Merge pull request open-webui#12373 from open-webui/dev

0.6.1

Author: tjbck

CHANGELOG.md

@@ -5,6 +5,29 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [0.6.1] - 2025-04-05
+
+### Added
+
+- 🛠️ **Global Tool Servers Configuration**: Admins can now centrally configure global external tool servers from Admin Settings > Tools, allowing seamless sharing of tool integrations across all users without manual setup per user.
+- 🔐 **Direct Tool Usage Permission for Users**: Introduced a new user-level permission toggle that grants non-admin users access to direct external tools, empowering broader team collaboration while maintaining control.
+- 🧠 **Mistral OCR Content Extraction Support**: Added native support for Mistral OCR as a high-accuracy document loader, drastically improving text extraction from scanned documents in RAG workflows.
+- 🖼️ **Tools Indicator UI Redesign**: Enhanced message input now smartly displays both built-in and external tools via a unified dropdown, making it simpler and more intuitive to activate tools during conversations.
+- 📄 **RAG Prompt Improved and More Coherent**: Default RAG system prompt has been revised to be more clear and citation-focused—admins can leave the template field empty to use this new gold-standard prompt.
+- 🧰 **Performance & Developer Improvements**: Major internal restructuring of several tool-related components, simplifying styling and merging external/internal handling logic, resulting in better maintainability and performance.
+- 🌍 **Improved Translations**: Updated translations for Tibetan, Polish, Chinese (Simplified & Traditional), Arabic, Russian, Ukrainian, Dutch, Finnish, and French to improve clarity and consistency across the interface.
+
+### Fixed
+
+- 🔑 **External Tool Server API Key Bug Resolved**: Fixed a critical issue where authentication headers were not being sent when calling tools from external OpenAPI tool servers, ensuring full security and smooth tool operations.
+- 🚫 **Conditional Export Button Visibility**: UI now gracefully hides export buttons when there's nothing to export in models, prompts, tools, or functions, improving visual clarity and reducing confusion.
+- 🧪 **Hybrid Search Failure Recovery**: Resolved edge case in parallel hybrid search where empty or unindexed collections caused backend crashes—these are now cleanly skipped to ensure system stability.
+- 📂 **Admin Folder Deletion Fix**: Addressed an issue where folders created in the admin workspace couldn't be deleted, restoring full organizational flexibility for admins.
+- 🔐 **Improved Generic Error Feedback on Login**: Authentication errors now show simplified, non-revealing messages for privacy and improved UX, especially with federated logins.
+- 📝 **Tool Message with Images Improved**: Enhanced how tool-generated messages with image outputs are shown in chat, making them more readable and consistent with the overall UI design.
+- ⚙️ **Auto-Exclusion for Broken RAG Collections**: Auto-skips document collections that fail to fetch data or return "None", preventing silent errors and streamlining retrieval workflows.
+- 📝 **Docling Text File Handling Fix**: Fixed file parsing inconsistency that broke docling-based RAG functionality for certain plain text files, ensuring wider file compatibility.
+
 ## [0.6.0] - 2025-03-31
 
 ### Added

backend/open_webui/config.py

@@ -331,12 +331,14 @@ def __getattr__(self, key):
 # OAuth config
 ####################################
 
+
 ENABLE_OAUTH_SIGNUP = PersistentConfig(
     "ENABLE_OAUTH_SIGNUP",
     "oauth.enable_signup",
     os.environ.get("ENABLE_OAUTH_SIGNUP", "False").lower() == "true",
 )
 
+
 OAUTH_MERGE_ACCOUNTS_BY_EMAIL = PersistentConfig(
     "OAUTH_MERGE_ACCOUNTS_BY_EMAIL",
     "oauth.merge_accounts_by_email",
@@ -466,6 +468,7 @@ def __getattr__(self, key):
     os.environ.get("OAUTH_USERNAME_CLAIM", "name"),
 )
 
+
 OAUTH_PICTURE_CLAIM = PersistentConfig(
     "OAUTH_PICTURE_CLAIM",
     "oauth.oidc.avatar_claim",
@@ -878,6 +881,17 @@ def oidc_oauth_register(client):
     pass
 OPENAI_API_BASE_URL = "https://api.openai.com/v1"
 
+####################################
+# TOOL_SERVERS
+####################################
+
+
+TOOL_SERVER_CONNECTIONS = PersistentConfig(
+    "TOOL_SERVER_CONNECTIONS",
+    "tool_server.connections",
+    [],
+)
+
 ####################################
 # WEBUI
 ####################################
@@ -1034,6 +1048,11 @@ def oidc_oauth_register(client):
     == "true"
 )
 
