Merge pull request #1144 from open-webui/dev

0.1.112

CHANGELOG.md

@@ -5,6 +5,14 @@ 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).
 
+## [0.1.112] - 2024-03-15
+
+### Fixed
+
+- 🗨️ Resolved chat malfunction after image generation.
+- 🎨 Fixed various RAG issues.
+- 🧪 Rectified experimental broken GGUF upload logic.
+
 ## [0.1.111] - 2024-03-10
 
 ### Added

backend/apps/images/main.py

@@ -293,6 +293,7 @@ def generate_image(
                 "size": form_data.size if form_data.size else app.state.IMAGE_SIZE,
                 "response_format": "b64_json",
             }
+
             r = requests.post(
                 url=f"https://api.openai.com/v1/images/generations",
                 json=data,
@@ -300,7 +301,6 @@ def generate_image(
             )
 
             r.raise_for_status()
-
             res = r.json()
 
             images = []
@@ -356,7 +356,10 @@ def generate_image(
             return images
 
     except Exception as e:
-        print(e)
-        if r:
-            print(r.json())
-        raise HTTPException(status_code=400, detail=ERROR_MESSAGES.DEFAULT(e))
+        error = e
+
+        if r != None:
+            data = r.json()
+            if "error" in data:
+                error = data["error"]["message"]
+        raise HTTPException(status_code=400, detail=ERROR_MESSAGES.DEFAULT(error))

backend/apps/ollama/main.py

@@ -123,6 +123,7 @@ async def get_all_models():
             map(lambda response: response["models"], responses)
         )
     }
+
     app.state.MODELS = {model["model"]: model for model in models["models"]}
 
     return models
@@ -181,11 +182,17 @@ async def get_ollama_versions(url_idx: Optional[int] = None):
         responses = await asyncio.gather(*tasks)
         responses = list(filter(lambda x: x is not None, responses))
 
-        lowest_version = min(
-            responses, key=lambda x: tuple(map(int, x["version"].split(".")))
-        )
+        if len(responses) > 0:
+            lowest_version = min(
+                responses, key=lambda x: tuple(map(int, x["version"].split(".")))
+            )
 
-        return {"version": lowest_version["version"]}
+            return {"version": lowest_version["version"]}
+        else:
+            raise HTTPException(
+                status_code=500,
+                detail=ERROR_MESSAGES.OLLAMA_NOT_FOUND,
+            )
     else:
         url = app.state.OLLAMA_BASE_URLS[url_idx]
         try:

backend/apps/openai/main.py

@@ -179,20 +179,26 @@ def merge_models_lists(model_lists):
 
 async def get_all_models():
     print("get_all_models")
-    tasks = [
-        fetch_url(f"{url}/models", app.state.OPENAI_API_KEYS[idx])
-        for idx, url in enumerate(app.state.OPENAI_API_BASE_URLS)
-    ]
-    responses = await asyncio.gather(*tasks)
-    responses = list(filter(lambda x: x is not None and "error" not in x, responses))
-    models = {
-        "data": merge_models_lists(
-            list(map(lambda response: response["data"], responses))
-        )
-    }
-    app.state.MODELS = {model["id"]: model for model in models["data"]}
 
-    return models
+    if len(app.state.OPENAI_API_KEYS) == 1 and app.state.OPENAI_API_KEYS[0] == "":
+        models = {"data": []}
+    else:
+        tasks = [
+            fetch_url(f"{url}/models", app.state.OPENAI_API_KEYS[idx])
+            for idx, url in enumerate(app.state.OPENAI_API_BASE_URLS)
+        ]
+        responses = await asyncio.gather(*tasks)
+        responses = list(
+            filter(lambda x: x is not None and "error" not in x, responses)
+        )
+        models = {
+            "data": merge_models_lists(
+                list(map(lambda response: response["data"], responses))
+            )
+        }
+        app.state.MODELS = {model["id"]: model for model in models["data"]}
+
+        return models
 
 
 @app.get("/models")

