diff --git a/README.md b/README.md
index c460eac..8c949de 100644
--- a/README.md
+++ b/README.md
@@ -142,12 +142,16 @@ Transform your smart home experience with powerful AI assistance powered by mult
 - **Claude Haiku 4.5** - The fastest and most economical option in the series
 
 #### DeepSeek Models
-- **DeepSeek-V3** - A general-purpose model for a wide range of tasks
-- **DeepSeek-R1** - A specialized model focused on reasoning and coding
+- **deepseek-v4-flash** - A fast general-purpose model for a wide range of tasks (default)
+- **deepseek-v4-pro** - A more capable model for reasoning and coding
+
+> The legacy model names `deepseek-chat` and `deepseek-reasoner` stop working on 2026-07-24. If your instance still uses one of them, switch the model in the integration options.
 
 #### Google Gemini Models
-- **Gemini 3.1 Pro** - The newest and most advanced model available
-- **Gemini 3.1 Flash Lite** - Fastest and most cost-efficient model for high-volume workloads
+- **gemini-3.5-flash** - Fast and cost-efficient, suitable for most tasks (default)
+- **Gemini 3.1 Pro** - The most advanced Gemini model available
+
+> Google shut down `gemini-2.0-flash` on 2026-06-01 and retires the 2.5 family on 2026-10-16. If your instance uses one of those, switch the model in the integration options.
 
 <details>
 <summary>🌐 Potentially Compatible Providers</summary>
@@ -199,8 +203,8 @@ If the integration is not found in the default repository:
 5. Click "Download"
 
 ### Manual Installation
-1. Download the latest release
-2. Extract and copy `custom_components/ha_text_ai` to your `custom_components` directory
+1. Download `ha_text_ai.zip` from the latest release
+2. Extract the archive and copy the `ha_text_ai` folder into your `custom_components` directory
 3. Restart Home Assistant
 4. Add configuration via UI (Settings → Devices & Services → Add Integration)
 
diff --git a/custom_components/ha_text_ai/__init__.py b/custom_components/ha_text_ai/__init__.py
index 5782a57..1d61eb0 100644
--- a/custom_components/ha_text_ai/__init__.py
+++ b/custom_components/ha_text_ai/__init__.py
@@ -80,8 +80,11 @@ SERVICE_SCHEMA_SET_SYSTEM_PROMPT = vol.Schema({
 
 SERVICE_SCHEMA_GET_HISTORY = vol.Schema({
     vol.Required("instance"): cv.string,
-    # Cap limit to protect automations from huge response payloads.
-    vol.Optional("limit", default=10): vol.All(cv.positive_int, vol.Range(min=1, max=100)),
+    # No default and no schema max: omitting limit returns the full history
+    # (pre-2.5.0 behavior) and oversized values are clamped to
+    # ABSOLUTE_MAX_HISTORY_SIZE in history.async_get_history instead of
+    # failing the whole service call.
+    vol.Optional("limit"): vol.All(cv.positive_int, vol.Range(min=1)),
     vol.Optional("filter_model"): cv.string,
     vol.Optional("start_date"): cv.string,
     vol.Optional("include_metadata"): cv.boolean,
@@ -244,7 +247,9 @@ async def async_check_api(session, endpoint: str, headers: dict, provider: str,
         check_url = f"{endpoint}{check_path}"
 
         async with asyncio.timeout(api_timeout):
-            async with session.get(check_url, headers=headers) as response:
+            async with session.get(
+                check_url, headers=headers, allow_redirects=False
+            ) as response:
                 if response.status == 200:
                     return True
                 elif response.status == 401:
@@ -264,6 +269,7 @@ async def async_setup_entry(hass: HomeAssistant, entry: ConfigEntry) -> bool:
     """Set up HA Text AI from a config entry."""
     _LOGGER.debug("Setting up HA Text AI entry: %s", safe_log_data(dict(entry.data)))
 
+    session = None
     try:
         # Get provider from data or options (options takes precedence)
         config = {**entry.data, **entry.options}
@@ -292,7 +298,9 @@ async def async_setup_entry(hass: HomeAssistant, entry: ConfigEntry) -> bool:
 
         # Pinned session closes DNS-rebinding TOCTOU and isolates cookies
         # from other integrations sharing the same endpoint hostname.
-        session = create_pinned_session(hass, endpoint, resolved_ips)
+        # The integration owns this session: APIClient.shutdown() closes it
+        # on unload, the except handler below closes it on failed setup.
+        session = create_pinned_session(endpoint, resolved_ips)
         # API key can now be updated via options
         api_key = config.get(CONF_API_KEY, entry.data.get(CONF_API_KEY))
         instance_name = entry.data.get(CONF_NAME, entry.entry_id)
@@ -359,6 +367,8 @@ async def async_setup_entry(hass: HomeAssistant, entry: ConfigEntry) -> bool:
 
     except Exception as err:
         _LOGGER.exception("Error setting up HA Text AI: %s", err)
+        if session is not None and not session.closed:
+            await session.close()
         raise
 
 async def async_update_options(hass: HomeAssistant, entry: ConfigEntry) -> None:
diff --git a/custom_components/ha_text_ai/api_client.py b/custom_components/ha_text_ai/api_client.py
index dbc5737..88c74a8 100644
--- a/custom_components/ha_text_ai/api_client.py
+++ b/custom_components/ha_text_ai/api_client.py
@@ -113,6 +113,9 @@ class APIClient:
                     json=payload,
                     headers=self.headers,
                     timeout=self.timeout,
+                    # The session pins DNS to validated IPs; following a
+                    # redirect would resolve a new host past that pin.
+                    allow_redirects=False,
                 ) as response:
                     _LOGGER.debug("Response status: %s", response.status)
                     if response.status == 200:
@@ -358,12 +361,18 @@ class APIClient:
         separate field alongside content. We skip the no_think append for
         this model and preserve reasoning_content in the response payload
         so it's available for logging/debug.
+
+        DeepSeek V4+ (deepseek-v4-flash/-pro) selects thinking mode via a
+        top-level "thinking" request parameter instead of the model name,
+        so /no_think does not apply there.
         """
         url = f"{self.endpoint}/chat/completions"
-        is_reasoner = "reasoner" in model.lower()
+        m_lower = model.lower()
+        is_reasoner = "reasoner" in m_lower
+        is_v4plus = re.search(r"deepseek-v[4-9]", m_lower) is not None
         final_messages = (
             self._apply_no_think_tag(messages)
-            if (disable_thinking and not is_reasoner)
+            if (disable_thinking and not is_reasoner and not is_v4plus)
             else messages
         )
         payload = {
@@ -373,6 +382,8 @@ class APIClient:
             "max_tokens": max_tokens,
             "stream": False,
         }
+        if disable_thinking and is_v4plus:
+            payload["thinking"] = {"type": "disabled"}
         self._apply_structured_output(payload, structured_output, json_schema)
 
         data = await self._make_request(url, payload)
@@ -627,15 +638,26 @@ class APIClient:
                     config.response_mime_type = "application/json"
                     config.response_schema = parsed_schema
 
-                # Disable thinking for Gemini 2.5+ models. Flash variants accept
-                # thinking_budget=0 (fully off); Pro variants reject 0 and require
-                # at least 128 tokens of thinking. 2.0 and earlier ignore the field.
+                # Disable thinking. Gemini 3.x+ replaced the numeric
+                # thinking_budget with a semantic thinking_level; Pro
+                # variants do not accept MINIMAL, their floor is LOW.
+                # Gemini 2.5: Flash accepts thinking_budget=0 (fully off),
+                # Pro rejects 0 and requires at least 128 tokens.
+                # 2.0 and earlier ignore the field.
                 if disable_thinking:
                     m_lower = model.lower()
-                    budget = 128 if "2.5-pro" in m_lower else 0
                     try:
-                        config.thinking_config = types.ThinkingConfig(thinking_budget=budget)
-                    except (AttributeError, TypeError) as err:
+                        if re.search(r"gemini-[3-9]", m_lower):
+                            level = "LOW" if "pro" in m_lower else "MINIMAL"
+                            config.thinking_config = types.ThinkingConfig(
+                                thinking_level=level
+                            )
+                        else:
+                            budget = 128 if "2.5-pro" in m_lower else 0
+                            config.thinking_config = types.ThinkingConfig(
+                                thinking_budget=budget
+                            )
+                    except (AttributeError, TypeError, ValueError) as err:
                         _LOGGER.debug(
                             "ThinkingConfig not supported by this google-genai version: %s", err
                         )
@@ -735,7 +757,11 @@ class APIClient:
             raise HomeAssistantError(f"Gemini API request failed: {e}") from e
 
