merged conflicts

0d127ffe · Jannik Streidl · f5afb0ca · f7e10e1f · 0d127ffe · 0d127ffe
Commit 0d127ffe authored Mar 07, 2024 by Jannik Streidl
20 changed files
--- a/.dockerignore
+++ b/.dockerignore
@@ -7,7 +7,6 @@ node_modules
 /package
 .env
 .env.*
-!.env.example
 vite.config.js.timestamp-*
 vite.config.ts.timestamp-*
 __pycache__

--- a/.env.example
+++ b/.env.example
 # Ollama URL for the backend to connect
-# The path '/ollama/api' will be redirected to the specified backend URL
+# The path '/ollama' will be redirected to the specified backend URL
-OLLAMA_API_BASE_URL='http://localhost:11434/api'
+OLLAMA_BASE_URL='http://localhost:11434'
 OPENAI_API_BASE_URL=''
 OPENAI_API_KEY=''

--- a/.github/workflows/build-release.yml
+++ b/.github/workflows/build-release.yml
@@ -19,24 +19,34 @@ jobs:
          echo "No changes to package.json"
          exit 1
        }
    - name: Get version number from package.json
      id: get_version
      run: |
        VERSION=$(jq -r '.version' package.json)
        echo "::set-output name=version::$VERSION"
+    - name: Extract latest CHANGELOG entry
+      id: changelog
+      run: |
+        CHANGELOG_CONTENT=$(awk '/^## \[/{n++} n==1' CHANGELOG.md)
+        echo "CHANGELOG_CONTENT<<EOF" 
+        echo "$CHANGELOG_CONTENT"
+        echo "EOF" 
+        echo "::set-output name=content::${CHANGELOG_CONTENT}"
    - name: Create GitHub release
      uses: actions/github-script@v5
      with:
        github-token: ${{ secrets.GITHUB_TOKEN }}
        script: |
+          const changelog = `${{ steps.changelog.outputs.content }}`;
          const release = await github.rest.repos.createRelease({
            owner: context.repo.owner,
            repo: context.repo.repo,
            tag_name: `v${{ steps.get_version.outputs.version }}`,
            name: `v${{ steps.get_version.outputs.version }}`,
-            body: 'Automatically created new release',
+            body: changelog,
          })
          console.log(`Created release ${release.data.html_url}`)

--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -5,6 +5,38 @@ 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).
+## [0.1.110] - 2024-03-06
+### Added
+- **🌐 Multiple OpenAI Servers Support**: Enjoy seamless integration with multiple OpenAI-compatible APIs, now supported natively.
+### Fixed
+- **🔍 OCR Issue**: Resolved PDF parsing issue caused by OCR malfunction.
+- **🚫 RAG Issue**: Fixed the RAG functionality, ensuring it operates smoothly.
+- **📄 "Add Docs" Model Button**: Addressed the non-functional behavior of the "Add Docs" model button.
+## [0.1.109] - 2024-03-06
+### Added
+- **🔄 Multiple Ollama Servers Support**: Enjoy enhanced scalability and performance with support for multiple Ollama servers in a single WebUI. Load balancing features are now available, providing improved efficiency (#788, #278).
+- **🔧 Support for Claude 3 and Gemini**: Responding to user requests, we've expanded our toolset to include Claude 3 and Gemini, offering a wider range of functionalities within our platform (#1064).
+- **🔍 OCR Functionality for PDF Loader**: We've augmented our PDF loader with Optical Character Recognition (OCR) capabilities. Now, extract text from scanned documents and images within PDFs, broadening the scope of content processing (#1050).
+### Fixed
+- **🛠️ RAG Collection**: Implemented a dynamic mechanism to recreate RAG collections, ensuring users have up-to-date and accurate data (#1031).
+- **📝 User Agent Headers**: Fixed issue of RAG web requests being sent with empty user_agent headers, reducing rejections from certain websites. Realistic headers are now utilized for these requests (#1024).
+- **⏹️ Playground Cancel Functionality**: Introducing a new "Cancel" option for stopping Ollama generation in the Playground, enhancing user control and usability (#1006).
+- **🔤 Typographical Error in 'ASSISTANT' Field**: Corrected a typographical error in the 'ASSISTANT' field within the GGUF model upload template for accuracy and consistency (#1061).
+### Changed
+- **🔄 Refactored Message Deletion Logic**: Streamlined message deletion process for improved efficiency and user experience, simplifying interactions within the platform (#1004).
+- **⚠️ Deprecation of `OLLAMA_API_BASE_URL`**: Deprecated `OLLAMA_API_BASE_URL` environment variable; recommend using `OLLAMA_BASE_URL` instead. Refer to our documentation for further details.
 ## [0.1.108] - 2024-03-02
 ### Added