+USER_PERMISSIONS_FEATURES_DIRECT_TOOL_SERVERS = (
+    os.environ.get("USER_PERMISSIONS_FEATURES_DIRECT_TOOL_SERVERS", "False").lower()
+    == "true"
+)
+
 USER_PERMISSIONS_FEATURES_WEB_SEARCH = (
     os.environ.get("USER_PERMISSIONS_FEATURES_WEB_SEARCH", "True").lower() == "true"
 )
@@ -1071,6 +1090,7 @@ def oidc_oauth_register(client):
         "temporary_enforced": USER_PERMISSIONS_CHAT_TEMPORARY_ENFORCED,
     },
     "features": {
+        "direct_tool_servers": USER_PERMISSIONS_FEATURES_DIRECT_TOOL_SERVERS,
         "web_search": USER_PERMISSIONS_FEATURES_WEB_SEARCH,
         "image_generation": USER_PERMISSIONS_FEATURES_IMAGE_GENERATION,
         "code_interpreter": USER_PERMISSIONS_FEATURES_CODE_INTERPRETER,
@@ -1727,6 +1747,11 @@ class BannerModel(BaseModel):
     os.getenv("DOCUMENT_INTELLIGENCE_KEY", ""),
 )
 
+MISTRAL_OCR_API_KEY = PersistentConfig(
+    "MISTRAL_OCR_API_KEY",
+    "rag.mistral_ocr_api_key",
+    os.getenv("MISTRAL_OCR_API_KEY", ""),
+)
 
 BYPASS_EMBEDDING_AND_RETRIEVAL = PersistentConfig(
     "BYPASS_EMBEDDING_AND_RETRIEVAL",
@@ -1875,26 +1900,25 @@ class BannerModel(BaseModel):
 )
 
 DEFAULT_RAG_TEMPLATE = """### Task:
-Respond to the user query using the provided context, incorporating inline citations in the format [source_id] **only when the <source_id> tag is explicitly provided** in the context.
+Respond to the user query using the provided context, incorporating inline citations in the format [id] **only when the <source> tag includes an explicit id attribute** (e.g., <source id="1">).
 
 ### Guidelines:
 - If you don't know the answer, clearly state that.
 - If uncertain, ask the user for clarification.
 - Respond in the same language as the user's query.
 - If the context is unreadable or of poor quality, inform the user and provide the best possible answer.
 - If the answer isn't present in the context but you possess the knowledge, explain this to the user and provide the answer using your own understanding.
-- **Only include inline citations using [source_id] (e.g., [1], [2]) when a `<source_id>` tag is explicitly provided in the context.**
-- Do not cite if the <source_id> tag is not provided in the context.  
+- **Only include inline citations using [id] (e.g., [1], [2]) when the <source> tag includes an id attribute.**
+- Do not cite if the <source> tag does not contain an id attribute.
 - Do not use XML tags in your response.
 - Ensure citations are concise and directly related to the information provided.
 
 ### Example of Citation:
-If the user asks about a specific topic and the information is found in "whitepaper.pdf" with a provided <source_id>, the response should include the citation like so:  
-* "According to the study, the proposed method increases efficiency by 20% [whitepaper.pdf]."
-If no <source_id> is present, the response should omit the citation.
+If the user asks about a specific topic and the information is found in a source with a provided id attribute, the response should include the citation like in the following example:
+* "According to the study, the proposed method increases efficiency by 20% [1]."
 
 ### Output:
-Provide a clear and direct response to the user's query, including inline citations in the format [source_id] only when the <source_id> tag is present in the context.
+Provide a clear and direct response to the user's query, including inline citations in the format [id] only when the <source> tag with id attribute is present in the context.
 
 <context>
 {{CONTEXT}}

backend/open_webui/main.py

@@ -105,6 +105,8 @@
     OPENAI_API_CONFIGS,
     # Direct Connections
     ENABLE_DIRECT_CONNECTIONS,
+    # Tool Server Configs
+    TOOL_SERVER_CONNECTIONS,
     # Code Execution
     ENABLE_CODE_EXECUTION,
     CODE_EXECUTION_ENGINE,
@@ -191,6 +193,7 @@
     DOCLING_SERVER_URL,
     DOCUMENT_INTELLIGENCE_ENDPOINT,
     DOCUMENT_INTELLIGENCE_KEY,
+    MISTRAL_OCR_API_KEY,
     RAG_TOP_K,
     RAG_TOP_K_RERANKER,
     RAG_TEXT_SPLITTER,
@@ -355,6 +358,7 @@
 
 from open_webui.utils.auth import (
     get_license_data,
+    get_http_authorization_cred,
     decode_token,
     get_admin_user,
     get_verified_user,
@@ -477,6 +481,15 @@ async def lifespan(app: FastAPI):
 