     async def shutdown(self) -> None:
-        """Shutdown API client."""
+        """Shutdown API client and close its dedicated session."""
         _LOGGER.debug("Shutting down API client")
         self._closed = True
-        # Do NOT close the shared Home Assistant session
+        # The session is dedicated to this config entry (pinned resolver,
+        # isolated cookie jar), so it must be closed here to release the
+        # connector; nothing else owns it.
+        if self.session is not None and not self.session.closed:
+            await self.session.close()
diff --git a/custom_components/ha_text_ai/config_flow.py b/custom_components/ha_text_ai/config_flow.py
index 7eca6ca..077545a 100644
--- a/custom_components/ha_text_ai/config_flow.py
+++ b/custom_components/ha_text_ai/config_flow.py
@@ -255,23 +255,28 @@ class HATextAIConfigFlow(config_entries.ConfigFlow, domain=DOMAIN):
                     return False
                 return True
 
-            # Pinned session ensures the reachability check goes to the same
-            # IP that will later be used by api_client (no DNS rebinding).
-            session = create_pinned_session(self.hass, endpoint, resolved_ips)
             headers = build_auth_headers(self._provider, user_input[CONF_API_KEY])
 
             from .providers import get_provider_config
             check_path = get_provider_config(self._provider).get("check_path", "/models")
             check_url = f"{endpoint}{check_path}"
 
-            async with session.get(check_url, headers=headers) as response:
-                if response.status == 401:
-                    self._errors["base"] = "invalid_auth"
-                    return False
-                elif response.status != 200:
-                    self._errors["base"] = "cannot_connect"
-                    return False
-                return True
+            # Pinned session ensures the reachability check goes to the same
+            # IP that will later be used by api_client (no DNS rebinding).
+            session = create_pinned_session(endpoint, resolved_ips)
+            try:
+                async with session.get(
+                    check_url, headers=headers, allow_redirects=False
+                ) as response:
+                    if response.status == 401:
+                        self._errors["base"] = "invalid_auth"
+                        return False
+                    elif response.status != 200:
+                        self._errors["base"] = "cannot_connect"
+                        return False
+                    return True
+            finally:
+                await session.close()
 
         except Exception as err:
             _LOGGER.error("API validation error: %s", str(err))
@@ -340,21 +345,26 @@ class OptionsFlowHandler(config_entries.OptionsFlow):
             if provider == API_PROVIDER_GEMINI:
                 return True
 
-            session = create_pinned_session(self.hass, endpoint, resolved_ips)
             headers = build_auth_headers(provider, api_key)
 
             from .providers import get_provider_config
             check_path = get_provider_config(provider).get("check_path", "/models")
             check_url = f"{endpoint}{check_path}"
 
-            async with session.get(check_url, headers=headers) as response:
-                if response.status == 401:
-                    self._errors["base"] = "invalid_auth"
-                    return False
-                elif response.status != 200:
-                    self._errors["base"] = "cannot_connect"
-                    return False
-                return True
+            session = create_pinned_session(endpoint, resolved_ips)
+            try:
+                async with session.get(
+                    check_url, headers=headers, allow_redirects=False
+                ) as response:
+                    if response.status == 401:
+                        self._errors["base"] = "invalid_auth"
+                        return False
+                    elif response.status != 200:
+                        self._errors["base"] = "cannot_connect"
+                        return False
+                    return True
+            finally:
+                await session.close()
 
