Update app.py

app.py

@@ -337,40 +337,91 @@ def normalize_assistant_text(x):
 # Query wrapper: tries modern chat, responses, legacy shapes
 
 def query_openai(system_prompt: str, user_message: str, max_tokens=800, temperature=0.2):
+    """
+    Robust OpenAI query that supports:
+    - modern client.chat.create(...)
+    - legacy openai.ChatCompletion.create(...)
+    - client.responses.create(...) (tries several kwarg combinations)
+    - older Completion.create(...)
+    Returns {"text": "..."} on success or {"error": "..."} on failure.
+    """
     client, err = make_openai_client(OPENAI_api_key)
     if err:
         return {"error": err}
+
     model = OPENAI_model or "gpt-4o-mini"
+
     try:
-        # modern client.chat.create
+        # 1) Modern OpenAI SDK: client.chat.create(...)
         if hasattr(client, "chat") and hasattr(client.chat, "create"):
             resp = client.chat.create(
                 model=model,
-                messages=[{"role": "system", "content": system_prompt}, {"role": "user", "content": user_message}],
+                messages=[
+                    {"role": "system", "content": system_prompt},
+                    {"role": "user", "content": user_message},
+                ],
                 temperature=temperature,
                 max_tokens=max_tokens,
             )
             return {"text": extract_text_from_response(resp)}
-        # legacy openai.ChatCompletion
-        if hasattr(client, "ChatCompletion"):
+
+        # 2) Legacy openai package: ChatCompletion.create(...)
+        if hasattr(client, "ChatCompletion") and hasattr(client.ChatCompletion, "create"):
             resp = client.ChatCompletion.create(
                 model=model,
-                messages=[{"role": "system", "content": system_prompt}, {"role": "user", "content": user_message}],
+                messages=[
+                    {"role": "system", "content": system_prompt},
+                    {"role": "user", "content": user_message},
+                ],
                 temperature=temperature,
                 max_tokens=max_tokens,
             )
             return {"text": extract_text_from_response(resp)}
-        # responses API
+
+        # 3) Responses API: client.responses.create(...)
         if hasattr(client, "responses") and hasattr(client.responses, "create"):
             prompt = system_prompt + "\n\n" + user_message
-            resp = client.responses.create(model=model, input=prompt, max_tokens=max_tokens, temperature=temperature)
+            response = None
+            last_exc = None
+
+            # Try several kwarg shapes because different SDK versions accept different names
+            candidate_kw_sets = [
+                {"model": model, "input": prompt, "max_output_tokens": max_tokens, "temperature": temperature},
+                {"model": model, "input": prompt, "max_tokens": max_tokens, "temperature": temperature},
+                {"model": model, "input": prompt, "temperature": temperature},
+                {"model": model, "input": prompt},
+            ]
+
+            for kw in candidate_kw_sets:
+                try:
+                    response = client.responses.create(**kw)
+                    break
+                except TypeError as te:
+                    # argument mismatch for this kw set — try next
+                    last_exc = te
+                    continue
+                except Exception as e:
+                    # other runtime errors (auth/quota) -> return as error
+                    return {"error": f"OpenAI responses request failed: {e}"}
+
+            if response is None:
+                return {"error": f"OpenAI responses client didn't accept tried args. Last error: {last_exc}"}
+
+            return {"text": extract_text_from_response(response)}
+
+        # 4) Very old Completion API fallback
+        if hasattr(client, "Completion") and hasattr(client.Completion, "create"):
+            prompt = system_prompt + "\n\n" + user_message
+            resp = client.Completion.create(model=model, prompt=prompt, max_tokens=max_tokens, temperature=temperature)
             return {"text": extract_text_from_response(resp)}
-        # fallback
-        return {"error": "OpenAI client present but no compatible chat method found."}
+
+        return {"error": "OpenAI client present but no compatible chat method found. Consider upgrading the openai package."}
+
     except Exception as e:
         return {"error": f"OpenAI request failed: {e}"}
 
 
+
 # System prompt builder that respects UI language
 
 def make_system_prompt(wound_type: str, victim_info: dict) -> str: