Merge pull request open-webui#7475 from open-webui/dev

0.4.7

CHANGELOG.md

@@ -5,6 +5,24 @@ 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).
 
+## [0.4.7] - 2024-12-01
+
+### Added
+
+- **✨ Prompt Input Auto-Completion**: Type a prompt and let AI intelligently suggest and complete your inputs. Simply press 'Tab' or swipe right on mobile to confirm. Available only with Rich Text Input (default setting). Disable via Admin Settings for full control.
+- **🌍 Improved Translations**: Enhanced localization for multiple languages, ensuring a more polished and accessible experience for international users.
+
+### Fixed
+
+- **🛠️ Tools Export Issue**: Resolved a critical issue where exporting tools wasn’t functioning, restoring seamless export capabilities.
+- **🔗 Model ID Registration**: Fixed an issue where model IDs weren’t registering correctly in the model editor, ensuring reliable model setup and tracking.
+- **🖋️ Textarea Auto-Expansion**: Corrected a bug where textareas didn’t expand automatically on certain browsers, improving usability for multi-line inputs.
+- **🔧 Ollama Embed Endpoint**: Addressed the /ollama/embed endpoint malfunction, ensuring consistent performance and functionality.
+
+### Changed
+
+- **🎨 Knowledge Base Styling**: Refined knowledge base visuals for a cleaner, more modern look, laying the groundwork for further enhancements in upcoming releases.
+
 ## [0.4.6] - 2024-11-26
 
 ### Added

backend/open_webui/apps/ollama/main.py

@@ -706,7 +706,7 @@ async def generate_embeddings(
     url_idx: Optional[int] = None,
     user=Depends(get_verified_user),
 ):
-    return generate_ollama_batch_embeddings(form_data, url_idx)
+    return await generate_ollama_batch_embeddings(form_data, url_idx)
 
 
 @app.post("/api/embeddings")

backend/open_webui/apps/retrieval/loaders/youtube.py

@@ -1,7 +1,12 @@
+import logging
+
 from typing import Any, Dict, Generator, List, Optional, Sequence, Union
 from urllib.parse import parse_qs, urlparse
 from langchain_core.documents import Document
+from open_webui.env import SRC_LOG_LEVELS
 
+log = logging.getLogger(__name__)
+log.setLevel(SRC_LOG_LEVELS["RAG"])
 
 ALLOWED_SCHEMES = {"http", "https"}
 ALLOWED_NETLOCS = {
@@ -51,12 +56,14 @@ def __init__(
         self,
         video_id: str,
         language: Union[str, Sequence[str]] = "en",
+        proxy_url: Optional[str] = None,
     ):
         """Initialize with YouTube video ID."""
         _video_id = _parse_video_id(video_id)
         self.video_id = _video_id if _video_id is not None else video_id
         self._metadata = {"source": video_id}
         self.language = language
+        self.proxy_url = proxy_url
         if isinstance(language, str):
             self.language = [language]
         else:
@@ -76,10 +83,22 @@ def load(self) -> List[Document]:
                 "Please install it with `pip install youtube-transcript-api`."
             )
 
+        if self.proxy_url:
+            youtube_proxies = {
+                "http": self.proxy_url,
+                "https": self.proxy_url,
+            }
+            # Don't log complete URL because it might contain secrets
+            log.debug(f"Using proxy URL: {self.proxy_url[:14]}...")
+        else:
+            youtube_proxies = None
+
         try:
-            transcript_list = YouTubeTranscriptApi.list_transcripts(self.video_id)
+            transcript_list = YouTubeTranscriptApi.list_transcripts(
+                self.video_id, proxies=youtube_proxies
+            )
         except Exception as e:
-            print(e)
+            log.exception("Loading YouTube transcript failed")
             return []
 
         try:

backend/open_webui/apps/retrieval/main.py

@@ -105,6 +105,7 @@
     TIKA_SERVER_URL,
     UPLOAD_DIR,
     YOUTUBE_LOADER_LANGUAGE,
+    YOUTUBE_LOADER_PROXY_URL,
     DEFAULT_LOCALE,
     AppConfig,
 )
@@ -171,6 +172,7 @@
 app.state.config.PDF_EXTRACT_IMAGES = PDF_EXTRACT_IMAGES
 
 app.state.config.YOUTUBE_LOADER_LANGUAGE = YOUTUBE_LOADER_LANGUAGE
+app.state.config.YOUTUBE_LOADER_PROXY_URL = YOUTUBE_LOADER_PROXY_URL
 app.state.YOUTUBE_LOADER_TRANSLATION = None
 