         except Exception as err:
             _LOGGER.error("API validation error: %s", str(err))
diff --git a/custom_components/ha_text_ai/const.py b/custom_components/ha_text_ai/const.py
index f3ab014..7312f73 100644
--- a/custom_components/ha_text_ai/const.py
+++ b/custom_components/ha_text_ai/const.py
@@ -58,8 +58,11 @@ MAX_HISTORY_FILE_SIZE = 1 * 1024 * 1024
 # Default values
 DEFAULT_MODEL: Final = "gpt-4o-mini"
 DEFAULT_ANTHROPIC_MODEL: Final = "claude-sonnet-4-6"
-DEFAULT_DEEPSEEK_MODEL: Final = "deepseek-chat"
-DEFAULT_GEMINI_MODEL: Final = "gemini-2.0-flash"
+# deepseek-chat/deepseek-reasoner are discontinued 2026-07-24; V4 models
+# select thinking mode via a request parameter instead of the model name.
+DEFAULT_DEEPSEEK_MODEL: Final = "deepseek-v4-flash"
+# gemini-2.0-flash was shut down 2026-06-01; 2.5-flash follows 2026-10-16.
+DEFAULT_GEMINI_MODEL: Final = "gemini-3.5-flash"
 DEFAULT_TEMPERATURE: Final = 0.1
 DEFAULT_MAX_TOKENS: Final = 1000
 DEFAULT_REQUEST_INTERVAL: Final = 1.0
diff --git a/custom_components/ha_text_ai/translations/de.json b/custom_components/ha_text_ai/translations/de.json
index a63e273..039cc72 100644
--- a/custom_components/ha_text_ai/translations/de.json
+++ b/custom_components/ha_text_ai/translations/de.json
@@ -46,6 +46,7 @@
       "name_exists": "Eine Instanz mit diesem Namen existiert bereits",
       "invalid_name": "Ungültiger Instanzname",
       "invalid_auth": "Authentifizierung fehlgeschlagen - überprüfen Sie Ihren API-Schlüssel",
+      "api_key_required": "API-Schlüssel ist erforderlich, wenn Anbieter oder Endpunkt geändert wird",
       "invalid_api_key": "Ungültiger API-Schlüssel - bitte überprüfen Sie Ihre Anmeldeinformationen",
       "cannot_connect": "Verbindung zum API-Dienst fehlgeschlagen",
       "invalid_model": "Ausgewähltes Modell ist nicht verfügbar",
@@ -59,7 +60,6 @@
       "invalid_instance": "Ungültige Instanz angegeben",
       "unknown": "Unerwarteter Fehler aufgetreten",
       "empty": "Name darf nicht leer sein",
-      "invalid_characters": "Name darf nur Buchstaben, Zahlen, Leerzeichen, Unterstriche und Bindestriche enthalten",
       "name_too_long": "Name darf höchstens 50 Zeichen lang sein"
     },
     "abort": {
@@ -218,8 +218,7 @@
           "rate_limited": "Rate limitiert",
           "maintenance": "Wartung",
           "initializing": "Initialisierung",
-          "retrying": "Wiederholen",
-          "queued": "In der Warteschlange"
+          "retrying": "Wiederholen"
         },
         "state_attributes": {
           "question": {
@@ -311,6 +310,21 @@
           },
           "min_latency": {
             "name": "Minimale Latenz"
+          },
+          "last_model": {
+            "name": "Zuletzt verwendetes Modell"
+          },
+          "last_timestamp": {
+            "name": "Zeitpunkt der letzten Antwort"
+          },
+          "instance_name": {
+            "name": "Instanzname"
+          },
+          "normalized_name": {
+            "name": "Normalisierter Name"
+          },
+          "conversation_history": {
+            "name": "Konversationsverlauf"
           }
         }
       }
diff --git a/custom_components/ha_text_ai/translations/es.json b/custom_components/ha_text_ai/translations/es.json
index f813987..d905d96 100644
--- a/custom_components/ha_text_ai/translations/es.json
+++ b/custom_components/ha_text_ai/translations/es.json
@@ -46,6 +46,7 @@
       "name_exists": "Ya existe una instancia con este nombre",
       "invalid_name": "Nombre de instancia no válido",
       "invalid_auth": "La autenticación falló - verifica tu clave API",
+      "api_key_required": "Se requiere la clave API al cambiar de proveedor o endpoint",
       "invalid_api_key": "Clave API no válida - verifica tus credenciales",
       "cannot_connect": "Error al conectar con el servicio de API",
       "invalid_model": "El modelo seleccionado no está disponible",
@@ -59,7 +60,6 @@
       "invalid_instance": "Instancia no válida especificada",
       "unknown": "Ocurrió un error inesperado",
       "empty": "El nombre no puede estar vacío",
-      "invalid_characters": "El nombre solo puede contener letras, números, espacios, guiones bajos y guiones",
       "name_too_long": "El nombre debe tener 50 caracteres o menos"
     },
     "abort": {
@@ -218,8 +218,7 @@
           "rate_limited": "Limitado por tasa",
           "maintenance": "Mantenimiento",
