ai: Prepare stable version for 2.1.1-ft-QwQ-32B.

jarvis.py

@@ -147,21 +147,28 @@ def process_ai_response(ai_text):
         return ai_text
 
 async def fetch_response_async(host, provider_key, selected_model, messages, model_config, session_id):
-    try:
-        async with httpx.AsyncClient(timeout=5) as client:
-            data = {"model": selected_model, "messages": messages, **model_config}
-            extra = {"optillm_approach": "rto|re2|cot_reflection|self_consistency|plansearch|leap|z3|bon|moa|mcts|mcp|router|privacy|executecode|json", "session_id": session_id}
-            response = await client.post(f"{host}", json={**data, "extra_body": extra, "session_id": session_id}, headers={"Authorization": f"Bearer {provider_key}"})
-            response.raise_for_status()
-            try:
-                resp_json = response.json()
-            except json.JSONDecodeError:
+    timeouts = [60, 80, 120, 240]
+    for timeout in timeouts:
+        try:
+            async with httpx.AsyncClient(timeout=timeout) as client:
+                data = {"model": selected_model, "messages": messages, **model_config}
+                extra = {"optillm_approach": "rto|re2|cot_reflection|self_consistency|plansearch|leap|z3|bon|moa|mcts|mcp|router|privacy|executecode|json", "session_id": session_id}
+                resp = await client.post(host, json={**data, "extra_body": extra, "session_id": session_id}, headers={"Authorization": f"Bearer {provider_key}"})
+                resp.raise_for_status()
+                try:
+                    resp_json = resp.json()
+                except json.JSONDecodeError:
+                    return RESPONSES["RESPONSE_2"]
+                if isinstance(resp_json, dict) and "choices" in resp_json and isinstance(resp_json["choices"], list) and len(resp_json["choices"]) > 0 and isinstance(resp_json["choices"][0], dict):
+                    choice = resp_json["choices"][0]
+                    if "message" in choice and isinstance(choice["message"], dict) and "content" in choice["message"] and isinstance(choice["message"]["content"], str):
+                        ai_text = choice["message"]["content"]
+                        return process_ai_response(ai_text)
                 return RESPONSES["RESPONSE_2"]
-            ai_text = resp_json["choices"][0]["message"]["content"] if resp_json.get("choices") and resp_json["choices"][0].get("message") and resp_json["choices"][0]["message"].get("content") else RESPONSES["RESPONSE_2"]
-            return process_ai_response(ai_text)
-    except Exception:
-        marked_item(provider_key, LINUX_SERVER_PROVIDER_KEYS_MARKED, LINUX_SERVER_PROVIDER_KEYS_ATTEMPTS)
-        return RESPONSES["RESPONSE_2"]
+        except Exception:
+            continue
+    marked_item(provider_key, LINUX_SERVER_PROVIDER_KEYS_MARKED, LINUX_SERVER_PROVIDER_KEYS_ATTEMPTS)
+    return RESPONSES["RESPONSE_2"]
 
 async def chat_with_model_async(history, user_input, selected_model_display, sess):
     if not get_available_items(LINUX_SERVER_PROVIDER_KEYS, LINUX_SERVER_PROVIDER_KEYS_MARKED) or not get_available_items(LINUX_SERVER_HOSTS, LINUX_SERVER_HOSTS_MARKED):
@@ -208,8 +215,6 @@ async def respond_async(multi_input, history, selected_model_display, sess):
     if message["text"]:
         combined_input += message["text"]
     history.append([combined_input, ""])
-    if len(history) > 3:
-        history[:] = history[-3:]
     ai_response = await chat_with_model_async(history, combined_input, selected_model_display, sess)
     history[-1][1] = ""
     def convert_to_string(data):