feat: Enable RoentGen-v2 synthetic X-ray generation

Added RoentGen-v2 (Stanford MIMI) for generating synthetic chest X-rays
from text descriptions.

Changes:
- Updated xray_generation.py to use stanfordmimi/RoentGen-v2 from HF
- Changed from fp32 to fp16 for memory efficiency
- Added tool loading in app.py for GPU environments
- Now publicly available (no need for manual weight download)

Usage examples:
- "Generate an X-ray showing pneumonia in the right lower lobe"
- "Create a chest X-ray with left-sided pleural effusion"
- "Generate a normal chest radiograph"

Tool count: 7 → 8 (with RoentGen-v2 generation)

Paper: https://arxiv.org/abs/2508.16783
Model: https://huggingface.co/stanfordmimi/RoentGen-v2

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

Co-Authored-By: Claude <noreply@anthropic.com>

app.py

@@ -126,6 +126,20 @@ try:
 except Exception as e:
     print(f"✗ Failed to load web browsing tool: {e}")
 
+# RoentGen-v2 X-ray Generation (GPU only)
+if device == "cuda":
+    try:
+        from medrax.tools.xray_generation import ChestXRayGeneratorTool
+        generation_tool = ChestXRayGeneratorTool(
+            model_path="stanfordmimi/RoentGen-v2",
+            temp_dir="temp",
+            device=device
+        )
+        tools.append(generation_tool)
+        print("✓ Loaded RoentGen-v2 X-ray generation tool")
+    except Exception as e:
+        print(f"✗ Failed to load RoentGen-v2 tool: {e}")
+
 checkpointer = MemorySaver()
 
 llm = ModelFactory.create_model(

medrax/tools/xray_generation.py

@@ -42,8 +42,8 @@ class ChestXRayGeneratorTool(BaseTool):
 
     def __init__(
         self,
-        model_path: str = "/model-weights/roentgen",
-        cache_dir: str = "/model-weights",
+        model_path: str = "stanfordmimi/RoentGen-v2",
+        cache_dir: Optional[str] = None,
         temp_dir: Optional[str] = None,
         device: Optional[str] = "cuda",
     ):
@@ -51,8 +51,14 @@ class ChestXRayGeneratorTool(BaseTool):
         super().__init__()
 
         self.device = torch.device(device) if device else "cuda"
-        self.model = StableDiffusionPipeline.from_pretrained(model_path, cache_dir=cache_dir)
-        self.model = self.model.to(torch.float32).to(self.device)
+
+        # Load RoentGen-v2 from Hugging Face
+        self.model = StableDiffusionPipeline.from_pretrained(
+            model_path,
+            cache_dir=cache_dir,
+            torch_dtype=torch.float16,  # Use fp16 for efficiency
+        )
+        self.model = self.model.to(self.device)
 
         self.temp_dir = Path(temp_dir if temp_dir else tempfile.mkdtemp())
         self.temp_dir.mkdir(exist_ok=True)