--- a/Dockerfile
+++ b/Dockerfile
@@ -20,7 +20,7 @@ FROM python:3.11-slim-bookworm as base
 ENV ENV=prod
 ENV PORT ""
-ENV OLLAMA_API_BASE_URL "/ollama/api"
+ENV OLLAMA_BASE_URL "/ollama"
 ENV OPENAI_API_BASE_URL ""
 ENV OPENAI_API_KEY ""
@@ -53,6 +53,8 @@ WORKDIR /app/backend
 # install python dependencies
 COPY ./backend/requirements.txt ./requirements.txt
+RUN apt-get update && apt-get install ffmpeg libsm6 libxext6  -y
 RUN pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu --no-cache-dir
 RUN pip3 install -r requirements.txt --no-cache-dir

--- a/README.md
+++ b/README.md
@@ -95,10 +95,10 @@ Don't forget to explore our sibling project, [Open WebUI Community](https://open
 - **If Ollama is on a Different Server**, use this command:
- To connect to Ollama on another server, change the `OLLAMA_API_BASE_URL` to the server's URL:
+- To connect to Ollama on another server, change the `OLLAMA_BASE_URL` to the server's URL:
  ```bash
-  docker run -d -p 3000:8080 -e OLLAMA_API_BASE_URL=https://example.com/api -v open-webui:/app/backend/data --name open-webui --restart always ghcr.io/open-webui/open-webui:main
+  docker run -d -p 3000:8080 -e OLLAMA_BASE_URL=https://example.com -v open-webui:/app/backend/data --name open-webui --restart always ghcr.io/open-webui/open-webui:main
  ```
 - After installation, you can access Open WebUI at [http://localhost:3000](http://localhost:3000). Enjoy! 😄
@@ -110,7 +110,7 @@ If you're experiencing connection issues, it’s often due to the WebUI docker c
 **Example Docker Command**:
 ```bash
-docker run -d --network=host -v open-webui:/app/backend/data -e OLLAMA_API_BASE_URL=http://127.0.0.1:11434/api --name open-webui --restart always ghcr.io/open-webui/open-webui:main
+docker run -d --network=host -v open-webui:/app/backend/data -e OLLAMA_BASE_URL=http://127.0.0.1:11434 --name open-webui --restart always ghcr.io/open-webui/open-webui:main
 ```
 ### Other Installation Methods

