Merge branch 'dev' into dev

backend/open_webui/env.py

@@ -473,6 +473,7 @@ WEBUI_ADMIN_NAME = os.environ.get('WEBUI_ADMIN_NAME', 'Admin')
 WEBUI_AUTH_TRUSTED_EMAIL_HEADER = os.environ.get('WEBUI_AUTH_TRUSTED_EMAIL_HEADER', None)
 WEBUI_AUTH_TRUSTED_NAME_HEADER = os.environ.get('WEBUI_AUTH_TRUSTED_NAME_HEADER', None)
 WEBUI_AUTH_TRUSTED_GROUPS_HEADER = os.environ.get('WEBUI_AUTH_TRUSTED_GROUPS_HEADER', None)
+WEBUI_AUTH_TRUSTED_ROLE_HEADER = os.environ.get('WEBUI_AUTH_TRUSTED_ROLE_HEADER', None)
 
 
 ENABLE_PASSWORD_VALIDATION = os.environ.get('ENABLE_PASSWORD_VALIDATION', 'False').lower() == 'true'
@@ -629,6 +630,13 @@ else:
         CHAT_RESPONSE_MAX_TOOL_CALL_RETRIES = 30
 
 
+# WARNING: Experimental. Only enable if your upstream Responses API endpoint
+# supports stateful sessions (i.e. server-side response storage with
+# previous_response_id anchoring). Most proxies and third-party endpoints
+# are stateless and will break if this is enabled.
+ENABLE_RESPONSES_API_STATEFUL = os.environ.get('ENABLE_RESPONSES_API_STATEFUL', 'False').lower() == 'true'
+
+
 CHAT_STREAM_RESPONSE_CHUNK_MAX_BUFFER_SIZE = os.environ.get('CHAT_STREAM_RESPONSE_CHUNK_MAX_BUFFER_SIZE', '')
 
 if CHAT_STREAM_RESPONSE_CHUNK_MAX_BUFFER_SIZE == '':

backend/open_webui/main.py

@@ -403,6 +403,7 @@ from open_webui.config import (
     EVALUATION_ARENA_MODELS,
     # WebUI (OAuth)
     ENABLE_OAUTH_ROLE_MANAGEMENT,
+    OAUTH_SUB_CLAIM,
     OAUTH_ROLES_CLAIM,
     OAUTH_EMAIL_CLAIM,
     OAUTH_PICTURE_CLAIM,
@@ -888,6 +889,7 @@ if any('access_control' in m.get('meta', {}) for m in arena_models):
         migrate_access_control(model.get('meta', {}))
     app.state.config.EVALUATION_ARENA_MODELS = arena_models
 
+app.state.config.OAUTH_SUB_CLAIM = OAUTH_SUB_CLAIM
 app.state.config.OAUTH_USERNAME_CLAIM = OAUTH_USERNAME_CLAIM
 app.state.config.OAUTH_PICTURE_CLAIM = OAUTH_PICTURE_CLAIM
 app.state.config.OAUTH_EMAIL_CLAIM = OAUTH_EMAIL_CLAIM

backend/open_webui/models/access_grants.py

@@ -192,6 +192,16 @@ def has_public_read_access_grant(access_grants: Optional[list]) -> bool:
     return False
 
 
+def has_public_write_access_grant(access_grants: Optional[list]) -> bool:
+    """
+    Returns True when a direct grant list includes wildcard public-write.
+    """
+    for grant in normalize_access_grants(access_grants):
+        if grant['principal_type'] == 'user' and grant['principal_id'] == '*' and grant['permission'] == 'write':
+            return True
+    return False
+
+
 def has_user_access_grant(access_grants: Optional[list]) -> bool:
     """
     Returns True when a direct grant list includes any non-wildcard user grant.

backend/open_webui/models/functions.py

@@ -2,9 +2,9 @@ import logging
 import time
 from typing import Optional
 
-from sqlalchemy.orm import Session
+from sqlalchemy.orm import Session, defer
 from open_webui.internal.db import Base, JSONField, get_db, get_db_context
-from open_webui.models.users import Users, UserModel
+from open_webui.models.users import Users, UserModel, UserResponse
 from pydantic import BaseModel, ConfigDict
 from sqlalchemy import BigInteger, Boolean, Column, String, Text, Index
 
@@ -75,10 +75,6 @@ class FunctionWithValvesModel(BaseModel):
 ####################
 
 
-class FunctionUserResponse(FunctionModel):
-    user: Optional[UserModel] = None
-
-
 class FunctionResponse(BaseModel):
     id: str
     user_id: str
@@ -90,6 +86,12 @@ class FunctionResponse(BaseModel):
     updated_at: int  # timestamp in epoch
     created_at: int  # timestamp in epoch
 
+    model_config = ConfigDict(from_attributes=True)
+
+
+class FunctionUserResponse(FunctionResponse):
+    user: Optional[UserResponse] = None
+
 
 class FunctionForm(BaseModel):
     id: str
@@ -224,7 +226,12 @@ class FunctionsTable:
 
     def get_function_list(self, db: Optional[Session] = None) -> list[FunctionUserResponse]:
         with get_db_context(db) as db:
-            functions = db.query(Function).order_by(Function.updated_at.desc()).all()
+            functions = (
+                db.query(Function)
+                .options(defer(Function.content))
+                .order_by(Function.updated_at.desc())
+                .all()
+            )
             user_ids = list(set(func.user_id for func in functions))
 
             users = Users.get_users_by_user_ids(user_ids, db=db) if user_ids else []
@@ -233,8 +240,17 @@ class FunctionsTable:
             return [
                 FunctionUserResponse.model_validate(
                     {
-                        **FunctionModel.model_validate(func).model_dump(),
-                        'user': (users_dict.get(func.user_id).model_dump() if func.user_id in users_dict else None),
+                        **FunctionResponse.model_validate(func).model_dump(),
+                        'user': (
+                            UserResponse(
+                                id=users_dict[func.user_id].id,
+                                name=users_dict[func.user_id].name,
+                                role=users_dict[func.user_id].role,
+                                email=users_dict[func.user_id].email,
+                            ).model_dump()
+                            if func.user_id in users_dict
+                            else None
+                        ),
                     }
                 )
                 for func in functions

backend/open_webui/retrieval/utils.py

@@ -404,11 +404,34 @@ def get_all_items_from_collections(collection_names: list[str]) -> dict:
 
 
 async def query_collection(
+    request,
     collection_names: list[str],
     queries: list[str],
     embedding_function,
     k: int,
 ) -> dict:
+    # When request is provided, try hybrid search + reranking if enabled
+    if request and request.app.state.config.ENABLE_RAG_HYBRID_SEARCH:
+        try:
+            reranking_function = (
+                (lambda query, documents: request.app.state.RERANKING_FUNCTION(query, documents))
+                if request.app.state.RERANKING_FUNCTION
+                else None
+            )
+            return await query_collection_with_hybrid_search(
+                collection_names=collection_names,
+                queries=queries,
+                embedding_function=embedding_function,
+                k=k,
+                reranking_function=reranking_function,
+                k_reranker=request.app.state.config.TOP_K_RERANKER,
+                r=request.app.state.config.RELEVANCE_THRESHOLD,
+                hybrid_bm25_weight=request.app.state.config.HYBRID_BM25_WEIGHT,
+                enable_enriched_texts=request.app.state.config.ENABLE_RAG_HYBRID_SEARCH_ENRICHED_TEXTS,
+            )
+        except Exception as e:
+            log.debug(f'Hybrid search failed, falling back to vector search: {e}')
+
     results = []
     error = False
 
@@ -1126,31 +1149,13 @@ async def get_sources_from_items(
                 if full_context:
                     query_result = get_all_items_from_collections(collection_names)
                 else:
-                    query_result = None  # Initialize to None
-                    if hybrid_search:
-                        try:
-                            query_result = await query_collection_with_hybrid_search(
-                                collection_names=collection_names,
-                                queries=queries,
-                                embedding_function=embedding_function,
-                                k=k,
-                                reranking_function=reranking_function,
-                                k_reranker=k_reranker,
-                                r=r,
-                                hybrid_bm25_weight=hybrid_bm25_weight,
-                                enable_enriched_texts=request.app.state.config.ENABLE_RAG_HYBRID_SEARCH_ENRICHED_TEXTS,
-                            )
-                        except Exception as e:
-                            log.debug('Error when using hybrid search, using non hybrid search as fallback.')
-
-                    # fallback to non-hybrid search
-                    if not hybrid_search and query_result is None:
-                        query_result = await query_collection(
-                            collection_names=collection_names,
-                            queries=queries,
-                            embedding_function=embedding_function,
-                            k=k,
-                        )
+                    query_result = await query_collection(
+                        request,
+                        collection_names=collection_names,
+                        queries=queries,
+                        embedding_function=embedding_function,
+                        k=k,
+                    )
             except Exception as e:
                 log.exception(e)
 

backend/open_webui/retrieval/web/serper.py

@@ -31,7 +31,7 @@ def search_serper(api_key: str, query: str, count: int, filter_list: Optional[li
         SearchResult(
             link=result['link'],
             title=result.get('title'),
-            snippet=result.get('description'),
+            snippet=result.get('snippet'),
         )
         for result in results[:count]
     ]

backend/open_webui/routers/auths.py

@@ -35,6 +35,7 @@ from open_webui.env import (
     WEBUI_AUTH_TRUSTED_EMAIL_HEADER,
     WEBUI_AUTH_TRUSTED_NAME_HEADER,
     WEBUI_AUTH_TRUSTED_GROUPS_HEADER,
+    WEBUI_AUTH_TRUSTED_ROLE_HEADER,
     WEBUI_AUTH_COOKIE_SAME_SITE,
     WEBUI_AUTH_COOKIE_SECURE,
     WEBUI_AUTH_SIGNOUT_REDIRECT_URL,
@@ -117,6 +118,7 @@ def create_session_response(request: Request, user, db, response: Response = Non
 
     if set_cookie and response:
         datetime_expires_at = datetime.datetime.fromtimestamp(expires_at, datetime.timezone.utc) if expires_at else None
+        max_age = int(expires_delta.total_seconds()) if expires_delta else None
         response.set_cookie(
             key='token',
             value=token,
@@ -124,6 +126,7 @@ def create_session_response(request: Request, user, db, response: Response = Non
             httponly=True,
             samesite=WEBUI_AUTH_COOKIE_SAME_SITE,
             secure=WEBUI_AUTH_COOKIE_SECURE,
+            **({'max_age': max_age} if max_age is not None else {}),
         )
 
     user_permissions = get_permissions(user.id, request.app.state.config.USER_PERMISSIONS, db=db)
@@ -181,6 +184,7 @@ async def get_session_user(
             )
 
         # Set the cookie token
+        max_age = int(expires_at - time.time()) if expires_at else None
         response.set_cookie(
             key='token',
             value=token,
@@ -188,6 +192,7 @@ async def get_session_user(
             httponly=True,  # Ensures the cookie is not accessible via JavaScript
             samesite=WEBUI_AUTH_COOKIE_SAME_SITE,
             secure=WEBUI_AUTH_COOKIE_SECURE,
+            **({'max_age': max_age} if max_age is not None else {}),
         )
 
     user_permissions = get_permissions(user.id, request.app.state.config.USER_PERMISSIONS, db=db)
@@ -500,7 +505,7 @@ async def ldap_auth(
             user = Auths.authenticate_user_by_email(email, db=db)
 
             if user:
-                if user.role != 'admin' and ENABLE_LDAP_GROUP_MANAGEMENT and user_groups:
+                if ENABLE_LDAP_GROUP_MANAGEMENT and user_groups:
                     if ENABLE_LDAP_GROUP_CREATION:
                         Groups.create_groups_by_group_names(user.id, user_groups, db=db)
                     try:
@@ -561,12 +566,23 @@ async def signin(
             )
 
         user = Auths.authenticate_user_by_email(email, db=db)
-        if WEBUI_AUTH_TRUSTED_GROUPS_HEADER and user and user.role != 'admin':
-            group_names = request.headers.get(WEBUI_AUTH_TRUSTED_GROUPS_HEADER, '').split(',')
-            group_names = [name.strip() for name in group_names if name.strip()]
+        if user:
+            if WEBUI_AUTH_TRUSTED_GROUPS_HEADER:
+                group_names = request.headers.get(WEBUI_AUTH_TRUSTED_GROUPS_HEADER, '').split(',')
+                group_names = [name.strip() for name in group_names if name.strip()]
 
