Update app.py

app.py

@@ -69,13 +69,29 @@ def chat_simple(req: SimpleChatRequest):
             top_p=req.top_p,
         )
 
-        reply = response.choices[0].message.content if response.choices else ""
+        choice = response.choices[0]
+
+        reply = ""
+        if choice.message and choice.message.content:
+            reply = choice.message.content
+
+        finish_reason = getattr(choice, "finish_reason", None)
+
+        usage = None
+        if getattr(response, "usage", None):
+            usage = {
+                "prompt_tokens": response.usage.prompt_tokens,
+                "completion_tokens": response.usage.completion_tokens,
+                "total_tokens": response.usage.total_tokens,
+            }
+
         return JSONResponse(
             {
                 "ok": True,
                 "model": MODEL_NAME,
                 "reply": reply,
-                "usage": response.usage.model_dump() if getattr(response, "usage", None) else None,
+                "finish_reason": finish_reason,
+                "usage": usage,
             }
         )
 
@@ -94,13 +110,29 @@ def chat(req: ChatRequest):
             top_p=req.top_p,
         )
 
-        reply = response.choices[0].message.content if response.choices else ""
+        choice = response.choices[0]
+
+        reply = ""
+        if choice.message and choice.message.content:
+            reply = choice.message.content
+
+        finish_reason = getattr(choice, "finish_reason", None)
+
+        usage = None
+        if getattr(response, "usage", None):
+            usage = {
+                "prompt_tokens": response.usage.prompt_tokens,
+                "completion_tokens": response.usage.completion_tokens,
+                "total_tokens": response.usage.total_tokens,
+            }
+
         return JSONResponse(
             {
                 "ok": True,
                 "model": MODEL_NAME,
                 "reply": reply,
-                "usage": response.usage.model_dump() if getattr(response, "usage", None) else None,
+                "finish_reason": finish_reason,
+                "usage": usage,
             }
         )