--- a/TROUBLESHOOTING.md
+++ b/TROUBLESHOOTING.md
@@ -4,7 +4,7 @@
 The Open WebUI system is designed to streamline interactions between the client (your browser) and the Ollama API. At the heart of this design is a backend reverse proxy, enhancing security and resolving CORS issues.
- **How it Works**: The Open WebUI is designed to interact with the Ollama API through a specific route. When a request is made from the WebUI to Ollama, it is not directly sent to the Ollama API. Initially, the request is sent to the Open WebUI backend via `/ollama/api` route. From there, the backend is responsible for forwarding the request to the Ollama API. This forwarding is accomplished by using the route specified in the `OLLAMA_API_BASE_URL` environment variable. Therefore, a request made to `/ollama/api` in the WebUI is effectively the same as making a request to `OLLAMA_API_BASE_URL` in the backend. For instance, a request to `/ollama/api/tags` in the WebUI is equivalent to `OLLAMA_API_BASE_URL/tags` in the backend.
+- **How it Works**: The Open WebUI is designed to interact with the Ollama API through a specific route. When a request is made from the WebUI to Ollama, it is not directly sent to the Ollama API. Initially, the request is sent to the Open WebUI backend via `/ollama` route. From there, the backend is responsible for forwarding the request to the Ollama API. This forwarding is accomplished by using the route specified in the `OLLAMA_BASE_URL` environment variable. Therefore, a request made to `/ollama` in the WebUI is effectively the same as making a request to `OLLAMA_BASE_URL` in the backend. For instance, a request to `/ollama/api/tags` in the WebUI is equivalent to `OLLAMA_BASE_URL/api/tags` in the backend.
 - **Security Benefits**: This design prevents direct exposure of the Ollama API to the frontend, safeguarding against potential CORS (Cross-Origin Resource Sharing) issues and unauthorized access. Requiring authentication to access the Ollama API further enhances this security layer.
@@ -15,7 +15,7 @@ If you're experiencing connection issues, it’s often due to the WebUI docker c
 **Example Docker Command**:
 ```bash
-docker run -d --network=host -v open-webui:/app/backend/data -e OLLAMA_API_BASE_URL=http://127.0.0.1:11434/api --name open-webui --restart always ghcr.io/open-webui/open-webui:main
+docker run -d --network=host -v open-webui:/app/backend/data -e OLLAMA_BASE_URL=http://127.0.0.1:11434 --name open-webui --restart always ghcr.io/open-webui/open-webui:main
 ```
 ### General Connection Errors
@@ -25,8 +25,8 @@ docker run -d --network=host -v open-webui:/app/backend/data -e OLLAMA_API_BASE_
 **Troubleshooting Steps**:
 1. **Verify Ollama URL Format**:
-   - When running the Web UI container, ensure the `OLLAMA_API_BASE_URL` is correctly set, including the `/api` suffix. (e.g., `http://192.168.1.1:11434/api` for different host setups).
+   - When running the Web UI container, ensure the `OLLAMA_BASE_URL` is correctly set. (e.g., `http://192.168.1.1:11434` for different host setups).
   - In the Open WebUI, navigate to "Settings" > "General".
-   - Confirm that the Ollama Server URL is correctly set to `[OLLAMA URL]/api` (e.g., `http://localhost:11434/api`), including the `/api` suffix.
+   - Confirm that the Ollama Server URL is correctly set to `[OLLAMA URL]` (e.g., `http://localhost:11434`).
 By following these enhanced troubleshooting steps, connection issues should be effectively resolved. For further assistance or queries, feel free to reach out to us on our community Discord.
--- a/backend/apps/ollama/main.py
+++ b/backend/apps/ollama/main.py
@@ -15,7 +15,7 @@ import asyncio
 from apps.web.models.users import Users
 from constants import ERROR_MESSAGES
 from utils.utils import decode_token, get_current_user, get_admin_user
-from config import OLLAMA_BASE_URL, WEBUI_AUTH
+from config import OLLAMA_BASE_URLS
 from typing import Optional, List, Union
@@ -29,8 +29,7 @@ app.add_middleware(
    allow_headers=["*"],
 )
-app.state.OLLAMA_BASE_URL = OLLAMA_BASE_URL
+app.state.OLLAMA_BASE_URLS = OLLAMA_BASE_URLS
-app.state.OLLAMA_BASE_URLS = [OLLAMA_BASE_URL]
 app.state.MODELS = {}
@@ -223,7 +222,7 @@ async def pull_model(
            r = requests.request(
                method="POST",
                url=f"{url}/api/pull",
-                data=form_data.model_dump_json(exclude_none=True),
+                data=form_data.model_dump_json(exclude_none=True).encode(),
                stream=True,
            )
@@ -295,7 +294,7 @@ async def push_model(
            r = requests.request(
                method="POST",
                url=f"{url}/api/push",
-                data=form_data.model_dump_json(exclude_none=True),
+                data=form_data.model_dump_json(exclude_none=True).encode(),
            )
            r.raise_for_status()
@@ -357,7 +356,7 @@ async def create_model(
            r = requests.request(
                method="POST",
                url=f"{url}/api/create",
-                data=form_data.model_dump_json(exclude_none=True),
+                data=form_data.model_dump_json(exclude_none=True).encode(),
                stream=True,
            )
@@ -420,7 +419,7 @@ async def copy_model(
        r = requests.request(
            method="POST",
            url=f"{url}/api/copy",
-            data=form_data.model_dump_json(exclude_none=True),
+            data=form_data.model_dump_json(exclude_none=True).encode(),
        )
        r.raise_for_status()
@@ -467,7 +466,7 @@ async def delete_model(
        r = requests.request(
            method="DELETE",
            url=f"{url}/api/delete",
-            data=form_data.model_dump_json(exclude_none=True),
+            data=form_data.model_dump_json(exclude_none=True).encode(),
        )
        r.raise_for_status()
@@ -507,7 +506,7 @@ async def show_model_info(form_data: ModelNameForm, user=Depends(get_current_use
        r = requests.request(
            method="POST",
            url=f"{url}/api/show",
-            data=form_data.model_dump_json(exclude_none=True),
+            data=form_data.model_dump_json(exclude_none=True).encode(),
        )
        r.raise_for_status()
@@ -559,7 +558,7 @@ async def generate_embeddings(
        r = requests.request(
            method="POST",
            url=f"{url}/api/embeddings",
-            data=form_data.model_dump_json(exclude_none=True),
+            data=form_data.model_dump_json(exclude_none=True).encode(),
        )
        r.raise_for_status()
@@ -645,7 +644,7 @@ async def generate_completion(
            r = requests.request(
                method="POST",
                url=f"{url}/api/generate",
-                data=form_data.model_dump_json(exclude_none=True),
+                data=form_data.model_dump_json(exclude_none=True).encode(),
                stream=True,
            )
@@ -715,7 +714,7 @@ async def generate_chat_completion(
    r = None
-    print(form_data.model_dump_json(exclude_none=True))
+    print(form_data.model_dump_json(exclude_none=True).encode())
    def get_request():
        nonlocal form_data
@@ -745,7 +744,7 @@ async def generate_chat_completion(
            r = requests.request(
                method="POST",
                url=f"{url}/api/chat",
-                data=form_data.model_dump_json(exclude_none=True),
+                data=form_data.model_dump_json(exclude_none=True).encode(),
                stream=True,
            )
@@ -757,6 +756,7 @@ async def generate_chat_completion(
                headers=dict(r.headers),
            )
        except Exception as e:
+            print(e)
            raise e
    try:
@@ -844,7 +844,7 @@ async def generate_openai_chat_completion(
            r = requests.request(
                method="POST",
                url=f"{url}/v1/chat/completions",
-                data=form_data.model_dump_json(exclude_none=True),
+                data=form_data.model_dump_json(exclude_none=True).encode(),
                stream=True,
            )