backend/apps/rag/utils.py

@@ -91,7 +91,92 @@ def query_collection(
 
 
 def rag_template(template: str, context: str, query: str):
-    template = re.sub(r"\[context\]", context, template)
-    template = re.sub(r"\[query\]", query, template)
-
+    template = template.replace("[context]", context)
+    template = template.replace("[query]", query)
     return template
+
+
+def rag_messages(docs, messages, template, k, embedding_function):
+    print(docs)
+
+    last_user_message_idx = None
+    for i in range(len(messages) - 1, -1, -1):
+        if messages[i]["role"] == "user":
+            last_user_message_idx = i
+            break
+
+    user_message = messages[last_user_message_idx]
+
+    if isinstance(user_message["content"], list):
+        # Handle list content input
+        content_type = "list"
+        query = ""
+        for content_item in user_message["content"]:
+            if content_item["type"] == "text":
+                query = content_item["text"]
+                break
+    elif isinstance(user_message["content"], str):
+        # Handle text content input
+        content_type = "text"
+        query = user_message["content"]
+    else:
+        # Fallback in case the input does not match expected types
+        content_type = None
+        query = ""
+
+    relevant_contexts = []
+
+    for doc in docs:
+        context = None
+
+        try:
+            if doc["type"] == "collection":
+                context = query_collection(
+                    collection_names=doc["collection_names"],
+                    query=query,
+                    k=k,
+                    embedding_function=embedding_function,
+                )
+            else:
+                context = query_doc(
+                    collection_name=doc["collection_name"],
+                    query=query,
+                    k=k,
+                    embedding_function=embedding_function,
+                )
+        except Exception as e:
+            print(e)
+            context = None
+
+        relevant_contexts.append(context)
+
+    context_string = ""
+    for context in relevant_contexts:
+        if context:
+            context_string += " ".join(context["documents"][0]) + "\n"
+
+    ra_content = rag_template(
+        template=template,
+        context=context_string,
+        query=query,
+    )
+
+    if content_type == "list":
+        new_content = []
+        for content_item in user_message["content"]:
+            if content_item["type"] == "text":
+                # Update the text item's content with ra_content
+                new_content.append({"type": "text", "text": ra_content})
+            else:
+                # Keep other types of content as they are
+                new_content.append(content_item)
+        new_user_message = {**user_message, "content": new_content}
+    else:
+        new_user_message = {
+            **user_message,
+            "content": ra_content,
+        }
+
+    messages[last_user_message_idx] = new_user_message
+
+    return messages

backend/apps/web/routers/utils.py

@@ -75,7 +75,7 @@ async def download_file_stream(url, file_path, file_name, chunk_size=1024 * 1024
                     hashed = calculate_sha256(file)
                     file.seek(0)
 
-                    url = f"{OLLAMA_BASE_URLS[0]}/blobs/sha256:{hashed}"
+                    url = f"{OLLAMA_BASE_URLS[0]}/api/blobs/sha256:{hashed}"
                     response = requests.post(url, data=file)
 
                     if response.ok:

backend/config.py

@@ -209,10 +209,6 @@ OLLAMA_API_BASE_URL = os.environ.get(
 
 OLLAMA_BASE_URL = os.environ.get("OLLAMA_BASE_URL", "")
 
-if ENV == "prod":
-    if OLLAMA_BASE_URL == "/ollama":
-        OLLAMA_BASE_URL = "http://host.docker.internal:11434"
-
 
 if OLLAMA_BASE_URL == "" and OLLAMA_API_BASE_URL != "":
     OLLAMA_BASE_URL = (
@@ -221,6 +217,11 @@ if OLLAMA_BASE_URL == "" and OLLAMA_API_BASE_URL != "":
         else OLLAMA_API_BASE_URL
     )
 
+if ENV == "prod":
+    if OLLAMA_BASE_URL == "/ollama":
+        OLLAMA_BASE_URL = "http://host.docker.internal:11434"
+
+
 OLLAMA_BASE_URLS = os.environ.get("OLLAMA_BASE_URLS", "")
 OLLAMA_BASE_URLS = OLLAMA_BASE_URLS if OLLAMA_BASE_URLS != "" else OLLAMA_BASE_URL
 
@@ -234,8 +235,6 @@ OLLAMA_BASE_URLS = [url.strip() for url in OLLAMA_BASE_URLS.split(";")]
 OPENAI_API_KEY = os.environ.get("OPENAI_API_KEY", "")
 OPENAI_API_BASE_URL = os.environ.get("OPENAI_API_BASE_URL", "")
 
-if OPENAI_API_KEY == "":
-    OPENAI_API_KEY = "none"
 
 if OPENAI_API_BASE_URL == "":
     OPENAI_API_BASE_URL = "https://api.openai.com/v1"

backend/constants.py

@@ -52,3 +52,4 @@ class ERROR_MESSAGES(str, Enum):
 
     MODEL_NOT_FOUND = lambda name="": f"Model '{name}' was not found"
     OPENAI_NOT_FOUND = lambda name="": f"OpenAI API was not found"
+    OLLAMA_NOT_FOUND = "WebUI could not connect to Ollama"

backend/main.py

@@ -28,7 +28,7 @@ from typing import List
 
 
 from utils.utils import get_admin_user
-from apps.rag.utils import query_doc, query_collection, rag_template
+from apps.rag.utils import rag_messages
 
 from config import (
     WEBUI_NAME,
@@ -60,19 +60,6 @@ app.state.MODEL_FILTER_LIST = MODEL_FILTER_LIST
 
 origins = ["*"]
 
-app.add_middleware(
-    CORSMiddleware,
-    allow_origins=origins,
-    allow_credentials=True,
-    allow_methods=["*"],
-    allow_headers=["*"],
-)
-
-
-@app.on_event("startup")
-async def on_startup():
-    await litellm_app_startup()
-
 
 class RAGMiddleware(BaseHTTPMiddleware):
     async def dispatch(self, request: Request, call_next):
@@ -91,98 +78,33 @@ class RAGMiddleware(BaseHTTPMiddleware):
             # Example: Add a new key-value pair or modify existing ones
             # data["modified"] = True  # Example modification
             if "docs" in data:
-                docs = data["docs"]
-                print(docs)
 
-                last_user_message_idx = None
-                for i in range(len(data["messages"]) - 1, -1, -1):
-                    if data["messages"][i]["role"] == "user":
-                        last_user_message_idx = i
-                        break
-
-                user_message = data["messages"][last_user_message_idx]
-
-                if isinstance(user_message["content"], list):
-                    # Handle list content input
-                    content_type = "list"
-                    query = ""
-                    for content_item in user_message["content"]:
-                        if content_item["type"] == "text":
-                            query = content_item["text"]
-                            break
-                elif isinstance(user_message["content"], str):
-                    # Handle text content input
-                    content_type = "text"
-                    query = user_message["content"]
-                else:
-                    # Fallback in case the input does not match expected types
-                    content_type = None
-                    query = ""
-
-                relevant_contexts = []
-
-                for doc in docs:
-                    context = None
-
-                    try:
-                        if doc["type"] == "collection":
-                            context = query_collection(
-                                collection_names=doc["collection_names"],
-                                query=query,
-                                k=rag_app.state.TOP_K,
-                                embedding_function=rag_app.state.sentence_transformer_ef,
-                            )
-                        else:
-                            context = query_doc(
-                                collection_name=doc["collection_name"],
-                                query=query,
-                                k=rag_app.state.TOP_K,
-                                embedding_function=rag_app.state.sentence_transformer_ef,
-                            )
-                    except Exception as e:
-                        print(e)
-                        context = None
-
-                    relevant_contexts.append(context)
-
-                context_string = ""
-                for context in relevant_contexts:
-                    if context:
-                        context_string += " ".join(context["documents"][0]) + "\n"
-
-                ra_content = rag_template(
-                    template=rag_app.state.RAG_TEMPLATE,
-                    context=context_string,
-                    query=query,
+                data = {**data}
+                data["messages"] = rag_messages(
+                    data["docs"],
+                    data["messages"],
+                    rag_app.state.RAG_TEMPLATE,
+                    rag_app.state.TOP_K,
+                    rag_app.state.sentence_transformer_ef,
                 )
-
-                if content_type == "list":
-                    new_content = []
-                    for content_item in user_message["content"]:
-                        if content_item["type"] == "text":
-                            # Update the text item's content with ra_content
-                            new_content.append({"type": "text", "text": ra_content})
-                        else:
-                            # Keep other types of content as they are
-                            new_content.append(content_item)
-                    new_user_message = {**user_message, "content": new_content}
-                else:
-                    new_user_message = {
-                        **user_message,
-                        "content": ra_content,
-                    }
-
-                data["messages"][last_user_message_idx] = new_user_message
                 del data["docs"]
 
                 print(data["messages"])
 
             modified_body_bytes = json.dumps(data).encode("utf-8")
 
-            # Create a new request with the modified body
-            scope = request.scope
-            scope["body"] = modified_body_bytes
-            request = Request(scope, receive=lambda: self._receive(modified_body_bytes))
+            # Replace the request body with the modified one
+            request._body = modified_body_bytes
+
+            # Set custom header to ensure content-length matches new body length
+            request.headers.__dict__["_list"] = [
+                (b"content-length", str(len(modified_body_bytes)).encode("utf-8")),
+                *[
+                    (k, v)
+                    for k, v in request.headers.raw
+                    if k.lower() != b"content-length"
+                ],
+            ]
 
         response = await call_next(request)
         return response
@@ -194,6 +116,15 @@ class RAGMiddleware(BaseHTTPMiddleware):
 app.add_middleware(RAGMiddleware)
 
 
+app.add_middleware(
+    CORSMiddleware,
+    allow_origins=origins,
+    allow_credentials=True,
+    allow_methods=["*"],
+    allow_headers=["*"],
+)
+
+
 @app.middleware("http")
 async def check_url(request: Request, call_next):
     start_time = int(time.time())
@@ -204,6 +135,11 @@ async def check_url(request: Request, call_next):
     return response
 