-            if group_names:
-                Groups.sync_groups_by_group_names(user.id, group_names, db=db)
+                if group_names:
+                    Groups.sync_groups_by_group_names(user.id, group_names, db=db)
+
+            if WEBUI_AUTH_TRUSTED_ROLE_HEADER:
+                trusted_role = request.headers.get(WEBUI_AUTH_TRUSTED_ROLE_HEADER, '').lower().strip()
+                if trusted_role in {'admin', 'user', 'pending'}:
+                    if user.role != trusted_role:
+                        Users.update_user_role_by_id(user.id, trusted_role, db=db)
+                elif trusted_role:
+                    log.warning(
+                        f'Ignoring invalid trusted role header value: {trusted_role}'
+                    )
 
     elif WEBUI_AUTH == False:
         admin_email = 'admin@localhost'

backend/open_webui/routers/channels.py

@@ -36,7 +36,7 @@ from open_webui.models.channels import (
     ChannelWebhookModel,
     ChannelWebhookForm,
 )
-from open_webui.models.access_grants import AccessGrants, has_public_read_access_grant
+from open_webui.models.access_grants import AccessGrants, has_public_read_access_grant, has_public_write_access_grant
 from open_webui.models.messages import (
     Messages,
     MessageModel,
@@ -88,7 +88,7 @@ def channel_has_access(
     ):
         return True
 
-    if not strict and permission == 'write' and has_public_read_access_grant(channel.access_grants):
+    if not strict and permission == 'write' and has_public_write_access_grant(channel.access_grants):
         return True
 
     return False

backend/open_webui/routers/files.py

@@ -722,7 +722,7 @@ async def get_file_content_by_id(id: str, user=Depends(get_verified_user), db: S
                 )
         else:
             # File path doesn’t exist, return the content as .txt if possible
-            file_content = file.content.get('content', '')
+            file_content = file.data.get('content', '')
             file_name = file.filename
 
             # Create a generator that encodes the file content

backend/open_webui/routers/notes.py

@@ -30,6 +30,7 @@ from open_webui.utils.auth import get_admin_user, get_verified_user
 from open_webui.utils.access_control import (
     has_permission,
     has_public_read_access_grant,
+    has_public_write_access_grant,
     filter_allowed_access_grants,
 )
 from open_webui.models.access_grants import AccessGrants
@@ -234,7 +235,7 @@ async def get_note_by_id(
             permission='write',
             db=db,
         )
-        or has_public_read_access_grant(note.access_grants)
+        or has_public_write_access_grant(note.access_grants)
     )
 
     return NoteResponse(**note.model_dump(), write_access=write_access)

backend/open_webui/routers/openai.py

@@ -2,6 +2,7 @@ import asyncio
 import hashlib
 import json
 import logging
+import re
 from typing import Optional
 from urllib.parse import urlparse
 
@@ -739,8 +740,16 @@ def get_azure_allowed_params(api_version: str) -> set[str]:
     return allowed_params
 
 
-def is_openai_reasoning_model(model: str) -> bool:
-    return model.lower().startswith(('o1', 'o3', 'o4', 'gpt-5'))
+def is_openai_new_model(model: str) -> bool:
+    model_lower = model.lower()
+    # o-series models (o1, o3, o4, o5, ...)
+    if re.match(r'^o\d+', model_lower):
+        return True
+    # gpt-N where N >= 5 (gpt-5, gpt-5.2, gpt-6, ...)
+    m = re.match(r'^gpt-(\d+)', model_lower)
+    if m and int(m.group(1)) >= 5:
+        return True
+    return False
 
 
 def convert_to_azure_payload(url, payload: dict, api_version: str):
@@ -750,7 +759,7 @@ def convert_to_azure_payload(url, payload: dict, api_version: str):
     allowed_params = get_azure_allowed_params(api_version)
 
     # Special handling for o-series models
-    if is_openai_reasoning_model(model):
+    if is_openai_new_model(model):
         # Convert max_tokens to max_completion_tokens for o-series models
         if 'max_tokens' in payload:
             payload['max_completion_tokens'] = payload['max_tokens']
@@ -770,6 +779,31 @@ def convert_to_azure_payload(url, payload: dict, api_version: str):
     return url, payload
 
 
+# Fields accepted by the Responses API for each input item type.
+RESPONSES_ALLOWED_FIELDS: dict[str, set[str]] = {
+    'message': {'type', 'role', 'content'},
+    'function_call': {'type', 'call_id', 'name', 'arguments', 'id'},
+    'function_call_output': {'type', 'call_id', 'output'},
+}
+
+
+def _normalize_stored_item(item: dict) -> dict:
+    """Strip local-only fields from a stored output item before replaying it.
+
+    Open WebUI stores extra bookkeeping fields (``id``, ``status``,
+    ``started_at``, ``ended_at``, ``duration``, ``_tag_type``,
+    ``attributes``, ``summary``, etc.) that the Responses API does
+    not accept.  This helper returns a copy containing only the
+    fields the API understands.
+    """
+    item_type = item.get('type', '')
+    allowed = RESPONSES_ALLOWED_FIELDS.get(item_type)
+    if allowed is None:
+        # Unknown type — pass through as-is (e.g. reasoning, extension items).
+        return item
+    return {k: v for k, v in item.items() if k in allowed}
+
+
 def convert_to_responses_payload(payload: dict) -> dict:
     """
     Convert Chat Completions payload to Responses API format.
@@ -789,7 +823,7 @@ def convert_to_responses_payload(payload: dict) -> dict:
         # Check for stored output items (from previous Responses API turn)
         stored_output = msg.get('output')
         if stored_output and isinstance(stored_output, list):
-            input_items.extend(stored_output)
+            input_items.extend(_normalize_stored_item(item) for item in stored_output)
             continue
 
         if role == 'system':
@@ -799,6 +833,38 @@ def convert_to_responses_payload(payload: dict) -> dict:
                 system_content = '\n'.join(p.get('text', '') for p in content if p.get('type') == 'text')
             continue
 
+        # Handle assistant messages with tool_calls (from convert_output_to_messages)
+        if role == 'assistant' and msg.get('tool_calls'):
+            # Add text content as message if present
+            if content:
+                text = content if isinstance(content, str) else '\n'.join(
+                    p.get('text', '') for p in content if p.get('type') == 'text'
+                )
+                if text.strip():
+                    input_items.append({
+                        'type': 'message', 'role': 'assistant',
+                        'content': [{'type': 'output_text', 'text': text}],
+                    })
+            # Convert each tool_call to a function_call input item
+            for tool_call in msg['tool_calls']:
+                func = tool_call.get('function', {})
+                input_items.append({
+                    'type': 'function_call',
+                    'call_id': tool_call.get('id', ''),
+                    'name': func.get('name', ''),
+                    'arguments': func.get('arguments', '{}'),
+                })
+            continue
+
+        # Handle tool result messages
+        if role == 'tool':
+            input_items.append({
+                'type': 'function_call_output',
+                'call_id': msg.get('tool_call_id', ''),
+                'output': msg.get('content', ''),
+            })
+            continue
+
         # Convert content format
         text_type = 'output_text' if role == 'assistant' else 'input_text'
 
@@ -820,12 +886,21 @@ def convert_to_responses_payload(payload: dict) -> dict:
 
     responses_payload = {**payload, 'input': input_items}
 
+    # Forward previous_response_id when the middleware has set it
+    # (only used when ENABLE_RESPONSES_API_STATEFUL is enabled).
+    previous_response_id = responses_payload.pop('previous_response_id', None)
+    if previous_response_id:
+        responses_payload['previous_response_id'] = previous_response_id
+
     if system_content:
         responses_payload['instructions'] = system_content
 
     if 'max_tokens' in responses_payload:
         responses_payload['max_output_tokens'] = responses_payload.pop('max_tokens')
 
+    if 'max_completion_tokens' in responses_payload:
+        responses_payload['max_output_tokens'] = responses_payload.pop('max_completion_tokens')
+
     # Remove Chat Completions-only parameters not supported by the Responses API
     for unsupported_key in (
         'stream_options',
@@ -864,11 +939,36 @@ def convert_to_responses_payload(payload: dict) -> dict:
 
 def convert_responses_result(response: dict) -> dict:
     """
-    Convert non-streaming Responses API result.
-    Just add done flag - pass through raw response, frontend handles output.
+    Convert non-streaming Responses API result to Chat Completions format.
+
+    Extracts text from message output items so all downstream consumers
+    (frontend tasks, get_content_from_response) work without modification.
     """
-    response['done'] = True
-    return response
+    output_items = response.get('output', [])
+
+    content = ''
+    for item in output_items:
+        if item.get('type') == 'message':
+            for part in item.get('content', []):
+                if part.get('type') == 'output_text':
+                    content += part.get('text', '')
+
+    return {
+        'id': response.get('id', ''),
+        'object': 'chat.completion',
+        'model': response.get('model', ''),
+        'choices': [
+            {
+                'index': 0,
+                'message': {
+                    'role': 'assistant',
+                    'content': content,
+                },
+                'finish_reason': 'stop',
+            }
+        ],
+        'usage': response.get('usage', {}),
+    }
 
 
 @router.post('/chat/completions')
@@ -980,7 +1080,7 @@ async def generate_chat_completion(
     key = request.app.state.config.OPENAI_API_KEYS[idx]
 
     # Check if model is a reasoning model that needs special handling
-    if is_openai_reasoning_model(payload['model']):
+    if is_openai_new_model(payload['model']):
         payload = openai_reasoning_model_handler(payload)
     elif 'api.openai.com' not in url:
         # Remove "max_completion_tokens" from the payload for backward compatibility

backend/open_webui/routers/retrieval.py

@@ -2464,6 +2464,7 @@ async def query_collection_handler(
             )
         else:
             return await query_collection(
+                request,
                 collection_names=form_data.collection_names,
                 queries=[form_data.query],
                 embedding_function=lambda query, prefix: request.app.state.EMBEDDING_FUNCTION(

backend/open_webui/socket/utils.py

@@ -82,13 +82,22 @@ class RedisDict:
         return [(k, json.loads(v)) for k, v in self.redis.hgetall(self.name).items()]
 
     def set(self, mapping: dict):
-        pipe = self.redis.pipeline()
+        if not mapping:
+            self.redis.delete(self.name)
+            return
 
-        pipe.delete(self.name)
-        if mapping:
-            pipe.hset(self.name, mapping={k: json.dumps(v) for k, v in mapping.items()})
+        # Fetch existing keys before writing so we know which ones to remove.
+        # HKEYS is cheap — it transfers only short key strings, not large JSON values.
+        existing_keys = set(self.redis.hkeys(self.name))
+        new_keys = set(mapping.keys())
+        keys_to_remove = existing_keys - new_keys
 
-        pipe.execute()
+        # HSET first (add/update all new values), then HDEL (remove stale keys).
+        # We never DELETE the whole hash — this eliminates the race window
+        # where concurrent readers would see an empty models dict.
+        self.redis.hset(self.name, mapping={k: json.dumps(v) for k, v in mapping.items()})
+        if keys_to_remove:
+            self.redis.hdel(self.name, *keys_to_remove)
 
     def get(self, key, default=None):
         try:

backend/open_webui/tools/builtin.py

@@ -1874,6 +1874,7 @@ async def query_knowledge_files(
         # Query vector collections if any
         if collection_names:
             query_results = await query_collection(
+                __request__,
                 collection_names=collection_names,
                 queries=[query],
                 embedding_function=embedding_function,

backend/open_webui/utils/access_control/__init__.py

@@ -5,6 +5,7 @@ from open_webui.models.users import UserModel
 from open_webui.models.groups import Groups
 from open_webui.models.access_grants import (
     has_public_read_access_grant,
+    has_public_write_access_grant,
     has_user_access_grant,
     strip_user_access_grants,
 )
@@ -225,7 +226,7 @@ def filter_allowed_access_grants(
         return access_grants
 
     # Check if user can share publicly
-    if has_public_read_access_grant(access_grants) and not has_permission(
+    if (has_public_read_access_grant(access_grants) or has_public_write_access_grant(access_grants)) and not has_permission(
         user_id,
         public_permission_key,
         default_permissions,

backend/open_webui/utils/auth.py

@@ -143,7 +143,7 @@ def get_license_data(app, key):
             pn, pt = nt(pb)
 
             data = json.loads(aesgcm.decrypt(pn, pt, None).decode())
-            if not data.get('exp') and data.get('exp') < datetime.now().date():
+            if not data.get('exp') or data.get('exp') < datetime.now().date():
                 return False
 
             data_handler(data)

backend/open_webui/utils/middleware.py

@@ -98,6 +98,7 @@ from open_webui.utils.misc import (
     get_content_from_message,
     convert_output_to_messages,
     merge_assistant_content_into_output_messages,
+    strip_empty_content_blocks,
 )
 from open_webui.utils.tools import (
     get_tools,
@@ -136,6 +137,7 @@ from open_webui.env import (
     ENABLE_FORWARD_USER_INFO_HEADERS,
     FORWARD_SESSION_INFO_HEADER_CHAT_ID,
     FORWARD_SESSION_INFO_HEADER_MESSAGE_ID,
+    ENABLE_RESPONSES_API_STATEFUL,
 )
 from open_webui.utils.headers import include_user_info_headers
 from open_webui.constants import TASKS
@@ -850,7 +852,11 @@ def handle_responses_streaming_event(
                 if item.get('type') == 'reasoning' and item.get('status') != 'completed':
                     item['status'] = 'completed'
 
-        return new_output, {'usage': response_data.get('usage'), 'done': True}
+        return new_output, {
+            'usage': response_data.get('usage'),
+            'done': True,
+            'response_id': response_data.get('id'),
+        }
 
     elif event_type == 'response.in_progress':
         # State Machine Event: In Progress
@@ -938,6 +944,13 @@ def process_tool_result(
     tool_result_embeds = []
     EXTERNAL_TOOL_TYPES = ('external', 'action', 'terminal')
 
+    # Support (HTMLResponse, result_context) tuples: the optional second
+    # element lets tool authors provide the LLM with actionable context
+    # about the generated embed instead of the generic fallback message.
+    result_context = None
+    if isinstance(tool_result, tuple) and len(tool_result) == 2 and isinstance(tool_result[0], HTMLResponse):
+        tool_result, result_context = tool_result
+
     if isinstance(tool_result, HTMLResponse):
         content_disposition = tool_result.headers.get('Content-Disposition', '')
         if 'inline' in content_disposition:
@@ -945,11 +958,14 @@ def process_tool_result(
             tool_result_embeds.append(content)
 
             if 200 <= tool_result.status_code < 300:
-                tool_result = {
-                    'status': 'success',
-                    'code': 'ui_component',
-                    'message': f'{tool_function_name}: Embedded UI result is active and visible to the user.',
-                }
+                if result_context is not None and isinstance(result_context, (str, dict, list)):
+                    tool_result = result_context
+                else:
+                    tool_result = {
+                        'status': 'success',
+                        'code': 'ui_component',
+                        'message': f'{tool_function_name}: Embedded UI result is active and visible to the user.',
+                    }
             elif 400 <= tool_result.status_code < 500:
                 tool_result = {
                     'status': 'error',
@@ -1000,20 +1016,37 @@ def process_tool_result(
                 )
 
                 if 'text/html' in content_type:
+                    # Support (html_content, result_context) nested tuple
+                    result_context = None
+                    html_content = tool_result
+                    if isinstance(tool_result, (tuple, list)) and len(tool_result) == 2:
+                        html_content, result_context = tool_result
+
                     # Display as iframe embed
-                    tool_result_embeds.append(tool_result)
-                    tool_result = {
-                        'status': 'success',
-                        'code': 'ui_component',
-                        'message': f'{tool_function_name}: Embedded UI result is active and visible to the user.',
-                    }
+                    tool_result_embeds.append(html_content)
+                    if result_context is not None and isinstance(result_context, (str, dict, list)):
+                        tool_result = result_context
+                    else:
+                        tool_result = {
+                            'status': 'success',
+                            'code': 'ui_component',
+                            'message': f'{tool_function_name}: Embedded UI result is active and visible to the user.',
+                        }
                 elif location:
+                    # Support (html_content, result_context) nested tuple for location embeds
+                    result_context = None
+                    if isinstance(tool_result, (tuple, list)) and len(tool_result) == 2:
+                        _, result_context = tool_result
+
                     tool_result_embeds.append(location)
-                    tool_result = {
-                        'status': 'success',
-                        'code': 'ui_component',
-                        'message': f'{tool_function_name}: Embedded UI result is active and visible to the user.',
-                    }
+                    if result_context is not None and isinstance(result_context, (str, dict, list)):
+                        tool_result = result_context
+                    else:
+                        tool_result = {
+                            'status': 'success',
+                            'code': 'ui_component',
+                            'message': f'{tool_function_name}: Embedded UI result is active and visible to the user.',
+                        }
 
     tool_result_files = []
 
@@ -2531,7 +2564,7 @@ async def process_chat_payload(request, form_data, user, metadata, model):
         # so system terminals work even when no other tools are selected)
         if terminal_id:
             try:
-                terminal_tools = await get_terminal_tools(
+                terminal_tools, system_prompt = await get_terminal_tools(
                     request,
                     terminal_id,
                     user,
@@ -2539,11 +2572,25 @@ async def process_chat_payload(request, form_data, user, metadata, model):
                 )
                 if terminal_tools:
                     tools_dict = {**tools_dict, **terminal_tools}
+                if system_prompt:
+                    form_data['messages'] = add_or_update_system_message(
+                        system_prompt,
+                        form_data['messages'],
+                        append=True,
+                    )
             except Exception as e:
                 log.exception(e)
 
         if direct_tool_servers:
             for tool_server in direct_tool_servers:
+                system_prompt = tool_server.pop('system_prompt', None)
+                if system_prompt:
+                    form_data['messages'] = add_or_update_system_message(
+                        system_prompt,
+                        form_data['messages'],
+                        append=True,
+                    )
+
                 tool_specs = tool_server.pop('specs', [])
 
                 for tool in tool_specs:
@@ -2641,6 +2688,10 @@ async def process_chat_payload(request, form_data, user, metadata, model):
             }
         )
 
+    # Strip empty text content blocks from multimodal messages
+    # to prevent errors from providers like Gemini and Claude
+    form_data['messages'] = strip_empty_content_blocks(form_data.get('messages', []))
+
     return form_data, metadata, events
 
 
@@ -2891,7 +2942,7 @@ async def background_tasks_handler(ctx):
                                 }
                             )
 
-                    if title == None and len(messages) == 2:
+                    if title == None and len(messages) == 2 and (not messages_map or len(messages_map) <= 2):
                         title = messages[0].get('content', user_message)
 
                         Chats.update_chat_title_by_id(metadata['chat_id'], title)
@@ -3034,6 +3085,7 @@ async def non_streaming_chat_response_handler(response, ctx):
                         metadata['chat_id'],
                         metadata['message_id'],
                         {
+                            'done': True,
                             'role': 'assistant',
                             'content': content,
                             'output': response_output,
@@ -3366,6 +3418,11 @@ async def streaming_chat_response_handler(response, ctx):
                     output = []
 
             usage = None
+            prior_output = []
+            last_response_id = None
+
+            def full_output():
+                return prior_output + output if prior_output else output
 
             reasoning_tags_param = metadata.get('params', {}).get('reasoning_tags')
             DETECT_REASONING_TAGS = reasoning_tags_param is not False
@@ -3400,6 +3457,8 @@ async def streaming_chat_response_handler(response, ctx):
                     nonlocal content
                     nonlocal usage
                     nonlocal output
+                    nonlocal prior_output
+                    nonlocal last_response_id
 
                     response_tool_calls = []
 
@@ -3471,19 +3530,29 @@ async def streaming_chat_response_handler(response, ctx):
                                     )
                                 # Check for Responses API events (type field starts with "response.")
                                 elif data.get('type', '').startswith('response.'):
+
                                     output, response_metadata = handle_responses_streaming_event(data, output)
 
                                     processed_data = {
-                                        'output': output,
-                                        'content': serialize_output(output),
+                                        'output': full_output(),
+                                        'content': serialize_output(full_output()),
                                     }
 
                                     # print(data)
                                     # print(processed_data)
 
-                                    # Merge any metadata (usage, done, etc.)
+                                    # Merge any metadata (usage, etc.)
+                                    # Strip 'done' — response.completed emits
+                                    # it but we may still need to execute tool
+                                    # calls. The outer middleware manages the
+                                    # actual completion signal.
                                     if response_metadata:
+                                        if ENABLE_RESPONSES_API_STATEFUL:
+                                            response_id = response_metadata.pop('response_id', None)
+                                            if response_id:
+                                                last_response_id = response_id
                                         processed_data.update(response_metadata)
+                                        processed_data.pop('done', None)
 
                                     await event_emitter(
                                         {
@@ -3825,13 +3894,13 @@ async def streaming_chat_response_handler(response, ctx):
                                                 metadata['chat_id'],
                                                 metadata['message_id'],
                                                 {
-                                                    'content': serialize_output(output),
-                                                    'output': output,
+                                                    'content': serialize_output(full_output()),
+                                                    'output': full_output(),
                                                 },
                                             )
                                         else:
                                             data = {
-                                                'content': serialize_output(output),
+                                                'content': serialize_output(full_output()),
                                             }
 
                                 if delta:
@@ -3888,6 +3957,34 @@ async def streaming_chat_response_handler(response, ctx):
                     if response_tool_calls:
                         tool_calls.append(_split_tool_calls(response_tool_calls))
 
+                    # Responses API path: extract function_call items from output
+                    if not response_tool_calls and output:
+                        # Collect call_ids that already have results,
+                        # including those from prior_output so we don't
+                        # re-process tool calls from a previous turn.
+                        handled_call_ids = {
+                            item.get('call_id')
+                            for item in (prior_output + output)
+                            if item.get('type') == 'function_call_output'
+                        }
+                        responses_api_tool_calls = []
+                        for item in output:
+                            if (
+                                item.get('type') == 'function_call'
+                                and item.get('call_id') not in handled_call_ids
+                            ):
+                                arguments = item.get('arguments', '{}')
+                                responses_api_tool_calls.append({
+                                    'id': item.get('call_id', ''),
+                                    'index': len(responses_api_tool_calls),
+                                    'function': {
+                                        'name': item.get('name', ''),
+                                        'arguments': arguments if isinstance(arguments, str) else json.dumps(arguments),
+                                    },
+                                })
+                        if responses_api_tool_calls:
+                            tool_calls.append(_split_tool_calls(responses_api_tool_calls))
+
                     if response.background:
                         await response.background()
 
@@ -3919,26 +4016,32 @@ async def streaming_chat_response_handler(response, ctx):
                     response_tool_calls = tool_calls.pop(0)
 
                     # Append function_call items for each tool call
+                    # (Responses API already has them from streaming, so skip duplicates)
+                    existing_call_ids = {
+                        item.get('call_id') for item in output
+                        if item.get('type') == 'function_call'
+                    }
                     for tc in response_tool_calls:
                         call_id = tc.get('id', '')
-                        func = tc.get('function', {})
-                        output.append(
-                            {
-                                'type': 'function_call',
-                                'id': call_id or output_id('fc'),
-                                'call_id': call_id,
-                                'name': func.get('name', ''),
-                                'arguments': func.get('arguments', '{}'),
-                                'status': 'in_progress',
-                            }
-                        )
+                        if call_id not in existing_call_ids:
+                            func = tc.get('function', {})
+                            output.append(
+                                {
+                                    'type': 'function_call',
+                                    'id': call_id or output_id('fc'),
+                                    'call_id': call_id,
+                                    'name': func.get('name', ''),
+                                    'arguments': func.get('arguments', '{}'),
+                                    'status': 'in_progress',
+                                }
+                            )
 
                     await event_emitter(
                         {
                             'type': 'chat:completion',
                             'data': {
-                                'content': serialize_output(output),
-                                'output': output,
+                                'content': serialize_output(full_output()),
+                                'output': full_output(),
                             },
                         }
                     )
@@ -4180,11 +4283,20 @@ async def streaming_chat_response_handler(response, ctx):
                             **form_data,
                             'model': model_id,
                             'stream': True,
-                            'messages': [
+                        }
+
+                        if ENABLE_RESPONSES_API_STATEFUL and last_response_id:
+                            system_message = get_system_message(form_data['messages'])
+                            new_form_data['messages'] = (
+                                ([system_message] if system_message else [])
+                                + convert_output_to_messages(output, raw=True)
+                            )
+                            new_form_data['previous_response_id'] = last_response_id
+                        else:
+                            new_form_data['messages'] = [
                                 *form_data['messages'],
                                 *convert_output_to_messages(output, raw=True),
-                            ],
-                        }
+                            ]
 
                         res = await generate_chat_completion(
                             request,
@@ -4194,7 +4306,31 @@ async def streaming_chat_response_handler(response, ctx):
                         )
 
                         if isinstance(res, StreamingResponse):
+                            # Save accumulated output and start fresh.
+                            # Responses API output_index values are relative
+                            # to the current response — a clean output list
+                            # keeps indices aligned. The display prefix
+                            # ensures the UI shows tool history during
+                            # streaming.
+                            prior_output = list(output)
+                            # Trim the trailing empty placeholder message
+                            # so it doesn't persist as a ghost item once
+                            # the new stream produces real content.
+                            if (
+                                prior_output
+                                and prior_output[-1].get('type') == 'message'
+                                and prior_output[-1].get('status') == 'in_progress'
+                            ):
+                                msg_parts = prior_output[-1].get('content', [])
+                                if (
+                                    not msg_parts
+                                    or (len(msg_parts) == 1 and not msg_parts[0].get('text', '').strip())
+                                ):
+                                    prior_output.pop()
+                            output = []
                             await stream_body_handler(res, new_form_data)
+                            output[:0] = prior_output
+                            prior_output = []
                         else:
                             break
                     except Exception as e:
@@ -4383,6 +4519,7 @@ async def streaming_chat_response_handler(response, ctx):
                         metadata['chat_id'],
                         metadata['message_id'],
                         {
+                            'done': True,
                             'content': serialize_output(output),
                             'output': output,
                             **({'usage': usage} if usage else {}),
@@ -4392,7 +4529,13 @@ async def streaming_chat_response_handler(response, ctx):
                     Chats.upsert_message_to_chat_by_id_and_message_id(
                         metadata['chat_id'],
                         metadata['message_id'],
-                        {'usage': usage},
+                        {'done': True, 'usage': usage},
+                    )
+                else:
+                    Chats.upsert_message_to_chat_by_id_and_message_id(
+                        metadata['chat_id'],
+                        metadata['message_id'],
+                        {'done': True},
                     )
 
                 # Send a webhook notification if the user is not active
@@ -4429,10 +4572,17 @@ async def streaming_chat_response_handler(response, ctx):
                         metadata['chat_id'],
                         metadata['message_id'],
                         {
+                            'done': True,
                             'content': serialize_output(output),
                             'output': output,
                         },
                     )
+                else:
+                    Chats.upsert_message_to_chat_by_id_and_message_id(
+                        metadata['chat_id'],
+                        metadata['message_id'],
+                        {'done': True},
+                    )
 
             if response.background is not None:
                 await response.background()

backend/open_webui/utils/misc.py

@@ -446,6 +446,31 @@ def append_or_update_assistant_message(content: str, messages: list[dict]):
     return messages
 
 
+def strip_empty_content_blocks(messages: list[dict]) -> list[dict]:
+    """
+    Remove empty text content blocks from multimodal message content arrays.
+
+    Providers like Gemini and Claude reject messages where a text block has
+    an empty string.  This can happen when a user sends only file/image
+    attachments without typing any text.
+    """
+    for message in messages:
+        content = message.get('content')
+        if isinstance(content, list):
+            cleaned = [
+                block
+                for block in content
+                if not (
+                    isinstance(block, dict)
+                    and block.get('type') == 'text'
+                    and not block.get('text', '').strip()
+                )
+            ]
+            if cleaned:
+                message['content'] = cleaned
+    return messages
+
+
 def openai_chat_message_template(model: str):
     return {
         'id': f'{model}-{str(uuid.uuid4())}',

backend/open_webui/utils/oauth.py

@@ -1508,7 +1508,7 @@ class OAuthManager:
                 data={'id': user.id},
                 expires_delta=parse_duration(auth_manager_config.JWT_EXPIRES_IN),
             )
-            if auth_manager_config.ENABLE_OAUTH_GROUP_MANAGEMENT and user.role != 'admin':
+            if auth_manager_config.ENABLE_OAUTH_GROUP_MANAGEMENT:
                 self.update_user_groups(
                     user=user,
                     user_data=user_data,
@@ -1533,6 +1533,10 @@ class OAuthManager:
 
         response = RedirectResponse(url=redirect_url, headers=response.headers)
 
+        # Compute cookie expiry from JWT lifetime
+        expires_delta = parse_duration(auth_manager_config.JWT_EXPIRES_IN)
+        cookie_max_age = int(expires_delta.total_seconds()) if expires_delta else None
+
         # Set the cookie token
         # Redirect back to the frontend with the JWT token
         response.set_cookie(
@@ -1541,6 +1545,7 @@ class OAuthManager:
             httponly=False,  # Required for frontend access
             samesite=WEBUI_AUTH_COOKIE_SAME_SITE,
             secure=WEBUI_AUTH_COOKIE_SECURE,
+            **({'max_age': cookie_max_age} if cookie_max_age is not None else {}),
         )
 
         # Legacy cookies for compatibility with older frontend versions
@@ -1551,6 +1556,7 @@ class OAuthManager:
                 httponly=True,
                 samesite=WEBUI_AUTH_COOKIE_SAME_SITE,
                 secure=WEBUI_AUTH_COOKIE_SECURE,
+                **({'max_age': cookie_max_age} if cookie_max_age is not None else {}),
             )
 
         try:
@@ -1588,6 +1594,7 @@ class OAuthManager:
                     httponly=True,
                     samesite=WEBUI_AUTH_COOKIE_SAME_SITE,
                     secure=WEBUI_AUTH_COOKIE_SECURE,
+                    **({'max_age': cookie_max_age, 'expires': cookie_expires} if cookie_max_age is not None else {}),
                 )
 
                 log.info(f'Stored OAuth session server-side for user {user.id}, provider {provider}')

backend/open_webui/utils/tools.py

@@ -828,6 +828,43 @@ async def get_terminal_cwd(
     return None
 
 
+async def get_terminal_system_prompt(
+    base_url: str,
+    headers: dict,
+    cookies: Optional[dict] = None,
+) -> Optional[str]:
+    """Fetch the system prompt from a terminal server.
+
+    Checks ``/api/config`` for the ``system`` feature flag first;
+    only fetches ``/system`` if the flag is present.  Returns *None*
+    silently when the server doesn't support the endpoint.
+    """
+    base = base_url.rstrip('/')
+    try:
+        async with aiohttp.ClientSession(
+            timeout=aiohttp.ClientTimeout(total=3),
+            trust_env=True,
+        ) as session:
+            # 1. Check feature flag
+            async with session.get(f'{base}/api/config') as resp:
+                if resp.status != 200:
+                    return None
+                config = await resp.json()
+                if not config.get('features', {}).get('system'):
+                    return None
+
+            # 2. Fetch system prompt
+            async with session.get(
+                f'{base}/system', headers=headers, cookies=cookies or {}
+            ) as resp:
+                if resp.status == 200:
+                    data = await resp.json()
+                    return data.get('prompt')
+    except Exception as e:
+        log.debug(f'Failed to fetch terminal system prompt: {e}')
+    return None
+
+
 async def set_terminal_servers(request: Request):
     """Load and cache OpenAPI specs from all TERMINAL_SERVER_CONNECTIONS."""
     connections = request.app.state.config.TERMINAL_SERVER_CONNECTIONS or []
@@ -867,6 +904,25 @@ async def set_terminal_servers(request: Request):
 
     request.app.state.TERMINAL_SERVERS = await get_tool_servers_data(server_configs)
 
+    # Fetch system prompts concurrently (runs at cache time, not per-request)
+    connections_by_id = {c.get('id'): c for c in connections if c.get('id')}
+
+    async def _fetch_system_prompt(server):
+        connection = connections_by_id.get(server.get('id'))
+        if not connection:
+            return
+        headers = {}
+        if connection.get('auth_type', 'bearer') == 'bearer':
+            headers['Authorization'] = f'Bearer {connection.get("key", "")}'
+        prompt = await get_terminal_system_prompt(server['url'], headers)
+        if prompt:
+            server['system_prompt'] = prompt
+
+    await asyncio.gather(
+        *[_fetch_system_prompt(s) for s in request.app.state.TERMINAL_SERVERS],
+        return_exceptions=True,
+    )
+
     if request.app.state.redis is not None:
         await request.app.state.redis.set('terminal_servers', json.dumps(request.app.state.TERMINAL_SERVERS))
 
@@ -894,7 +950,7 @@ async def get_terminal_tools(
     terminal_id: str,
     user: UserModel,
     extra_params: dict,
-) -> dict[str, dict]:
+) -> tuple[dict[str, dict], Optional[str]]:
     """Resolve tools for a terminal server identified by terminal_id.
 
     - Finds the connection in TERMINAL_SERVER_CONNECTIONS
@@ -941,14 +997,14 @@ async def get_terminal_tools(
             headers['Authorization'] = f'Bearer {oauth_token.get("access_token", "")}'
     # auth_type == "none": no Authorization header
 
+    system_prompt = server_data.get('system_prompt')
     terminal_cwd = await get_terminal_cwd(connection.get('url', ''), headers, cookies)
 
     tools_dict = {}
     for spec in specs:
         function_name = spec['name']
-
-        # Inject CWD into run_command description
         tool_spec = clean_openai_tool_schema(spec)
+
         if function_name == 'run_command' and terminal_cwd:
             tool_spec['description'] = (
                 tool_spec.get('description', '') + f'\n\nThe current working directory is: {terminal_cwd}'
@@ -977,7 +1033,7 @@ async def get_terminal_tools(
             'type': 'terminal',
         }
 
-    return tools_dict
+    return tools_dict, system_prompt
 
 
 async def get_tool_server_data(url: str, headers: Optional[dict]) -> Dict[str, Any]:

pyproject.toml

@@ -216,6 +216,10 @@ dev = [
     "ruff>=0.15.5",
 ]
 
+[tool.black]
+line-length = 120
+skip-string-normalization = true
+
 [tool.ruff]
 line-length = 120
 

src/lib/apis/index.ts

@@ -392,12 +392,38 @@ export const getToolServersData = async (servers: object[]) => {
 							specs: convertOpenApiToToolPayload(res)
 						};
 
-						return {
+						const result: Record<string, any> = {
 							url: server?.url,
 							openapi: openapi,
 							info: info,
 							specs: specs
 						};
+
+						// Fetch system prompt if the server supports it
+						try {
+							const baseUrl = (server?.url ?? '').replace(/\/$/, '');
+							const configRes = await fetch(`${baseUrl}/api/config`);
+							if (configRes.ok) {
+								const config = await configRes.json();
+								if (config?.features?.system) {
+									const headers: Record<string, string> = {};
+									if (toolServerToken) {
+										headers['Authorization'] = `Bearer ${toolServerToken}`;
+									}
+									const systemRes = await fetch(`${baseUrl}/system`, { headers });
+									if (systemRes.ok) {
+										const systemData = await systemRes.json();
+										if (systemData?.prompt) {
+											result.system_prompt = systemData.prompt;
+										}
+									}
+								}
+							}
+						} catch (e) {
+							// Server doesn't support /system — that's fine
+						}
+
+						return result;
 					} else if (error) {
 						return {
 							error,

src/lib/components/ImportModal.svelte

@@ -6,7 +6,7 @@
 	import Spinner from '$lib/components/common/Spinner.svelte';
 	import Modal from '$lib/components/common/Modal.svelte';
 	import XMark from '$lib/components/icons/XMark.svelte';
-	import { extractFrontmatter } from '$lib/utils';
+	import { extractFrontmatter, nameToId } from '$lib/utils';
 
 	export let show = false;
 
@@ -42,7 +42,7 @@
 			toast.success(successMessage);
 
 			let func = res;
-			func.id = func.id || func.name.replace(/\s+/g, '_').toLowerCase();
+			func.id = func.id || nameToId(func.name);
 
 			const frontmatter = extractFrontmatter(res.content); // Ensure frontmatter is extracted
 

src/lib/components/admin/Functions/FunctionEditor.svelte

@@ -4,6 +4,7 @@
 
 	const i18n = getContext('i18n');
 
+	import { nameToId } from '$lib/utils';
 	import CodeEditor from '$lib/components/common/CodeEditor.svelte';
 	import ConfirmDialog from '$lib/components/common/ConfirmDialog.svelte';
 	import Badge from '$lib/components/common/Badge.svelte';
@@ -36,7 +37,7 @@
 	};
 
 	$: if (name && !edit && !clone) {
-		id = name.replace(/\s+/g, '_').toLowerCase();
+		id = nameToId(name);
 	}
 
 	let codeEditor;

src/lib/components/channel/Messages/Message.svelte

@@ -66,7 +66,7 @@
 	let showDeleteConfirmDialog = false;
 
 	const loadMessageData = async () => {
-		if (message && message?.data) {
+		if (message && message?.data === true) {
 			const res = await getMessageData(localStorage.token, channel?.id, message.id);
 			if (res) {
 				message.data = res;
@@ -75,7 +75,7 @@
 	};
 
 	onMount(async () => {
-		if (message && message?.data) {
+		if (message && message?.data === true) {
 			await loadMessageData();
 		}
 	});
@@ -285,12 +285,15 @@
 								e.currentTarget.src = '/favicon.png';
 							}}
 						/>
+					{:else if message.user?.role === 'webhook'}
+						<ProfileImage
+							src={`${WEBUI_API_BASE_URL}/channels/webhooks/${message.user?.id}/profile/image`}
+							className={'size-8 ml-0.5'}
+						/>
 					{:else}
 						<ProfilePreview user={message.user}>
 							<ProfileImage
-								src={message.user?.role === 'webhook'
-									? `${WEBUI_API_BASE_URL}/channels/webhooks/${message.user?.id}/profile/image`
-									: `${WEBUI_API_BASE_URL}/users/${message.user?.id}/profile/image`}
+								src={`${WEBUI_API_BASE_URL}/users/${message.user?.id}/profile/image`}
 								className={'size-8 ml-0.5'}
 							/>
 						</ProfilePreview>

src/lib/components/chat/Chat.svelte

@@ -45,7 +45,8 @@
 		showEmbeds,
 		selectedTerminalId,
 		showFileNavPath,
-		showFileNavDir
+		showFileNavDir,
+		chatRequestQueues
 	} from '$lib/stores';
 
 	import { WEBUI_API_BASE_URL } from '$lib/constants';
@@ -170,8 +171,7 @@
 	let files = [];
 	let params = {};
 
-	// Message queue for storing messages while generating
-	let messageQueue: { id: string; prompt: string; files: any[] }[] = [];
+
 
 	$: if (chatIdProp) {
 		navigateHandler();
@@ -180,16 +180,10 @@
 	const navigateHandler = async () => {
 		loading = true;
 
-		// Save current queue to sessionStorage before navigating away
-		if (messageQueue.length > 0 && $chatId) {
-			sessionStorage.setItem(`chat-queue-${$chatId}`, JSON.stringify(messageQueue));
-		}
-
 		prompt = '';
 		messageInput?.setText('');
 
 		files = [];
-		messageQueue = [];
 		selectedToolIds = [];
 		selectedFilterIds = [];
 		webSearchEnabled = false;
@@ -206,28 +200,11 @@
 
 			await tick();
 
-			// Restore queue from sessionStorage
-			const storedQueueData = sessionStorage.getItem(`chat-queue-${chatIdProp}`);
-			if (storedQueueData) {
-				try {
-					const restoredQueue = JSON.parse(storedQueueData);
-
-					if (restoredQueue.length > 0) {
-						sessionStorage.removeItem(`chat-queue-${chatIdProp}`);
-						// Check if there are pending tasks (still generating)
-						const hasPendingTask = taskIds !== null && taskIds.length > 0;
-						if (!hasPendingTask) {
-							// No pending tasks - process the queue
-							files = restoredQueue.flatMap((m) => m.files);
-							await tick();
-							const combinedPrompt = restoredQueue.map((m) => m.prompt).join('\n\n');
-							await submitPrompt(combinedPrompt);
-						} else {
-							// Has pending tasks - show as queued (chatCompletedHandler will process)
-							messageQueue = restoredQueue;
-						}
-					}
-				} catch (e) {}
+			// Process any queued requests if the chat is idle
+			const lastMessage = history.currentId ? history.messages[history.currentId] : null;
+			const isIdle = !lastMessage || lastMessage.role !== 'assistant' || lastMessage.done;
+			if (isIdle) {
+				await processNextInQueue(chatIdProp);
 			}
 
 			if (storageChatInput) {
@@ -438,11 +415,15 @@
 				} else if (type === 'chat:completion') {
 					chatCompletionEventHandler(data, message, event.chat_id);
 				} else if (type === 'chat:tasks:cancel') {
-					taskIds = null;
-					const responseMessage = history.messages[history.currentId];
-					// Set all response messages to done
-					for (const messageId of history.messages[responseMessage.parentId].childrenIds) {
-						history.messages[messageId].done = true;
+					if (event.message_id === history.currentId) {
+						taskIds = null;
+						// Set all response messages to done
+						for (const messageId of history.messages[message.parentId].childrenIds) {
+							history.messages[messageId].done = true;
+						}
+						await processNextInQueue($chatId);
+					} else {
+						message.done = true;
 					}
 				} else if (type === 'chat:message:delta' || type === 'message') {
 					message.content += data.content;
@@ -559,6 +540,9 @@
 
 				history.messages[event.message_id] = message;
 			}
+		} else {
+			// Non-active chat completion: queue stays in the global store.
+			// navigateHandler will process it when the user returns to that chat.
 		}
 	};
 
@@ -566,11 +550,20 @@
 		origin: string;
 		data: { type: string; text: string };
 	}) => {
-		if (event.origin !== window.origin) {
+		const isSameOrigin = event.origin === window.origin;
+		const type = event.data?.type;
+
+		// Prompt-related message types only submit text to the chat input —
+		// functionally equivalent to the user typing.  When same-origin is
+		// enabled they go through immediately.  When it is disabled (opaque
+		// origin) we show a confirmation dialog so the user stays in control.
+		const iframePromptTypes = ['input:prompt', 'input:prompt:submit', 'action:submit'];
+
+		if (!isSameOrigin && !iframePromptTypes.includes(type)) {
 			return;
 		}
 
-		if (event.data.type === 'action:submit') {
+		if (type === 'action:submit') {
 			console.debug(event.data.text);
 
 			if (prompt !== '') {
@@ -579,8 +572,7 @@
 			}
 		}
 
-		// Replace with your iframe's origin
-		if (event.data.type === 'input:prompt') {
+		if (type === 'input:prompt') {
 			console.debug(event.data.text);
 
 			const inputElement = document.getElementById('chat-input');
@@ -591,12 +583,26 @@
 			}
 		}
 
-		if (event.data.type === 'input:prompt:submit') {
+		if (type === 'input:prompt:submit') {
 			console.debug(event.data.text);
 
 			if (event.data.text !== '') {
-				await tick();
-				submitPrompt(event.data.text);
+				if (isSameOrigin) {
+					await tick();
+					submitPrompt(event.data.text);
+				} else {
+					// Cross-origin: ask user to confirm before submitting
+					eventConfirmationInput = false;
+					eventConfirmationTitle = $i18n.t('Confirm Prompt from Embed');
+					eventConfirmationMessage = event.data.text;
+					eventCallback = async (confirmed: boolean) => {
+						if (confirmed) {
+							await tick();
+							submitPrompt(event.data.text);
+						}
+					};
+					showEventConfirmation = true;
+				}
 			}
 		}
 	};
@@ -637,11 +643,14 @@
 		const audioQueueInstance = new AudioQueue(document.getElementById('audioElement'));
 		audioQueue.set(audioQueueInstance);
 
-		// Reset direct terminal enabled states — selectedTerminalId starts null on every page load
-		if ($settings?.terminalServers?.some((s) => s.enabled)) {
+		// Restore direct terminal enabled states based on persisted selectedTerminalId
+		if ($settings?.terminalServers?.length) {
 			settings.set({
 				...$settings,
-				terminalServers: ($settings.terminalServers ?? []).map((s) => ({ ...s, enabled: false }))
+				terminalServers: ($settings.terminalServers ?? []).map((s) => ({
+					...s,
+					enabled: $selectedTerminalId !== null && s.url === $selectedTerminalId
+				}))
 			});
 		}
 
@@ -936,11 +945,11 @@
 
 	const onHistoryChange = (history) => {
 		if (history) {
-			cancelAnimationFrame(contentsRAF);
-			contentsRAF = requestAnimationFrame(() => {
+			clearTimeout(contentsRAF);
+			contentsRAF = setTimeout(() => {
 				getContents();
 				contentsRAF = null;
-			});
+			}, 0);
 		} else {
 			artifactContents.set([]);
 		}
@@ -955,13 +964,12 @@
 			if (message?.role !== 'user' && message?.content) {
 				const {
 					codeBlocks: codeBlocks,
-					html: htmlContent,
-					css: cssContent,
-					js: jsContent
+					htmlGroups: htmlGroups
 				} = getCodeBlockContents(message.content);
 
-				if (htmlContent || cssContent || jsContent) {
-					const renderedContent = `
+				if (htmlGroups && htmlGroups.length > 0) {
+					htmlGroups.forEach((group) => {
+						const renderedContent = `
                         <!DOCTYPE html>
                         <html lang="en">
                         <head>
@@ -972,19 +980,20 @@
 									background-color: white; /* Ensure the iframe has a white background */
 								}
 
-								${cssContent}
+								${group.css}
 							</${''}style>
                         </head>
                         <body>
-                            ${htmlContent}
+                            ${group.html}
 
 							<${''}script>
-                            	${jsContent}
+                            	${group.js}
 							</${''}script>
                         </body>
                         </html>
                     `;
-					contents = [...contents, { type: 'iframe', content: renderedContent }];
+						contents = [...contents, { type: 'iframe', content: renderedContent }];
+					});
 				} else {
 					// Check for SVG content
 					for (const block of codeBlocks) {
@@ -1130,7 +1139,6 @@
 		chatFiles = [];
 		params = {};
 		taskIds = null;
-		messageQueue = [];
 
 		if ($page.url.searchParams.get('youtube')) {
 			await uploadWeb(`https://www.youtube.com/watch?v=${$page.url.searchParams.get('youtube')}`);
@@ -1239,7 +1247,7 @@
 
 				if (history.currentId) {
 					for (const message of Object.values(history.messages)) {
-						if (message && message.role === 'assistant') {
+						if (message && message.role === 'assistant' && message.done !== false) {
 							message.done = true;
 						}
 					}
@@ -1282,6 +1290,24 @@
 			});
 		}
 	};
+
+	const processNextInQueue = async (targetChatId: string) => {
+		const queue = $chatRequestQueues[targetChatId];
+		if (!queue || queue.length === 0) return;
+
+		const combinedPrompt = queue.map((m) => m.prompt).join('\n\n');
+		const combinedFiles = queue.flatMap((m) => m.files);
+
+		chatRequestQueues.update((q) => {
+			const { [targetChatId]: _, ...rest } = q;
+			return rest;
+		});
+
+		files = combinedFiles;
+		await tick();
+		await submitPrompt(combinedPrompt);
+	};
+
 	const chatCompletedHandler = async (_chatId, modelId, responseMessageId, messages) => {
 		const res = await chatCompleted(localStorage.token, {
 			model: modelId,
@@ -1691,16 +1717,8 @@
 				createMessagesList(history, message.id)
 			);
 
-			// Process message queue immediately after main response finishes
-			if (messageQueue.length > 0) {
-				const combinedPrompt = messageQueue.map((m) => m.prompt).join('\n\n');
-				const combinedFiles = messageQueue.flatMap((m) => m.files);
-				messageQueue = [];
-
-				files = combinedFiles;
-				await tick();
-				await submitPrompt(combinedPrompt);
-			}
+			// Process next queued request if any
+			await processNextInQueue(chatId);
 		}
 
 		console.log(data);
@@ -1764,16 +1782,15 @@
 
 		if (isGenerating) {
 			if ($settings?.enableMessageQueue ?? true) {
-				// Queue the message
+				// Enqueue the request
 				const _files = structuredClone(files);
-				messageQueue = [
-					...messageQueue,
-					{
-						id: uuidv4(),
-						prompt: userPrompt,
-						files: _files
-					}
-				];
+				chatRequestQueues.update((q) => ({
+					...q,
+					[$chatId]: [
+						...(q[$chatId] ?? []),
+						{ id: uuidv4(), prompt: userPrompt, files: _files }
+					]
+				}));
 				// Clear input
 				messageInput?.setText('');
 				prompt = '';
@@ -2375,6 +2392,8 @@
 			generationController?.abort();
 			generationController = null;
 		}
+
+		await processNextInQueue($chatId);
 	};
 
 	const submitMessage = async (parentId, prompt) => {
@@ -2638,12 +2657,8 @@
 			currentChatPage.set(1);
 			initNewChat();
 			await goto('/');
-			getChatList(localStorage.token, $currentChatPage).then((chats) => {
-				chats.set(chats);
-			});
-			getPinnedChatList(localStorage.token).then((pinnedChats) => {
-				pinnedChats.set(pinnedChats);
-			});
+			chats.set(await getChatList(localStorage.token, $currentChatPage));
+			pinnedChats.set(await getPinnedChatList(localStorage.token));
 			toast.success($i18n.t('Chat archived.'));
 		} catch (error) {
 			console.error('Error archiving chat:', error);
@@ -2811,7 +2826,7 @@
 								</div>
 							</div>
 
