ai: Back to 5 minute of cool down.

jarvis.py

@@ -34,6 +34,8 @@ LINUX_SERVER_PROVIDER_KEYS = [key for key in json.loads(os.getenv("LINUX_SERVER_
 LINUX_SERVER_PROVIDER_KEYS_MARKED = set()
 LINUX_SERVER_PROVIDER_KEYS_ATTEMPTS = {}
 
+LINUX_SERVER_ERRORS = set(map(int, os.getenv("LINUX_SERVER_ERROR").split(",")))
+
 AI_TYPES = {f"AI_TYPE_{i}": os.getenv(f"AI_TYPE_{i}") for i in range(1, 7)}
 RESPONSES = {f"RESPONSE_{i}": os.getenv(f"RESPONSE_{i}") for i in range(1, 10)}
 
@@ -60,7 +62,7 @@ def marked_item(item, marked, attempts):
         def remove_fail():
             marked.discard(item)
             attempts.pop(item, None)
-        threading.Timer(3600, remove_fail).start()
+        threading.Timer(300, remove_fail).start()
 
 class SessionWithID(requests.Session):
     def __init__(self):
@@ -114,19 +116,21 @@ async def fetch_response_async(host, provider_key, selected_model, messages, mod
     for timeout in timeouts:
         try:
             async with httpx.AsyncClient(timeout=timeout) as client:
-                data = {"model": selected_model, "messages": messages, **model_config}
-                resp = await client.post(host, json={**data, "session_id": session_id}, headers={"Authorization": f"Bearer {provider_key}"})
+                resp = await client.post(host, json={"model": selected_model, "messages": messages, **model_config, "session_id": session_id}, headers={"Authorization": f"Bearer {provider_key}"})
+                if resp.status_code in LINUX_SERVER_ERRORS:
+                    marked_item(provider_key, LINUX_SERVER_PROVIDER_KEYS_MARKED, LINUX_SERVER_PROVIDER_KEYS_ATTEMPTS)
+                    return None
                 resp.raise_for_status()
                 resp_json = resp.json()
                 if isinstance(resp_json, dict) and resp_json.get("choices"):
                     choice = resp_json["choices"][0]
                     if choice.get("message") and isinstance(choice["message"].get("content"), str):
                         return choice["message"]["content"]
-                return RESPONSES["RESPONSE_2"]
+                return None
         except Exception:
             continue
     marked_item(provider_key, LINUX_SERVER_PROVIDER_KEYS_MARKED, LINUX_SERVER_PROVIDER_KEYS_ATTEMPTS)
-    return RESPONSES["RESPONSE_2"]
+    return None
 
 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):
@@ -142,7 +146,7 @@ async def chat_with_model_async(history, user_input, selected_model_display, ses
     global ACTIVE_CANDIDATE
     if ACTIVE_CANDIDATE:
         result = await fetch_response_async(ACTIVE_CANDIDATE[0], ACTIVE_CANDIDATE[1], selected_model, messages, model_config, sess.session_id)
-        if result != RESPONSES["RESPONSE_2"]:
+        if result:
             return result
         ACTIVE_CANDIDATE = None
     keys = get_available_items(LINUX_SERVER_PROVIDER_KEYS, LINUX_SERVER_PROVIDER_KEYS_MARKED)
@@ -151,7 +155,7 @@ async def chat_with_model_async(history, user_input, selected_model_display, ses
     random.shuffle(candidates)
     for host, key in candidates:
         result = await fetch_response_async(host, key, selected_model, messages, model_config, sess.session_id)
-        if result != RESPONSES["RESPONSE_2"]:
+        if result:
             ACTIVE_CANDIDATE = (host, key)
             return result
     return RESPONSES["RESPONSE_2"]
@@ -197,6 +201,6 @@ with gr.Blocks(fill_height=True, fill_width=True, title=AI_TYPES["AI_TYPE_4"], h
         msg = gr.MultimodalTextbox(show_label=False, placeholder=RESPONSES["RESPONSE_5"], interactive=True, file_count="single", file_types=ALLOWED_EXTENSIONS)
     with gr.Accordion(AI_TYPES["AI_TYPE_6"], open=False):
         model_dropdown = gr.Dropdown(show_label=False, choices=MODEL_CHOICES, value=MODEL_CHOICES[0])
-    model_dropdown.change(fn=change_model, inputs=[model_dropdown], outputs=[user_history, user_session, selected_model], show_progress="full")
+    model_dropdown.change(fn=change_model, inputs=[model_dropdown], outputs=[user_history, user_session, selected_model])
     msg.submit(fn=respond_async, inputs=[msg, user_history, selected_model, user_session], outputs=[chatbot, msg, user_session], api_name=INTERNAL_AI_GET_SERVER)
 jarvis.launch(max_file_size="1mb")