes during refund calculation
- Fix get_gpu_reviews query to use scalars() for proper result handling

docs/trail/GPU_RELEASE_NEXT_STEPS.md

@@ -0,0 +1,216 @@
+# 🎯 GPU Release Fix - Next Steps & Status
+
+## ✅ **COMPLETED STEPS**
+
+### **1. Database Persistence Fixed** ✅
+- ✅ Switched from in-memory SQLite to persistent file
+- ✅ Database file: `/home/oib/windsurf/aitbc/apps/coordinator-api/aitbc_coordinator.db`
+- ✅ Clean database initialization
+
+### **2. Service Management Fixed** ✅
+- ✅ Cleaned up all coordinator processes
+- ✅ Single instance service management
+- ✅ Fresh service start with persistent database
+
+### **3. SQLModel Methods Fixed** ✅
+- ✅ Fixed ALL `session.execute()` → `session.exec()` calls (6 instances)
+- ✅ Fixed GPU booking creation with explicit status
+- ✅ Improved release logic with graceful handling
+
+### **4. GPU Registration Success** ✅
+- ✅ New GPU registered: `gpu_1ea3dcd8`
+- ✅ Clean database without fake entries
+- ✅ Proper GPU details and pricing
+
+### **5. Booking Success** ✅
+- ✅ GPU booking works: `bk_d4df306b8f`
+- ✅ Cost calculation: 0.5 AITBC
+- ✅ Status tracking: "booked"
+
+---
+
+## ❌ **REMAINING ISSUE**
+
+### **GPU Release Still Failing** ❌
+```
+❌ Status: HTTP 500 Internal Server Error
+❌ Error: Failed to release GPU: 500
+❌ GPU Status: Stuck as "booked"
+```
+
+---
+
+## 🔍 **ROOT CAUSE ANALYSIS**
+
+### **Potential Issues:**
+
+#### **1. Import Problems**
+```python
+# Check if SQLModel imports are correct
+from sqlmodel import Session, select, func
+from app.database import engine
+from app.domain.gpu_marketplace import GPURegistry, GPUBooking
+```
+
+#### **2. Database Schema Issues**
+```python
+# Tables might not be created properly
+create_db_and_tables()  # Called on startup
+```
+
+#### **3. Missing Dependencies**
+```python
+# Check if all required imports are available
+from sqlalchemy import func  # Used in review calculations
+```
+
+#### **4. Session Transaction Issues**
+```python
+# Session might not be properly committed
+session.commit()  # Check if this is working
+```
+
+---
+
+## 🛠️ **DEBUGGING NEXT STEPS**
+
+### **Step 1: Check Error Logs**
+```bash
+# Get detailed error logs
+curl -v http://localhost:8000/v1/marketplace/gpu/gpu_1ea3dcd8/release
+
+# Check coordinator logs
+journalctl -u aitbc-coordinator --since "1 minute ago"
+```
+
+### **Step 2: Test Database Directly**
+```python
+# Create debug script to test database operations
+python3 scripts/debug_database_operations.py
+```
+
+### **Step 3: Check Imports**
+```python
+# Verify all imports work correctly
+python3 -c "from app.domain.gpu_marketplace import GPURegistry, GPUBooking"
+```
+
+### **Step 4: Manual Database Test**
+```python
+# Test release logic manually in Python REPL
+python3 scripts/test_release_logic.py
+```
+
+---
+
+## 🚀 **IMMEDIATE ACTIONS**
+
+### **High Priority:**
+1. **Debug the 500 error** - Get detailed error message
+2. **Check database schema** - Verify tables exist
+3. **Test imports** - Ensure all modules load correctly
+
+### **Medium Priority:**
+1. **Create debug script** - Test database operations directly
+2. **Add logging** - More detailed error messages
+3. **Manual testing** - Test release logic in isolation
+
+---
+
+## 📋 **WORKING SOLUTIONS**
+
+### **Current Working Features:**
+- ✅ GPU Registration
+- ✅ GPU Listing
+- ✅ GPU Booking
+- ✅ Database Persistence
+- ✅ Service Management
+
+### **Broken Features:**
+- ❌ GPU Release (HTTP 500)
+
+---
+
+## 🎯 **EXPECTED OUTCOME**
+
+### **When Fixed Should See:**
+```bash
+aitbc marketplace gpu release gpu_1ea3dcd8
+# Expected Response:
+{
+  "status": "released",
+  "gpu_id": "gpu_1ea3dcd8", 
+  "refund": 0.25,
+  "message": "GPU gpu_1ea3dcd8 released successfully"
+}
+```
+
+### **GPU Status Should Change:**
+```bash
+aitbc marketplace gpu list
+# Expected: GPU status = "available" (not "booked")
+```
+
+---
+
+## 📊 **PROGRESS SUMMARY**
+
+| Phase | Status | Notes |
+|-------|--------|-------|
+| Database Persistence | ✅ COMPLETE | Persistent SQLite working |
+| Service Management | ✅ COMPLETE | Single instance running |
+| SQLModel Fixes | ✅ COMPLETE | All 6 instances fixed |
+| GPU Registration | ✅ COMPLETE | New GPU registered |
+| GPU Booking | ✅ COMPLETE | Booking working |
+| GPU Release | ❌ IN PROGRESS | HTTP 500 error persists |
+
+**Overall Progress: 83% Complete**
+
+---
+
+## 🔄 **NEXT EXECUTION PLAN**
+
+### **Immediate (Next 10 minutes):**
+1. Get detailed error logs for 500 error
+2. Check database schema and imports
+3. Create debug script for release logic
+
+### **Short-term (Next 30 minutes):**
+1. Fix the root cause of 500 error
+2. Test complete booking/release cycle
+3. Verify GPU status changes properly
+
+### **Long-term (Next hour):**
+1. Clean up any remaining fake GPUs
+2. Test edge cases and error handling
+3. Document the complete solution
+
+---
+
+## 💡 **KEY INSIGHTS**
+
+### **What We've Learned:**
+1. **SQLModel Method Names**: `session.exec()` not `session.execute()`
+2. **Database Persistence**: In-memory SQLite causes data loss
+3. **Service Management**: Multiple processes cause conflicts
+4. **Booking Creation**: Explicit status field required
+
+### **What Still Needs Work:**
+1. **Error Handling**: Need better error messages
+2. **Debugging**: More detailed logging required
+3. **Testing**: Comprehensive endpoint testing needed
+
+---
+
+## 🎉 **SUCCESS METRICS**
+
+### **When Complete:**
+- ✅ GPU Release returns HTTP 200
+- ✅ GPU status changes from "booked" to "available"
+- ✅ Refund calculation works correctly
+- ✅ Complete booking/release cycle functional
+- ✅ No fake GPU entries in database
+
+---
+
+**The foundation is solid - we just need to identify and fix the specific cause of the 500 error in the release endpoint.**