-							<div class=" pb-2 z-10">
+							<div class=" pb-2 {dragged ? 'z-0' : 'z-10'}">
 								<MessageInput
 									bind:this={messageInput}
 									{history}
@@ -2833,12 +2848,16 @@
 									{stopResponse}
 									{createMessagePair}
 									{onUpload}
-									{messageQueue}
+									messageQueue={$chatRequestQueues[$chatId] ?? []}
 									onQueueSendNow={async (id) => {
-										const item = messageQueue.find((m) => m.id === id);
+										const queue = $chatRequestQueues[$chatId] ?? [];
+										const item = queue.find((m) => m.id === id);
 										if (item) {
 											// Remove from queue
-											messageQueue = messageQueue.filter((m) => m.id !== id);
+											chatRequestQueues.update((q) => ({
+												...q,
+												[$chatId]: queue.filter((m) => m.id !== id)
+											}));
 											// Stop current generation first
 											await stopResponse();
 											await tick();
@@ -2849,17 +2868,25 @@
 										}
 									}}
 									onQueueEdit={(id) => {
-										const item = messageQueue.find((m) => m.id === id);
+										const queue = $chatRequestQueues[$chatId] ?? [];
+										const item = queue.find((m) => m.id === id);
 										if (item) {
 											// Remove from queue
-											messageQueue = messageQueue.filter((m) => m.id !== id);
+											chatRequestQueues.update((q) => ({
+												...q,
+												[$chatId]: queue.filter((m) => m.id !== id)
+											}));
 											// Set files and restore prompt to input
 											files = item.files;
 											messageInput?.setText(item.prompt);
 										}
 									}}
 									onQueueDelete={(id) => {
-										messageQueue = messageQueue.filter((m) => m.id !== id);
+										const queue = $chatRequestQueues[$chatId] ?? [];
+										chatRequestQueues.update((q) => ({
+											...q,
+											[$chatId]: queue.filter((m) => m.id !== id)
+										}));
 									}}
 									onChange={(data) => {
 										if (!$temporaryChatEnabled) {

src/lib/components/chat/MessageInput.svelte

@@ -55,6 +55,7 @@
 	import { uploadFile } from '$lib/apis/files';
 	import { generateAutoCompletion } from '$lib/apis';
 	import { deleteFileById } from '$lib/apis/files';
+	import { getChatById } from '$lib/apis/chats';
 	import { getSessionUser } from '$lib/apis/auths';
 	import { getTools } from '$lib/apis/tools';
 
@@ -806,8 +807,11 @@
 	const onDragOver = (e: DragEvent) => {
 		e.preventDefault();
 
-		// Check if a file is being dragged.
-		if (e.dataTransfer?.types?.includes('Files')) {
+		// Check if a file or a sidebar chat item is being dragged.
+		if (
+			e.dataTransfer?.types?.includes('Files') ||
+			e.dataTransfer?.types?.includes('text/plain')
+		) {
 			dragged = true;
 		} else {
 			dragged = false;
@@ -825,6 +829,35 @@
 		e.preventDefault();
 		console.log(e);
 
+		// Check if the dropped data is a sidebar chat item
+		const textData = e.dataTransfer?.getData('text/plain');
+		if (textData) {
+			try {
+				const data = JSON.parse(textData);
+				if (data.type === 'chat' && data.id) {
+					// Fetch the chat to get its title, then add as a reference chat
+					const chat = await getChatById(localStorage.token, data.id);
+					if (chat) {
+						const chatItem = {
+							type: 'chat',
+							id: chat.id,
+							name: chat.title,
+							collection_name: '',
+							status: 'processed'
+						};
+						if (!files.find((f) => f.id === chatItem.id)) {
+							files = [...files, chatItem];
+						}
+					}
+					dragged = false;
+					e.stopPropagation();
+					return;
+				}
+			} catch (_) {
+				// Not valid JSON — fall through to file handling
+			}
+		}
+
 		if (e.dataTransfer?.files) {
 			const inputFiles = Array.from(e.dataTransfer?.files);
 			if (inputFiles && inputFiles.length > 0) {
@@ -1022,8 +1055,8 @@
 
 			dropzoneElement = document.getElementById('chat-pane');
 			if (dropzoneElement) {
-				dropzoneElement.addEventListener('dragover', onDragOver);
-				dropzoneElement.addEventListener('drop', onDrop);
+				dropzoneElement.addEventListener('dragover', onDragOver, true);
+				dropzoneElement.addEventListener('drop', onDrop, true);
 				dropzoneElement.addEventListener('dragleave', onDragLeave);
 			}
 
@@ -1041,8 +1074,8 @@
 			window.removeEventListener('blur', onBlur);
 
 			if (dropzoneElement) {
-				dropzoneElement.removeEventListener('dragover', onDragOver);
-				dropzoneElement.removeEventListener('drop', onDrop);
+				dropzoneElement.removeEventListener('dragover', onDragOver, true);
+				dropzoneElement.removeEventListener('drop', onDrop, true);
 				dropzoneElement.removeEventListener('dragleave', onDragLeave);
 			}
 		};
@@ -1674,7 +1707,7 @@
 											</Tooltip>
 										{/if}
 
-										{#each selectedFilterIds as filterId}
+										{#each selectedFilterIds as filterId (filterId)}
 											{@const filter = toggleFilters.find((f) => f.id === filterId)}
 											{#if filter}
 												<Tooltip content={filter?.name} placement="top">

src/lib/components/chat/MessageInput/InputVariablesModal.svelte

@@ -17,7 +17,7 @@
 
 	export let onSave = (e) => {};
 
-	let loading = false;
+	let loading = true;
 	let variableValues = {};
 
 	const submitHandler = async () => {
@@ -33,14 +33,17 @@
 
 	const init = async () => {
 		loading = true;
-		variableValues = {};
-		for (const variable of Object.keys(variables)) {
-			if (variables[variable]?.default !== undefined) {
-				variableValues[variable] = variables[variable].default;
+		const newValues = {};
+		const keys = Object.keys(variables ?? {});
+		for (const key of keys) {
+			const variable = variables[key];
+			if (variable?.default !== undefined) {
+				newValues[key] = variable.default;
 			} else {
-				variableValues[variable] = '';
+				newValues[key] = '';
 			}
 		}
+		variableValues = newValues;
 		loading = false;
 
 		await tick();

src/lib/components/chat/Messages/Markdown.svelte

@@ -8,6 +8,7 @@
 	import markedKatexExtension from '$lib/utils/marked/katex-extension';
 	import { disableSingleTilde } from '$lib/utils/marked/strikethrough-extension';
 	import { mentionExtension } from '$lib/utils/marked/mention-extension';
+	import colonFenceExtension from '$lib/utils/marked/colon-fence-extension';
 
 	import MarkdownTokens from './Markdown/MarkdownTokens.svelte';
 	import footnoteExtension from '$lib/utils/marked/footnote-extension';
@@ -48,6 +49,7 @@
 	marked.use(markedExtension(options));
 	marked.use(citationExtension(options));
 	marked.use(footnoteExtension(options));
+	marked.use(colonFenceExtension(options));
 	marked.use(disableSingleTilde);
 	marked.use({
 		extensions: [

src/lib/components/chat/Messages/Markdown/ColonFenceBlock.svelte

@@ -0,0 +1,76 @@
+<script lang="ts">
+	import { getContext } from 'svelte';
+	const i18n = getContext('i18n');
+
+	import { copyToClipboard } from '$lib/utils';
+	import { settings } from '$lib/stores';
+	import MarkdownTokens from './MarkdownTokens.svelte';
+	import Tooltip from '$lib/components/common/Tooltip.svelte';
+	import DocumentDuplicate from '$lib/components/icons/DocumentDuplicate.svelte';
+
+	export let id: string = '';
+	export let token: any;
+	export let tokenIdx: number = 0;
+
+	export let done: boolean = true;
+	export let editCodeBlock: boolean = true;
+	export let sourceIds: string[] = [];
+	export let onTaskClick: Function = () => {};
+	export let onSourceClick: Function = () => {};
+
+	const fenceType: string = token.fenceType ?? 'default';
+
+	const label = fenceType.replace(/_/g, ' ').replace(/\b\w/g, (c) => c.toUpperCase());
+
+	let copied = false;
+
+	const copyText = async () => {
+		copied = true;
+		await copyToClipboard(token.text, null, $settings?.copyFormatted ?? false);
+		setTimeout(() => {
+			copied = false;
+		}, 1000);
+	};
+</script>
+
+<div class="relative group my-2 rounded-2xl border border-gray-100 dark:border-gray-800 px-4 py-3">
+	<!-- Header row: type badge + copy button -->
+	<div class="flex items-center justify-between mb-2">
+		<span class="text-xs font-medium text-gray-500 dark:text-gray-400">
+			{label}
+		</span>
+
+		<div class="invisible group-hover:visible flex gap-0.5">
+			<Tooltip content={copied ? $i18n.t('Copied') : $i18n.t('Copy')}>
+				<button
+					class="p-1 rounded-lg bg-transparent hover:bg-black/5 dark:hover:bg-white/5 transition"
+					on:click={(e) => {
+						e.stopPropagation();
+						copyText();
+					}}
+				>
+					{#if copied}
+						<svg xmlns="http://www.w3.org/2000/svg" fill="none" viewBox="0 0 24 24" stroke-width="1.5" stroke="currentColor" class="size-3.5 text-green-500">
+							<path stroke-linecap="round" stroke-linejoin="round" d="m4.5 12.75 6 6 9-13.5" />
+						</svg>
+					{:else}
+						<DocumentDuplicate className="size-3.5" strokeWidth="1.5" />
+					{/if}
+				</button>
+			</Tooltip>
+		</div>
+	</div>
+
+	<!-- Content -->
+	<div class="prose-sm" dir="auto">
+		<MarkdownTokens
+			id={`${id}-${tokenIdx}-cf`}
+			tokens={token.tokens}
+			{done}
+			{editCodeBlock}
+			{sourceIds}
+			{onTaskClick}
+			{onSourceClick}
+		/>
+	</div>
+</div>

src/lib/components/chat/Messages/Markdown/MarkdownTokens.svelte

@@ -23,6 +23,7 @@
 
 	import HtmlToken from './HTMLToken.svelte';
 	import Clipboard from '$lib/components/icons/Clipboard.svelte';
+	import ColonFenceBlock from './ColonFenceBlock.svelte';
 
 	export let id: string;
 	export let tokens: Token[];
@@ -241,7 +242,7 @@
 					<li class="text-start">
 						{#if item?.task}
 							<input
-								class=" translate-y-[1px] -translate-x-1"
+								class=" translate-y-[1px] -translate-x-1 flex-shrink-0"
 								type="checkbox"
 								checked={item.checked}
 								on:change={(e) => {
@@ -276,7 +277,7 @@
 					<li class="text-start {item?.task ? 'flex -translate-x-6.5 gap-3 ' : ''}">
 						{#if item?.task}
 							<input
-								class=""
+								class="flex-shrink-0"
 								type="checkbox"
 								checked={item.checked}
 								on:change={(e) => {
@@ -434,6 +435,17 @@
 		{#if token.text}
 			<KatexRenderer content={token.text} displayMode={token?.displayMode ?? false} />
 		{/if}
+	{:else if token.type === 'colonFence'}
+		<ColonFenceBlock
+			id={`${id}-${tokenIdx}`}
+			{token}
+			{tokenIdx}
+			{done}
+			{editCodeBlock}
+			{sourceIds}
+			{onTaskClick}
+			{onSourceClick}
+		/>
 	{:else if token.type === 'space'}
 		<div class="my-2" />
 	{:else}

src/lib/components/chat/Messages/ResponseMessage.svelte

@@ -168,6 +168,11 @@
 	let model = null;
 	$: model = $models.find((m) => m.id === message.model);
 
