Update app.py

app.py

@@ -125,7 +125,9 @@ demo = gr.TabbedInterface(
 # Launch Gradio non-blocking on port 7860
 demo.launch(prevent_thread_lock=True)
 
-# 7. FastAPI Schemas & Custom Endpoints attached to demo.app
+# 7. Create an APIRouter for custom endpoints
+api_router = APIRouter()
+
 class EmbedRequest(BaseModel):
     input: Union[str, List[str]]
     input_type: Optional[str] = "document"
@@ -143,7 +145,7 @@ def extract_token(authorization: Optional[str], x_api_key: Optional[str]) -> Opt
         return authorization[7:].strip()
     return None
 
-@demo.app.get("/health")
+@api_router.get("/health")
 def health_check():
     return {
         "status": "online",
@@ -151,7 +153,7 @@ def health_check():
         "endpoints": ["/v1/embeddings", "/v1/systemone", "/v1/decide"]
     }
 
-@demo.app.post("/v1/embeddings")
+@api_router.post("/v1/embeddings")
 def create_embeddings(
     req: EmbedRequest,
     authorization: Optional[str] = Header(default=None),
@@ -167,8 +169,8 @@ def create_embeddings(
         "data": data
     }
 
-@demo.app.post("/v1/systemone")
-@demo.app.post("/v1/decide")
+@api_router.post("/v1/systemone")
+@api_router.post("/v1/decide")
 def create_decision(
     req: DecisionRequest,
     authorization: Optional[str] = Header(default=None),
@@ -179,4 +181,7 @@ def create_decision(
 
     return compute_decision(token, req.state, req.questions)
 
+# Prepend custom routes BEFORE Gradio's catch-all GET /{path:path} route
+demo.app.router.routes = api_router.routes + demo.app.router.routes
+
 demo.block_thread()