 app.state.OPENAI_MODELS = {}
 
+########################################
+#
+# TOOL SERVERS
+#
+########################################
+
+app.state.config.TOOL_SERVER_CONNECTIONS = TOOL_SERVER_CONNECTIONS
+app.state.TOOL_SERVERS = []
+
 ########################################
 #
 # DIRECT CONNECTIONS
@@ -582,6 +595,7 @@ async def lifespan(app: FastAPI):
 app.state.config.DOCLING_SERVER_URL = DOCLING_SERVER_URL
 app.state.config.DOCUMENT_INTELLIGENCE_ENDPOINT = DOCUMENT_INTELLIGENCE_ENDPOINT
 app.state.config.DOCUMENT_INTELLIGENCE_KEY = DOCUMENT_INTELLIGENCE_KEY
+app.state.config.MISTRAL_OCR_API_KEY = MISTRAL_OCR_API_KEY
 
 app.state.config.TEXT_SPLITTER = RAG_TEXT_SPLITTER
 app.state.config.TIKTOKEN_ENCODING_NAME = TIKTOKEN_ENCODING_NAME
@@ -862,6 +876,10 @@ async def commit_session_after_request(request: Request, call_next):
 @app.middleware("http")
 async def check_url(request: Request, call_next):
     start_time = int(time.time())
+    request.state.token = get_http_authorization_cred(
+        request.headers.get("Authorization")
+    )
+
     request.state.enable_api_key = app.state.config.ENABLE_API_KEY
     response = await call_next(request)
     process_time = int(time.time()) - start_time

backend/open_webui/retrieval/loaders/main.py

@@ -20,6 +20,9 @@
     YoutubeLoader,
 )
 from langchain_core.documents import Document
+
+from open_webui.retrieval.loaders.mistral import MistralLoader
+
 from open_webui.env import SRC_LOG_LEVELS, GLOBAL_LOG_LEVEL
 
 logging.basicConfig(stream=sys.stdout, level=GLOBAL_LOG_LEVEL)
@@ -181,13 +184,16 @@ def load(
             for doc in docs
         ]
 
+    def _is_text_file(self, file_ext: str, file_content_type: str) -> bool:
+        return file_ext in known_source_ext or (
+            file_content_type and file_content_type.find("text/") >= 0
+        )
+
     def _get_loader(self, filename: str, file_content_type: str, file_path: str):
         file_ext = filename.split(".")[-1].lower()
 
         if self.engine == "tika" and self.kwargs.get("TIKA_SERVER_URL"):
-            if file_ext in known_source_ext or (
-                file_content_type and file_content_type.find("text/") >= 0
-            ):
+            if self._is_text_file(file_ext, file_content_type):
                 loader = TextLoader(file_path, autodetect_encoding=True)
             else:
                 loader = TikaLoader(
@@ -196,11 +202,14 @@ def _get_loader(self, filename: str, file_content_type: str, file_path: str):
                     mime_type=file_content_type,
                 )
         elif self.engine == "docling" and self.kwargs.get("DOCLING_SERVER_URL"):
-            loader = DoclingLoader(
-                url=self.kwargs.get("DOCLING_SERVER_URL"),
-                file_path=file_path,
-                mime_type=file_content_type,
-            )
+            if self._is_text_file(file_ext, file_content_type):
+                loader = TextLoader(file_path, autodetect_encoding=True)
+            else:
+                loader = DoclingLoader(
+                    url=self.kwargs.get("DOCLING_SERVER_URL"),
+                    file_path=file_path,
+                    mime_type=file_content_type,
+                )
         elif (
             self.engine == "document_intelligence"
             and self.kwargs.get("DOCUMENT_INTELLIGENCE_ENDPOINT") != ""
@@ -222,6 +231,15 @@ def _get_loader(self, filename: str, file_content_type: str, file_path: str):
                 api_endpoint=self.kwargs.get("DOCUMENT_INTELLIGENCE_ENDPOINT"),
                 api_key=self.kwargs.get("DOCUMENT_INTELLIGENCE_KEY"),
             )
+        elif (
+            self.engine == "mistral_ocr"
+            and self.kwargs.get("MISTRAL_OCR_API_KEY") != ""
+            and file_ext
+            in ["pdf"]  # Mistral OCR currently only supports PDF and images
+        ):
+            loader = MistralLoader(
+                api_key=self.kwargs.get("MISTRAL_OCR_API_KEY"), file_path=file_path
+            )
         else:
             if file_ext == "pdf":
                 loader = PyPDFLoader(
@@ -257,9 +275,7 @@ def _get_loader(self, filename: str, file_content_type: str, file_path: str):
                 loader = UnstructuredPowerPointLoader(file_path)
             elif file_ext == "msg":
                 loader = OutlookMessageLoader(file_path)
-            elif file_ext in known_source_ext or (
-                file_content_type and file_content_type.find("text/") >= 0
-            ):
+            elif self._is_text_file(file_ext, file_content_type):
                 loader = TextLoader(file_path, autodetect_encoding=True)
             else:
                 loader = TextLoader(file_path, autodetect_encoding=True)