           "initializing": "Inicializando",
-          "retrying": "Reintentando",
-          "queued": "En cola"
+          "retrying": "Reintentando"
         },
         "state_attributes": {
           "question": {
@@ -311,6 +310,21 @@
           },
           "min_latency": {
             "name": "Latencia Mínima"
+          },
+          "last_model": {
+            "name": "Último modelo utilizado"
+          },
+          "last_timestamp": {
+            "name": "Hora de la última respuesta"
+          },
+          "instance_name": {
+            "name": "Nombre de instancia"
+          },
+          "normalized_name": {
+            "name": "Nombre normalizado"
+          },
+          "conversation_history": {
+            "name": "Historial de conversación"
           }
         }
       }
diff --git a/custom_components/ha_text_ai/translations/hi.json b/custom_components/ha_text_ai/translations/hi.json
index ecefc56..393d996 100644
--- a/custom_components/ha_text_ai/translations/hi.json
+++ b/custom_components/ha_text_ai/translations/hi.json
@@ -46,6 +46,7 @@
       "name_exists": "इस नाम के साथ एक उदाहरण पहले से मौजूद है",
       "invalid_name": "अमान्य उदाहरण नाम",
       "invalid_auth": "प्रमाणीकरण विफल - अपनी एपीआई कुंजी की जांच करें",
+      "api_key_required": "प्रदाता या endpoint बदलते समय API कुंजी आवश्यक है",
       "invalid_api_key": "अमान्य एपीआई कुंजी - कृपया अपनी क्रेडेंशियल्स की पुष्टि करें",
       "cannot_connect": "एपीआई सेवा से कनेक्ट करने में विफल",
       "invalid_model": "चुना हुआ मॉडल उपलब्ध नहीं है",
@@ -59,7 +60,6 @@
       "invalid_instance": "अमान्य उदाहरण निर्दिष्ट किया गया",
       "unknown": "अप्रत्याशित त्रुटि हुई",
       "empty": "नाम खाली नहीं हो सकता",
-      "invalid_characters": "नाम में केवल अक्षर, अंक, रिक्त स्थान, अंडरस्कोर और हाइफन हो सकते हैं",
       "name_too_long": "नाम 50 अक्षरों या उससे कम होना चाहिए"
     },
     "abort": {
@@ -218,8 +218,7 @@
           "rate_limited": "रेट सीमित",
           "maintenance": "रखरखाव",
           "initializing": "प्रारंभिककरण",
-          "retrying": "पुनः प्रयास कर रहा है",
-          "queued": "क्यू में"
+          "retrying": "पुनः प्रयास कर रहा है"
         },
         "state_attributes": {
           "question": {
@@ -311,9 +310,24 @@
           },
           "min_latency": {
             "name": "न्यूनतम विलंबता"
+          },
+          "last_model": {
+            "name": "अंतिम उपयोग किया गया मॉडल"
+          },
+          "last_timestamp": {
+            "name": "अंतिम प्रतिक्रिया समय"
+          },
+          "instance_name": {
+            "name": "इंस्टेंस नाम"
+          },
+          "normalized_name": {
+            "name": "सामान्यीकृत नाम"
+          },
+          "conversation_history": {
+            "name": "वार्तालाप इतिहास"
           }
         }
       }
     }
   }
-}
\ No newline at end of file
+}
diff --git a/custom_components/ha_text_ai/translations/it.json b/custom_components/ha_text_ai/translations/it.json
index 7791c07..732cfe8 100644
--- a/custom_components/ha_text_ai/translations/it.json
+++ b/custom_components/ha_text_ai/translations/it.json
@@ -46,6 +46,7 @@
       "name_exists": "Esiste già un'istanza con questo nome",
       "invalid_name": "Nome dell'istanza non valido",
       "invalid_auth": "Autenticazione fallita - controlla la tua chiave API",
+      "api_key_required": "La chiave API è obbligatoria quando si cambia provider o endpoint",
       "invalid_api_key": "Chiave API non valida - verifica le tue credenziali",
       "cannot_connect": "Impossibile connettersi al servizio API",
       "invalid_model": "Il modello selezionato non è disponibile",
@@ -59,7 +60,6 @@
       "invalid_instance": "Istanze specificata non valida",
       "unknown": "Si è verificato un errore imprevisto",
       "empty": "Il nome non può essere vuoto",
-      "invalid_characters": "Il nome può contenere solo lettere, numeri, spazi, trattini bassi e trattini",
       "name_too_long": "Il nome deve essere lungo 50 caratteri o meno"
     },
     "abort": {
@@ -218,8 +218,7 @@
           "rate_limited": "Limite di frequenza",
           "maintenance": "Manutenzione",
           "initializing": "Inizializzazione",
-          "retrying": "Riprova",
-          "queued": "In coda"
+          "retrying": "Riprova"
         },
         "state_attributes": {
           "question": {
@@ -311,6 +310,21 @@
           },
           "min_latency": {
             "name": "Latenza minima"