+	$: statusEntries = message?.statusHistory ?? [...(message?.status ? [message?.status] : [])];
+	$: hasVisibleStatus = (model?.info?.meta?.capabilities?.status_updates ?? true)
+		&& statusEntries.length > 0
+		&& !(statusEntries.at(-1)?.hidden ?? false);
+
 	let edit = false;
 	let editedContent = '';
 	let editTextAreaElement: HTMLTextAreaElement;
@@ -197,7 +202,7 @@
 	const stopAudio = () => {
 		try {
 			speechSynthesis.cancel();
-			$audioQueue.stop();
+			$audioQueue?.stop();
 		} catch {}
 
 		if (speaking) {
@@ -779,7 +784,7 @@
 							class="w-full flex flex-col relative {edit ? 'hidden' : ''}"
 							id="response-content-container"
 						>
-							{#if message.content === '' && !message.error && ((model?.info?.meta?.capabilities?.status_updates ?? true) ? (message?.statusHistory ?? [...(message?.status ? [message?.status] : [])]).length === 0 || (message?.statusHistory?.at(-1)?.hidden ?? false) : true)}
+							{#if message.content === '' && !message.done && !message.error && !hasVisibleStatus}
 								<Skeleton />
 							{:else if message.content && message.error !== true}
 								<!-- always show message contents even if there's an error -->
@@ -1010,7 +1015,7 @@
 									</button>
 								</Tooltip>
 
-								{#if $user?.role === 'admin' || ($user?.permissions?.chat?.tts ?? true)}
+								{#if !readOnly && ($user?.role === 'admin' || ($user?.permissions?.chat?.tts ?? true))}
 									<Tooltip content={$i18n.t('Read Aloud')} placement="bottom">
 										<button
 											aria-label={$i18n.t('Read Aloud')}

src/lib/components/chat/Messages/UserMessage.svelte

@@ -135,7 +135,7 @@
 	{#if !($settings?.chatBubble ?? true)}
 		<div class={`shrink-0 ltr:mr-3 rtl:ml-3 mt-1`}>
 			<ProfileImage
-				src={`${WEBUI_API_BASE_URL}/users/${user.id}/profile/image`}
+				src={user?.id ? `${WEBUI_API_BASE_URL}/users/${user.id}/profile/image` : `${WEBUI_BASE_URL}/static/favicon.png`}
 				className={'size-8 user-message-profile-image'}
 			/>
 		</div>
@@ -147,8 +147,8 @@
 					{#if message.user}
 						{$i18n.t('You')}
 						<span class=" text-gray-500 text-sm font-medium">{message?.user ?? ''}</span>
-					{:else if $settings.showUsername || $_user.name !== user.name}
-						{user.name}
+					{:else if $settings.showUsername || $_user?.name !== user?.name}
+						{user?.name ?? $i18n.t('You')}
 					{:else}
 						{$i18n.t('You')}
 					{/if}

src/lib/components/chat/Settings/Personalization/ManageModal.svelte

@@ -49,6 +49,7 @@
 	let selectedMemory = null;
 
 	let showClearConfirmDialog = false;
+	let showDeleteConfirm = false;
 
 	$: filteredMemories = query
 		? memories.filter((m) => m.content?.toLowerCase().includes(query.toLowerCase()))
@@ -225,7 +226,8 @@
 											<Tooltip content={$i18n.t('Edit')}>
 												<button
 													class="self-center w-fit text-sm p-1.5 hover:bg-black/5 dark:hover:bg-white/5 rounded-xl"
-													on:click={() => {
+													on:click={(e) => {
+														e.stopPropagation();
 														selectedMemory = memory;
 														showEditMemoryModal = true;
 													}}
@@ -237,19 +239,10 @@
 											<Tooltip content={$i18n.t('Delete')}>
 												<button
 													class="self-center w-fit text-sm p-1.5 hover:bg-black/5 dark:hover:bg-white/5 rounded-xl"
-													on:click={async () => {
-														const res = await deleteMemoryById(
-															localStorage.token,
-															memory.id
-														).catch((error) => {
-															toast.error(`${error}`);
-															return null;
-														});
-
-														if (res) {
-															toast.success($i18n.t('Memory deleted successfully'));
-															memories = await getMemories(localStorage.token);
-														}
+													on:click={(e) => {
+														e.stopPropagation();
+														selectedMemory = memory;
+														showDeleteConfirm = true;
 													}}
 												>
 													<GarbageBin className="size-4" strokeWidth="1.5" />
@@ -302,6 +295,33 @@
 	}}
 />
 
+<ConfirmDialog
+	title={$i18n.t('Delete Memory?')}
+	show={showDeleteConfirm}
+	on:confirm={async () => {
+		const res = await deleteMemoryById(localStorage.token, selectedMemory.id).catch((error) => {
+			toast.error(`${error}`);
+			return null;
+		});
+
+		if (res) {
+			toast.success($i18n.t('Memory deleted successfully'));
+			memories = await getMemories(localStorage.token);
+		}
+		showDeleteConfirm = false;
+	}}
+	on:cancel={() => {
+		showDeleteConfirm = false;
+	}}
+>
+	<div class=" text-sm text-gray-500 flex-1">
+		{$i18n.t('Are you sure you want to delete this memory? This action cannot be undone.')}
+		<div class=" mt-2 bg-gray-50 dark:bg-gray-900 p-3 rounded-xl border border-gray-100 dark:border-gray-800 text-black dark:text-white whitespace-pre-wrap break-words max-h-32 overflow-y-auto">
+			{selectedMemory?.content}
+		</div>
+	</div>
+</ConfirmDialog>
+
 <AddMemoryModal
 	bind:show={showAddMemoryModal}
 	on:save={async () => {

src/lib/components/common/DragGhost.svelte

@@ -29,7 +29,7 @@
 
 <div
 	bind:this={popupElement}
-	class="fixed top-0 left-0 w-screen h-[100dvh] z-50 touch-none pointer-events-none"
+	class="fixed top-0 left-0 w-screen h-[100dvh] z-[99999] touch-none pointer-events-none"
 >
 	<div class=" absolute text-white z-99999" style="top: {y + 10}px; left: {x + 10}px;">
 		<slot></slot>

src/lib/components/common/RichTextInput.svelte

@@ -801,8 +801,8 @@
 									if (!editor || !editor.view || editor.isDestroyed) {
 										return false;
 									}
-									// default logic
-									return from !== to;
+									// Only show when editor is focused and text is selected
+									return view.hasFocus() && from !== to;
 								}
 							}),
 							FloatingMenu.configure({
@@ -905,6 +905,24 @@
 			},
 			editorProps: {
 				attributes: { id },
+				handleDrop: (view, event) => {
+					// Intercept sidebar chat item drops to prevent ProseMirror
+					// from inserting the raw JSON as text. The actual handling
+					// (adding as Reference Chat) is done by MessageInput's onDrop.
+					const textData = event.dataTransfer?.getData('text/plain');
+					if (textData) {
+						try {
+							const data = JSON.parse(textData);
+							if (data.type === 'chat' && data.id) {
+								// Swallow the drop — let the parent handler deal with it
+								return true;
+							}
+						} catch (_) {
+							// Not JSON, let ProseMirror handle normally
+						}
+					}
+					return false;
+				},
 				handlePaste: (view, event) => {
 					// Force plain-text pasting when richText === false
 					if (!richText) {
@@ -1170,6 +1188,18 @@
 				}
 			},
 			onSelectionUpdate: onSelectionUpdate,
+			onBlur: () => {
+				// Force-hide floating menus when editor loses focus.
+				// shouldShow alone isn't enough because it only runs on transactions.
+				if (bubbleMenuElement) {
+					bubbleMenuElement.style.visibility = 'hidden';
+					bubbleMenuElement.style.opacity = '0';
+				}
+				if (floatingMenuElement) {
+					floatingMenuElement.style.visibility = 'hidden';
+					floatingMenuElement.style.opacity = '0';
+				}
+			},
 			enableInputRules: richText,
 			enablePasteRules: richText
 		});

src/lib/components/layout/ArchivedChatsModal.svelte

@@ -21,6 +21,7 @@
 
 	export let show = false;
 	export let onUpdate = () => {};
+	export let onDelete: (id: string) => void = () => {};
 
 	let loading = false;
 	let chatList: any[] | null = null;
@@ -158,6 +159,9 @@
 	onUpdate={() => {
 		init();
 	}}
+	onDelete={(id) => {
+		onDelete(id);
+	}}
 	loadHandler={loadMoreChats}
 	{unarchiveHandler}
 >

src/lib/components/layout/ChatsModal.svelte

@@ -50,6 +50,7 @@
 	let showDeleteConfirmDialog = false;
 
 	export let onUpdate = () => {};
+	export let onDelete: (id: string) => void = () => {};
 
 	export let loadHandler: null | Function = null;
 	export let unarchiveHandler: null | Function = null;
@@ -69,6 +70,9 @@
 			toast.error(`${error}`);
 		});
 
+		if (res) {
+			onDelete(chatId);
+		}
 		onUpdate();
 	};
 </script>

src/lib/components/layout/SearchModal.svelte

@@ -427,7 +427,7 @@
 			</div>
 			<div
 				id="chat-preview"
-				class="hidden md:flex md:flex-1 w-full overflow-y-auto h-96 md:h-[40rem] scrollbar-hidden"
+				class="hidden md:flex md:flex-1 w-full overflow-y-auto h-96 md:h-[40rem] scrollbar-hidden @container"
 			>
 				{#if messages === null}
 					<div

src/lib/components/layout/Sidebar.svelte

@@ -592,6 +592,12 @@
 	onUpdate={async () => {
 		await initChatList();
 	}}
+	onDelete={(id) => {
+		if ($chatId === id) {
+			chatId.set('');
+			window.history.replaceState({}, '', '/');
+		}
+	}}
 />
 
 <ChannelModal

src/lib/components/layout/Sidebar/ChatItem.svelte

@@ -155,6 +155,12 @@
 	const archiveChatHandler = async (id) => {
 		try {
 			await archiveChatById(localStorage.token, id);
+
+			if ($chatId === id) {
+				await goto('/');
+				chatId.set('');
+			}
+
 			dispatch('change');
 			toast.success($i18n.t('Chat archived.'));
 		} catch (error) {

src/lib/components/layout/Sidebar/ChatMenu.svelte

@@ -384,6 +384,7 @@
 				draggable="false"
 				class="flex gap-2 items-center px-3 py-1.5 text-sm cursor-pointer hover:bg-gray-50 dark:hover:bg-gray-800 rounded-xl w-full"
 				on:click={() => {
+					show = false;
 					cloneChatHandler();
 				}}
 			>

src/lib/components/layout/Sidebar/UserMenu.svelte

@@ -240,8 +240,13 @@
 					href="/playground"
 					draggable="false"
 					class="flex rounded-xl py-1.5 px-3 w-full hover:bg-gray-50 dark:hover:bg-gray-800 transition cursor-pointer select-none"
-					on:click={async () => {
+					on:click={async (e) => {
+						if (e.metaKey || e.ctrlKey || e.shiftKey || e.button === 1) {
+							return;
+						}
+						e.preventDefault();
 						show = false;
+						goto('/playground');
 						if ($mobile) {
 							await tick();
 							showSidebar.set(false);
@@ -257,8 +262,13 @@
 					href="/admin"
 					draggable="false"
 					class="flex rounded-xl py-1.5 px-3 w-full hover:bg-gray-50 dark:hover:bg-gray-800 transition cursor-pointer select-none"
-					on:click={async () => {
+					on:click={async (e) => {
+						if (e.metaKey || e.ctrlKey || e.shiftKey || e.button === 1) {
+							return;
+						}
+						e.preventDefault();
 						show = false;
+						goto('/admin');
 						if ($mobile) {
 							await tick();
 							showSidebar.set(false);

src/lib/components/workspace/Prompts.svelte

@@ -64,6 +64,7 @@
 		loading = true;
 		clearTimeout(searchDebounceTimer);
 		searchDebounceTimer = setTimeout(() => {
+			page = 1;
 			getPromptList();
 		}, 300);
 	}

src/lib/components/workspace/Tools/ToolkitEditor.svelte

@@ -8,6 +8,7 @@
 	import { user } from '$lib/stores';
 	import { updateToolAccessGrants } from '$lib/apis/tools';
 