 
@@ -471,6 +473,7 @@ async def get_rag_config(user=Depends(get_admin_user)):
         "youtube": {
             "language": app.state.config.YOUTUBE_LOADER_LANGUAGE,
             "translation": app.state.YOUTUBE_LOADER_TRANSLATION,
+            "proxy_url": app.state.config.YOUTUBE_LOADER_PROXY_URL,
         },
         "web": {
             "web_loader_ssl_verification": app.state.config.ENABLE_RAG_WEB_LOADER_SSL_VERIFICATION,
@@ -518,6 +521,7 @@ class ChunkParamUpdateForm(BaseModel):
 class YoutubeLoaderConfig(BaseModel):
     language: list[str]
     translation: Optional[str] = None
+    proxy_url: str = ""
 
 
 class WebSearchConfig(BaseModel):
@@ -580,6 +584,7 @@ async def update_rag_config(form_data: ConfigUpdateForm, user=Depends(get_admin_
 
     if form_data.youtube is not None:
         app.state.config.YOUTUBE_LOADER_LANGUAGE = form_data.youtube.language
+        app.state.config.YOUTUBE_LOADER_PROXY_URL = form_data.youtube.proxy_url
         app.state.YOUTUBE_LOADER_TRANSLATION = form_data.youtube.translation
 
     if form_data.web is not None:
@@ -640,6 +645,7 @@ async def update_rag_config(form_data: ConfigUpdateForm, user=Depends(get_admin_
         },
         "youtube": {
             "language": app.state.config.YOUTUBE_LOADER_LANGUAGE,
+            "proxy_url": app.state.config.YOUTUBE_LOADER_PROXY_URL,
             "translation": app.state.YOUTUBE_LOADER_TRANSLATION,
         },
         "web": {
@@ -867,7 +873,7 @@ def save_docs_to_vector_db(
         return True
     except Exception as e:
         log.exception(e)
-        return False
+        raise e
 
 
 class ProcessFileForm(BaseModel):
@@ -897,7 +903,7 @@ def process_file(
 
             docs = [
                 Document(
-                    page_content=form_data.content,
+                    page_content=form_data.content.replace("<br/>", "\n"),
                     metadata={
                         **file.meta,
                         "name": file.filename,
@@ -1081,7 +1087,9 @@ def process_youtube_video(form_data: ProcessUrlForm, user=Depends(get_verified_u
             collection_name = calculate_sha256_string(form_data.url)[:63]
 
         loader = YoutubeLoader(
-            form_data.url, language=app.state.config.YOUTUBE_LOADER_LANGUAGE
+            form_data.url,
+            language=app.state.config.YOUTUBE_LOADER_LANGUAGE,
+            proxy_url=app.state.config.YOUTUBE_LOADER_PROXY_URL,
         )
 
         docs = loader.load()

backend/open_webui/apps/webui/models/tools.py

@@ -76,6 +76,10 @@ class ToolModel(BaseModel):
 ####################
 
 
+class ToolUserModel(ToolModel):
+    user: Optional[UserResponse] = None
+
+
 class ToolResponse(BaseModel):
     id: str
     user_id: str
@@ -138,13 +142,13 @@ def get_tool_by_id(self, id: str) -> Optional[ToolModel]:
         except Exception:
             return None
 
-    def get_tools(self) -> list[ToolUserResponse]:
+    def get_tools(self) -> list[ToolUserModel]:
         with get_db() as db:
             tools = []
             for tool in db.query(Tool).order_by(Tool.updated_at.desc()).all():
                 user = Users.get_user_by_id(tool.user_id)
                 tools.append(
-                    ToolUserResponse.model_validate(
+                    ToolUserModel.model_validate(
                         {
                             **ToolModel.model_validate(tool).model_dump(),
                             "user": user.model_dump() if user else None,
@@ -155,7 +159,7 @@ def get_tools(self) -> list[ToolUserResponse]:
 
     def get_tools_by_user_id(
         self, user_id: str, permission: str = "write"
-    ) -> list[ToolUserResponse]:
+    ) -> list[ToolUserModel]:
         tools = self.get_tools()
 
         return [

backend/open_webui/config.py

@@ -583,6 +583,12 @@ def load_oauth_providers():
 )
 
 OLLAMA_BASE_URL = os.environ.get("OLLAMA_BASE_URL", "")
+if OLLAMA_BASE_URL:
+    # Remove trailing slash
+    OLLAMA_BASE_URL = (
+        OLLAMA_BASE_URL[:-1] if OLLAMA_BASE_URL.endswith("/") else OLLAMA_BASE_URL
+    )
+
 
 K8S_FLAG = os.environ.get("K8S_FLAG", "")
 USE_OLLAMA_DOCKER = os.environ.get("USE_OLLAMA_DOCKER", "false")
@@ -998,6 +1004,66 @@ class BannerModel(BaseModel):
 </chat_history>
 """
 