+          },
+          "last_model": {
+            "name": "Ultimo modello utilizzato"
+          },
+          "last_timestamp": {
+            "name": "Ora dell'ultima risposta"
+          },
+          "instance_name": {
+            "name": "Nome istanza"
+          },
+          "normalized_name": {
+            "name": "Nome normalizzato"
+          },
+          "conversation_history": {
+            "name": "Cronologia conversazione"
           }
         }
       }
diff --git a/custom_components/ha_text_ai/translations/sr.json b/custom_components/ha_text_ai/translations/sr.json
index 2a6e4bd..19b641f 100644
--- a/custom_components/ha_text_ai/translations/sr.json
+++ b/custom_components/ha_text_ai/translations/sr.json
@@ -46,6 +46,7 @@
       "name_exists": "Инстанца са овим именом већ постоји",
       "invalid_name": "Неважеће име инстанце",
       "invalid_auth": "Аутентификација није успела - проверите ваш API кључ",
+      "api_key_required": "API кључ је обавезан при промени провајдера или endpoint-а",
       "invalid_api_key": "Неважећи API кључ - молимо проверите ваше акредитиве",
       "cannot_connect": "Неуспело повезивање са API сервисом",
       "invalid_model": "Изабрани модел није доступан",
@@ -59,7 +60,6 @@
       "invalid_instance": "Неважећа инстанца је назначена",
       "unknown": "Дошло је до неочекиване грешке",
       "empty": "Име не може бити празно",
-      "invalid_characters": "Име може садржавати само слова, цифре, размаке, подцрта и цртице",
       "name_too_long": "Име мора бити 50 знакова или мање"
     },
     "abort": {
@@ -218,8 +218,7 @@
           "rate_limited": "Ограничење захтева",
           "maintenance": "Одржавање",
           "initializing": "Инициализује се",
-          "retrying": "Покушава поново",
-          "queued": "У реду"
+          "retrying": "Покушава поново"
         },
         "state_attributes": {
           "question": {
@@ -311,9 +310,24 @@
           },
           "min_latency": {
             "name": "Минимална латенција"
+          },
+          "last_model": {
+            "name": "Последњи коришћени модел"
+          },
+          "last_timestamp": {
+            "name": "Време последњег одговора"
+          },
+          "instance_name": {
+            "name": "Назив инстанце"
+          },
+          "normalized_name": {
+            "name": "Нормализовани назив"
+          },
+          "conversation_history": {
+            "name": "Историја разговора"
           }
         }
       }
     }
   }
-}
\ No newline at end of file
+}
diff --git a/custom_components/ha_text_ai/translations/zh.json b/custom_components/ha_text_ai/translations/zh.json
index 78e2566..df2709c 100644
--- a/custom_components/ha_text_ai/translations/zh.json
+++ b/custom_components/ha_text_ai/translations/zh.json
@@ -46,6 +46,7 @@
       "name_exists": "具有此名称的实例已存在",
       "invalid_name": "无效的实例名称",
       "invalid_auth": "身份验证失败 - 检查您的API密钥",
+      "api_key_required": "更改提供商或端点时需要输入 API 密钥",
       "invalid_api_key": "无效的API密钥 - 请验证您的凭据",
       "cannot_connect": "无法连接到API服务",
       "invalid_model": "所选模型不可用",
@@ -59,7 +60,6 @@
       "invalid_instance": "指定的实例无效",
       "unknown": "发生意外错误",
       "empty": "名称不能为空",
-      "invalid_characters": "名称只能包含字母、数字、空格、下划线和连字符",
       "name_too_long": "名称必须少于50个字符"
     },
     "abort": {
@@ -218,8 +218,7 @@
           "rate_limited": "速率限制",
           "maintenance": "维护中",
           "initializing": "初始化中",
-          "retrying": "重试中",
-          "queued": "排队中"
+          "retrying": "重试中"
         },
         "state_attributes": {
           "question": {
@@ -311,9 +310,24 @@
           },
           "min_latency": {
             "name": "最小延迟"
+          },
+          "last_model": {
+            "name": "最近使用的模型"
+          },
+          "last_timestamp": {
+            "name": "最近响应时间"
+          },
+          "instance_name": {
+            "name": "实例名称"
+          },
+          "normalized_name": {
+            "name": "规范化名称"
+          },
+          "conversation_history": {
+            "name": "对话历史"
           }
         }
       }
     }
   }
-}
\ No newline at end of file
+}
diff --git a/custom_components/ha_text_ai/utils.py b/custom_components/ha_text_ai/utils.py
index 1965aec..887e4e2 100644
--- a/custom_components/ha_text_ai/utils.py
+++ b/custom_components/ha_text_ai/utils.py
@@ -21,7 +21,6 @@ from aiohttp.resolver import DefaultResolver
 