+	import { nameToId } from '$lib/utils';
 	import CodeEditor from '$lib/components/common/CodeEditor.svelte';
 	import ConfirmDialog from '$lib/components/common/ConfirmDialog.svelte';
 	import ChevronLeft from '$lib/components/icons/ChevronLeft.svelte';
@@ -45,7 +46,7 @@
 	};
 
 	$: if (name && !edit && !clone) {
-		id = name.replace(/\s+/g, '_').toLowerCase();
+		id = nameToId(name);
 	}
 
 	let codeEditor;

src/lib/components/workspace/common/AccessControl.svelte

@@ -12,6 +12,7 @@
 	import Plus from '$lib/components/icons/Plus.svelte';
 	import AddAccessModal from './AddAccessModal.svelte';
 	import Tooltip from '$lib/components/common/Tooltip.svelte';
+	import Switch from '$lib/components/common/Switch.svelte';
 
 	type AccessGrant = {
 		id?: string;
@@ -159,6 +160,12 @@
 				grant.principal_type === 'user' && grant.principal_id === '*' && grant.permission === 'read'
 		);
 
+	const hasPublicWriteGrant = (grants: AccessGrant[]): boolean =>
+		grants.some(
+			(grant) =>
+				grant.principal_type === 'user' && grant.principal_id === '*' && grant.permission === 'write'
+		);
+
 	const currentGrants = (): AccessGrant[] =>
 		Array.isArray(accessGrants) ? (accessGrants as AccessGrant[]) : [];
 
@@ -194,12 +201,12 @@
 	};
 
 	const setPublic = (isPublic: boolean) => {
+		// Remove all user:* grants
 		const filtered = currentGrants().filter(
 			(grant) =>
 				!(
 					grant.principal_type === 'user' &&
-					grant.principal_id === '*' &&
-					grant.permission === 'read'
+					grant.principal_id === '*'
 				)
 		);
 		if (isPublic) {
@@ -212,6 +219,23 @@
 		commitAccessGrants(filtered);
 	};
 
+	const togglePublicWrite = () => {
+		let next = [...currentGrants()];
+		if (hasPublicWriteGrant(next)) {
+			next = next.filter(
+				(grant) =>
+					!(
+						grant.principal_type === 'user' &&
+						grant.principal_id === '*' &&
+						grant.permission === 'write'
+					)
+			);
+		} else {
+			next = upsertPrincipalGrant('user', '*', 'write', next);
+		}
+		commitAccessGrants(next);
+	};
+
 	const upsertPrincipalGrant = (
 		principalType: 'user' | 'group',
 		principalId: string,
@@ -491,6 +515,20 @@
 				</div>
 			</div>
 		</div>
+
+		{#if hasPublicReadGrant(accessGrants ?? []) && accessRoles.includes('write')}
+			<div class="flex w-full justify-between mt-2 ml-0.5">
+				<div class="self-center text-xs">
+					{$i18n.t('Allow public write access')}
+				</div>
+				<Switch
+					state={hasPublicWriteGrant(accessGrants ?? [])}
+					on:change={() => {
+						togglePublicWrite();
+					}}
+				/>
+			</div>
+		{/if}
 	</div>
 
 	{#if share}

src/lib/i18n/locales/languages.json

@@ -15,6 +15,10 @@
 		"code": "ar-BH",
 		"title": "Arabic (Bahrain)"
 	},
+	{
+        "code": "az-AZ",
+        "title": "Azərbaycanca"
+    },
 	{
 		"code": "eu-ES",
 		"title": "Basque (Euskara)"

src/lib/stores/index.ts

@@ -80,6 +80,9 @@ export const banners: Writable<Banner[]> = writable([]);
 export const settings: Writable<Settings> = writable({});
 
 export const audioQueue = writable<AudioQueue | null>(null);
+export const chatRequestQueues: Writable<
+	Record<string, { id: string; prompt: string; files: any[] }[]>
+> = writable({});
 
 export const sidebarWidth = writable(260);
 

src/lib/utils/index.ts

@@ -836,8 +836,12 @@ export const isYoutubeUrl = (url: string) => {
 };
 
 export const removeEmojis = (str: string) => {
-	// Regular expression to match emojis
-	const emojiRegex = /[\uD800-\uDBFF][\uDC00-\uDFFF]|\uD83C[\uDC00-\uDFFF]|\uD83D[\uDC00-\uDE4F]/g;
+	// Use Unicode property escape with the 'v' flag (ES2024) to match all
+	// standardised emoji sequences, including text-presentation emoji + variation
+	// selector (e.g. ❤️, ☀️, ✅), keycap sequences (e.g. 1️⃣), ZWJ families
+	// (e.g. 👨‍👩‍👧‍👦) and flag sequences (e.g. 🏳️‍🌈).
+	// The previous surrogate-pair regex missed the entire BMP emoji category.
+	const emojiRegex = /\p{RGI_Emoji}/gv;
 
 	// Replace emojis with an empty string
 	return str.replace(emojiRegex, '');
@@ -885,13 +889,19 @@ export const removeDetails = (content, types) => {
 			);
 		}
 		return segment;
-	});
+	}).trim();
 };
 
 export const removeAllDetails = (content) => {
+	// First pass: strip <details> blocks on the full string before code-fence
+	// splitting, so blocks whose body contains triple backticks are caught.
+	// (replaceOutsideCode splits on ``` fences, which breaks the <details>
+	// regex when the opening and closing tags land in different segments.)
+	content = content.replace(/<details[^>]*>[\s\S]*?<\/details>/gi, '');
+	// Second pass: catch any remaining blocks that live outside code fences
 	return replaceOutsideCode(content, (segment) => {
 		return segment.replace(/<details[^>]*>.*?<\/details>/gis, '');
-	});
+	}).trim();
 };
 
 export const processDetails = (content) => {
@@ -1401,6 +1411,22 @@ export const slugify = (str: string): string => {
 	);
 };
 
+/**
+ * Convert a display name into a safe, underscore-delimited identifier.
+ * Strips emojis, accents, and any non-alphanumeric characters so the
+ * result is always accepted by backend validation.
+ *
+ * e.g. "My Tool 😄" → "my_tool"
+ */
+export const nameToId = (name: string): string => {
+	return name
+		.normalize('NFD')
+		.replace(/[\u0300-\u036f]/g, '')
+		.replace(/[^\w]+/g, '_')
+		.replace(/^_+|_+$/g, '')
+		.toLowerCase();
+};
+
 export const extractInputVariables = (text: string): Record<string, any> => {
 	const regex = /{{\s*([^|}\s]+)\s*\|\s*([^}]+)\s*}}/g;
 	const regularRegex = /{{\s*([^|}\s]+)\s*}}/g;
@@ -1703,9 +1729,18 @@ export const getCodeBlockContents = (content: string): object => {
 
 	let codeBlocks = [];
 
-	let htmlContent = '';
-	let cssContent = '';
-	let jsContent = '';
+	// Groups of related HTML/CSS/JS blocks. Each HTML block starts a new group;
+	// CSS and JS blocks attach to the current (most recent) group.
+	// This preserves the existing behaviour for "dumb" models that output
+	// separate html/css/js blocks meant to form a single page, while also
+	// allowing multiple distinct HTML blocks to produce separate artifacts.
+	let htmlGroups: Array<{ html: string; css: string; js: string }> = [];
+
+	const initDefaultGroup = () => {
+		if (htmlGroups.length === 0) {
+			htmlGroups.push({ html: '', css: '', js: '' });
+		}
+	};
 
 	if (codeBlockContents) {
 		codeBlockContents.forEach((block) => {
@@ -1718,11 +1753,14 @@ export const getCodeBlockContents = (content: string): object => {
 			const { lang, code } = block;
 
 			if (lang === 'html') {
-				htmlContent += code + '\n';
+				// Each HTML block starts a new group
+				htmlGroups.push({ html: code + '\n', css: '', js: '' });
 			} else if (lang === 'css') {
-				cssContent += code + '\n';
+				initDefaultGroup();
+				htmlGroups[htmlGroups.length - 1].css += code + '\n';
 			} else if (lang === 'javascript' || lang === 'js') {
-				jsContent += code + '\n';
+				initDefaultGroup();
+				htmlGroups[htmlGroups.length - 1].js += code + '\n';
 			}
 		});
 	} else {
@@ -1737,28 +1775,42 @@ export const getCodeBlockContents = (content: string): object => {
 		if (inlineHtml) {
 			inlineHtml.forEach((block) => {
 				const content = block.replace(/<\/?html>/gi, ''); // Remove <html> tags
-				htmlContent += content + '\n';
+				htmlGroups.push({ html: content + '\n', css: '', js: '' });
 			});
 		}
 		if (inlineCss) {
 			inlineCss.forEach((block) => {
 				const content = block.replace(/<\/?style>/gi, ''); // Remove <style> tags
-				cssContent += content + '\n';
+				initDefaultGroup();
+				htmlGroups[htmlGroups.length - 1].css += content + '\n';
 			});
 		}
 		if (inlineJs) {
 			inlineJs.forEach((block) => {
 				const content = block.replace(/<\/?script>/gi, ''); // Remove <script> tags
-				jsContent += content + '\n';
+				initDefaultGroup();
+				htmlGroups[htmlGroups.length - 1].js += content + '\n';
 			});
 		}
 	}
 
+	// Backward-compatible flat fields (merged from all groups)
+	const htmlContent = htmlGroups.map((g) => g.html).join('');
+	const cssContent = htmlGroups.map((g) => g.css).join('');
+	const jsContent = htmlGroups.map((g) => g.js).join('');
+
 	return {
 		codeBlocks: codeBlocks,
 		html: htmlContent.trim(),
 		css: cssContent.trim(),
-		js: jsContent.trim()
+		js: jsContent.trim(),
+		htmlGroups: htmlGroups
+			.filter((g) => g.html.trim() || g.css.trim() || g.js.trim())
+			.map((g) => ({
+				html: g.html.trim(),
+				css: g.css.trim(),
+				js: g.js.trim()
+			}))
 	};
 };
 export const parseFrontmatter = (content) => {

src/lib/utils/marked/colon-fence-extension.ts

@@ -0,0 +1,56 @@
+/**
+ * Marked extension for colon-fence blocks (:::type ... :::)
+ *
+ * Used by newer OpenAI chat models to wrap semantically distinct content:
+ *   :::writing    – reusable prose (letters, articles, docs)
+ *   :::code_execution – code execution output
+ *   :::search_results  – web search results
+ *
+ * The extension is generic and will tokenize any :::<identifier> block.
+ */
+
+function colonFenceTokenizer(this: any, src: string) {
+	// Match :::type at the start of a line, optionally followed by content, then closing :::
+	const match = /^:::([\w-]+)\n([\s\S]*?)(?:\n:::(?:\s*$|\n))/m.exec(src);
+	if (match) {
+		const fenceType = match[1];
+		const text = match[2].trim();
+		const raw = match[0];
+
+		const tokens: any[] = [];
+		this.lexer.blockTokens(text, tokens);
+
+		return {
+			type: 'colonFence',
+			raw,
+			fenceType,
+			text,
+			tokens
+		};
+	}
+}
+
+function colonFenceStart(src: string) {
+	const idx = src.match(/^:::\w/m);
+	return idx ? idx.index! : -1;
+}
+
+function colonFenceRenderer(token: any) {
+	return `<div class="colon-fence colon-fence-${token.fenceType}">${token.text}</div>`;
+}
+
+function colonFenceExtension() {
+	return {
+		name: 'colonFence',
+		level: 'block' as const,
+		start: colonFenceStart,
+		tokenizer: colonFenceTokenizer,
+		renderer: colonFenceRenderer
+	};
+}
+
+export default function (options = {}) {
+	return {
+		extensions: [colonFenceExtension()]
+	};
+}

src/routes/(app)/+layout.svelte

@@ -34,6 +34,7 @@
 		temporaryChatEnabled,
 		toolServers,
 		terminalServers,
+		selectedTerminalId,
 		showSearch,
 		showSidebar,
 		showControls,
@@ -350,6 +351,16 @@
 			localStorage.showControls = value ? 'true' : 'false';
 		});
 
+		// Persist selectedTerminalId across page loads
+		selectedTerminalId.set(localStorage.selectedTerminalId ?? null);
+		selectedTerminalId.subscribe((value) => {
+			if (value === null) {
+				delete localStorage.selectedTerminalId;
+			} else {
+				localStorage.selectedTerminalId = value;
+			}
+		});
+
 		await tick();
 
 		loaded = true;