 
+@app.on_event("startup")
+async def on_startup():
+    await litellm_app_startup()
+
+
 app.mount("/api/v1", webui_app)
 app.mount("/litellm/api", litellm_app)
 

package.json

@@ -1,6 +1,6 @@
 {
 	"name": "open-webui",
-	"version": "0.1.111",
+	"version": "0.1.112",
 	"private": true,
 	"scripts": {
 		"dev": "vite dev --host",

src/lib/components/chat/Settings/Images.svelte

@@ -116,11 +116,13 @@
 	class="flex flex-col h-full justify-between space-y-3 text-sm"
 	on:submit|preventDefault={async () => {
 		loading = true;
-		await updateOpenAIKey(localStorage.token, OPENAI_API_KEY);
+
+		if (imageGenerationEngine === 'openai') {
+			await updateOpenAIKey(localStorage.token, OPENAI_API_KEY);
+		}
 
 		await updateDefaultImageGenerationModel(localStorage.token, selectedModel);
 
-		await updateDefaultImageGenerationModel(localStorage.token, selectedModel);
 		await updateImageSize(localStorage.token, imageSize).catch((error) => {
 			toast.error(error);
 			return null;

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

@@ -140,7 +140,9 @@
 	};
 
 	const scrollToBottom = () => {
-		messagesContainerElement.scrollTop = messagesContainerElement.scrollHeight;
+		if (messagesContainerElement) {
+			messagesContainerElement.scrollTop = messagesContainerElement.scrollHeight;
+		}
 	};
 
 	//////////////////////////
@@ -308,7 +310,7 @@
 					.map((file) => file.url.slice(file.url.indexOf(',') + 1));
 
 				// Add images array only if it contains elements
-				if (imageUrls && imageUrls.length > 0) {
+				if (imageUrls && imageUrls.length > 0 && message.role === 'user') {
 					baseMessage.images = imageUrls;
 				}
 
@@ -532,7 +534,8 @@
 					.filter((message) => message)
 					.map((message, idx, arr) => ({
 						role: message.role,
-						...(message.files?.filter((file) => file.type === 'image').length > 0 ?? false
+						...((message.files?.filter((file) => file.type === 'image').length > 0 ?? false) &&
+						message.role === 'user'
 							? {
 									content: [
 										{

src/routes/(app)/c/[id]/+page.svelte

@@ -160,7 +160,9 @@
 	};
 
 	const scrollToBottom = () => {
-		messagesContainerElement.scrollTop = messagesContainerElement.scrollHeight;
+		if (messagesContainerElement) {
+			messagesContainerElement.scrollTop = messagesContainerElement.scrollHeight;
+		}
 	};
 
 	//////////////////////////
@@ -321,7 +323,7 @@
 					.map((file) => file.url.slice(file.url.indexOf(',') + 1));
 
 				// Add images array only if it contains elements
-				if (imageUrls && imageUrls.length > 0) {
+				if (imageUrls && imageUrls.length > 0 && message.role === 'user') {
 					baseMessage.images = imageUrls;
 				}
 
@@ -545,7 +547,8 @@
 					.filter((message) => message)
 					.map((message, idx, arr) => ({
 						role: message.role,
-						...(message.files?.filter((file) => file.type === 'image').length > 0 ?? false
+						...((message.files?.filter((file) => file.type === 'image').length > 0 ?? false) &&
+						message.role === 'user'
 							? {
 									content: [
 										{
@@ -688,7 +691,12 @@
 
 		if (messages.length == 2) {
 			window.history.replaceState(history.state, '', `/c/${_chatId}`);
-			await setChatTitle(_chatId, userPrompt);
+
+			if ($settings?.titleAutoGenerateModel) {
+				await generateChatTitle(_chatId, userPrompt);
+			} else {
+				await setChatTitle(_chatId, userPrompt);
+			}
 		}
 	};
 

src/tailwind.css

@@ -3,16 +3,13 @@
 @tailwind utilities;
 
 @layer base {
-	html {
+	html, pre {
 		font-family: -apple-system, 'Arimo', ui-sans-serif, system-ui, 'Segoe UI', Roboto, Ubuntu,
 			Cantarell, 'Noto Sans', sans-serif, 'Helvetica Neue', Arial, 'Apple Color Emoji',
 			'Segoe UI Emoji', 'Segoe UI Symbol', 'Noto Color Emoji';
 	}
 
-	pre {
-		font-family: -apple-system, 'Arimo', ui-sans-serif, system-ui, 'Segoe UI', Roboto, Ubuntu,
-			Cantarell, 'Noto Sans', sans-serif, 'Helvetica Neue', Arial, 'Apple Color Emoji',
-			'Segoe UI Emoji', 'Segoe UI Symbol', 'Noto Color Emoji';
-		white-space: pre-wrap;
-	}
+  pre {
+    white-space: pre-wrap;
+  }
 }