 from homeassistant.const import CONF_API_KEY
 from homeassistant.core import HomeAssistant
-from homeassistant.helpers.aiohttp_client import async_create_clientsession
 
 _LOGGER = logging.getLogger(__name__)
 
@@ -150,7 +149,6 @@ async def resolve_hostname_ips(
     return ips
 
 def create_pinned_session(
-    hass: HomeAssistant,
     endpoint: str,
     resolved_ips: list[str],
 ) -> aiohttp.ClientSession:
@@ -161,6 +159,13 @@ def create_pinned_session(
       rather than re-resolving the hostname on each request.
     - Cookie pollution: DummyCookieJar prevents cookies from leaking between
       this integration and other HA components sharing the same domain.
+
+    Built directly on aiohttp: HA's async_create_clientsession always
+    injects its own pooled connector and rejects a caller-supplied one,
+    so a custom resolver cannot go through the helper. The caller owns
+    the session and must close it. Requests must pass
+    allow_redirects=False so a 3xx response cannot route past the pinned
+    resolver to an unvalidated host.
     """
     parsed = urlparse(endpoint)
     hostname = (parsed.hostname or "").lower()
@@ -169,10 +174,9 @@ def create_pinned_session(
         hostname: [(ip, port) for ip in resolved_ips]
     }
     connector = aiohttp.TCPConnector(resolver=_PinnedResolver(pinned))
-    return async_create_clientsession(
-        hass,
-        cookie_jar=aiohttp.DummyCookieJar(),
+    return aiohttp.ClientSession(
         connector=connector,
+        cookie_jar=aiohttp.DummyCookieJar(),
     )
 
 async def validate_endpoint(