+ENABLE_AUTOCOMPLETE_GENERATION = PersistentConfig(
+    "ENABLE_AUTOCOMPLETE_GENERATION",
+    "task.autocomplete.enable",
+    os.environ.get("ENABLE_AUTOCOMPLETE_GENERATION", "True").lower() == "true",
+)
+
+AUTOCOMPLETE_GENERATION_INPUT_MAX_LENGTH = PersistentConfig(
+    "AUTOCOMPLETE_GENERATION_INPUT_MAX_LENGTH",
+    "task.autocomplete.input_max_length",
+    int(os.environ.get("AUTOCOMPLETE_GENERATION_INPUT_MAX_LENGTH", "-1")),
+)
+
+AUTOCOMPLETE_GENERATION_PROMPT_TEMPLATE = PersistentConfig(
+    "AUTOCOMPLETE_GENERATION_PROMPT_TEMPLATE",
+    "task.autocomplete.prompt_template",
+    os.environ.get("AUTOCOMPLETE_GENERATION_PROMPT_TEMPLATE", ""),
+)
+
+
+DEFAULT_AUTOCOMPLETE_GENERATION_PROMPT_TEMPLATE = """### Task:
+You are an autocompletion system. Continue the text in `<text>` based on the **completion type** in `<type>` and the given language.  
+
+### **Instructions**:
+1. Analyze `<text>` for context and meaning.  
+2. Use `<type>` to guide your output:  
+   - **General**: Provide a natural, concise continuation.  
+   - **Search Query**: Complete as if generating a realistic search query.  
+3. Start as if you are directly continuing `<text>`. Do **not** repeat, paraphrase, or respond as a model. Simply complete the text.  
+4. Ensure the continuation:
+   - Flows naturally from `<text>`.  
+   - Avoids repetition, overexplaining, or unrelated ideas.  
+5. If unsure, return: `{ "text": "" }`.  
+
+### **Output Rules**:
+- Respond only in JSON format: `{ "text": "<your_completion>" }`.
+
+### **Examples**:
+#### Example 1:  
+Input:  
+<type>General</type>  
+<text>The sun was setting over the horizon, painting the sky</text>  
+Output:  
+{ "text": "with vibrant shades of orange and pink." }
+
+#### Example 2:  
+Input:  
+<type>Search Query</type>  
+<text>Top-rated restaurants in</text>  
+Output:  
+{ "text": "New York City for Italian cuisine." }  
+
+---
+### Context:
+<chat_history>
+{{MESSAGES:END:6}}
+</chat_history>
+<type>{{TYPE}}</type>  
+<text>{{PROMPT}}</text>  
+#### Output:
+"""
 
 TOOLS_FUNCTION_CALLING_PROMPT_TEMPLATE = PersistentConfig(
     "TOOLS_FUNCTION_CALLING_PROMPT_TEMPLATE",
@@ -1259,6 +1325,12 @@ class BannerModel(BaseModel):
     os.getenv("YOUTUBE_LOADER_LANGUAGE", "en").split(","),
 )
 
+YOUTUBE_LOADER_PROXY_URL = PersistentConfig(
+    "YOUTUBE_LOADER_PROXY_URL",
+    "rag.youtube_loader_proxy_url",
+    os.getenv("YOUTUBE_LOADER_PROXY_URL", ""),
+)
+
 
 ENABLE_RAG_WEB_SEARCH = PersistentConfig(
     "ENABLE_RAG_WEB_SEARCH",

backend/open_webui/constants.py

@@ -113,5 +113,6 @@ def __str__(self) -> str:
     TAGS_GENERATION = "tags_generation"
     EMOJI_GENERATION = "emoji_generation"
     QUERY_GENERATION = "query_generation"
+    AUTOCOMPLETE_GENERATION = "autocomplete_generation"
     FUNCTION_CALLING = "function_calling"
     MOA_RESPONSE_GENERATION = "moa_response_generation"

backend/open_webui/env.py

@@ -373,7 +373,7 @@ def parse_section(section):
         AIOHTTP_CLIENT_TIMEOUT = 300
 
 AIOHTTP_CLIENT_TIMEOUT_OPENAI_MODEL_LIST = os.environ.get(
-    "AIOHTTP_CLIENT_TIMEOUT_OPENAI_MODEL_LIST", "3"
+    "AIOHTTP_CLIENT_TIMEOUT_OPENAI_MODEL_LIST", "5"
 )
 
 if AIOHTTP_CLIENT_TIMEOUT_OPENAI_MODEL_LIST == "":
@@ -384,7 +384,7 @@ def parse_section(section):
             AIOHTTP_CLIENT_TIMEOUT_OPENAI_MODEL_LIST
         )
     except Exception:
-        AIOHTTP_CLIENT_TIMEOUT_OPENAI_MODEL_LIST = 3
+        AIOHTTP_CLIENT_TIMEOUT_OPENAI_MODEL_LIST = 5
 
 ####################################
 # OFFLINE_MODE