fix: Fix model auto-detection test and add CI debug script

- Initialize model registry before creating AmbiguityResolver
- Add debug script to check CI environment and API keys
- Fix test to properly handle model registry initialization
- Add more detailed error reporting in test

This should fix the failing CI tests by ensuring models are properly
initialized before attempting auto-detection.

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <[email protected]>

Author: jeremymanning

.github/workflows/test-models.yml

@@ -27,6 +27,10 @@ jobs:
         pip install -e .
         pip install transformers torch requests
     
+    - name: Debug CI Environment
+      run: |
+        python test_ci_env.py || true
+    
     - name: Test HuggingFace model integration
       env:
         CI: true

test_ci_env.py

@@ -0,0 +1,29 @@
+#!/usr/bin/env python3
+"""Simple test to verify CI environment."""
+
+import os
+import sys
+
+# Just print environment info
+print("=== CI Environment Test ===")
+print(f"CI={os.environ.get('CI', 'not set')}")
+print(f"GITHUB_ACTIONS={os.environ.get('GITHUB_ACTIONS', 'not set')}")
+
+# Check API keys (don't print values)
+for key in ["OPENAI_API_KEY", "ANTHROPIC_API_KEY", "GOOGLE_AI_API_KEY", "HF_TOKEN"]:
+    value = os.environ.get(key)
+    if value:
+        print(f"{key}: SET (len={len(value)})")
+    else:
+        print(f"{key}: NOT SET")
+
+# Exit with error if no keys found
+keys_found = sum(1 for k in ["OPENAI_API_KEY", "ANTHROPIC_API_KEY", "GOOGLE_AI_API_KEY", "HF_TOKEN"] if os.environ.get(k))
+print(f"\nTotal keys found: {keys_found}")
+
+if keys_found == 0:
+    print("ERROR: No API keys found!")
+    sys.exit(1)
+else:
+    print("SUCCESS: API keys detected")
+    sys.exit(0)

tests/integration/test_models_comprehensive.py

@@ -132,19 +132,39 @@ async def test_model_auto_detection():
 
     try:
         from orchestrator.compiler.ambiguity_resolver import AmbiguityResolver
-
-        # Create resolver without specifying model (should auto-detect)
-        resolver = AmbiguityResolver()
-        print(f"✅ Auto-detected model: {resolver.model.name}")
-        print(f"   Provider: {resolver.model.provider}")
-
-        # Test simple resolution
-        resolved = await resolver.resolve("Choose format", "test.format")
-        print(f"✅ Resolved 'Choose format': '{resolved}'")
-        return True
+        from orchestrator import init_models
+        from orchestrator.models.registry_singleton import get_model_registry
+
+        # Initialize models first
+        init_models()
+        registry = get_model_registry()
+        
+        # Check if we have any models available
+        available_models = registry.list_models()
+        if not available_models:
+            print("❌ No models available in registry after init_models()")
+            return False
+            
+        print(f"✅ Found {len(available_models)} models in registry")
+
+        # Create resolver with model registry (should auto-detect)
+        resolver = AmbiguityResolver(model_registry=registry)
+        print(f"✅ Auto-detected model: {resolver.model.name if resolver.model else 'None'}")
+        if resolver.model:
+            print(f"   Provider: {resolver.model.provider}")
+
+            # Test simple resolution
+            resolved = await resolver.resolve("Choose format", "test.format")
+            print(f"✅ Resolved 'Choose format': '{resolved}'")
+            return True
+        else:
+            print("❌ No model was auto-detected")
+            return False
 
     except Exception as e:
         print(f"❌ Auto-detection test failed: {e}")
+        import traceback
+        traceback.print_exc()
         return False