--- a/backend/apps/openai/main.py
+++ b/backend/apps/openai/main.py
@@ -3,7 +3,10 @@ from fastapi.middleware.cors import CORSMiddleware
 from fastapi.responses import StreamingResponse, JSONResponse, FileResponse
 import requests
+import aiohttp
+import asyncio
 import json
 from pydantic import BaseModel
@@ -15,7 +18,9 @@ from utils.utils import (
    get_verified_user,
    get_admin_user,
 )
-from config import OPENAI_API_BASE_URL, OPENAI_API_KEY, CACHE_DIR
+from config import OPENAI_API_BASE_URLS, OPENAI_API_KEYS, CACHE_DIR
+from typing import List, Optional
 import hashlib
 from pathlib import Path
@@ -29,116 +34,207 @@ app.add_middleware(
    allow_headers=["*"],
 )
-app.state.OPENAI_API_BASE_URL = OPENAI_API_BASE_URL
+app.state.OPENAI_API_BASE_URLS = OPENAI_API_BASE_URLS
-app.state.OPENAI_API_KEY = OPENAI_API_KEY
+app.state.OPENAI_API_KEYS = OPENAI_API_KEYS
+app.state.MODELS = {}
+@app.middleware("http")
+async def check_url(request: Request, call_next):
+    if len(app.state.MODELS) == 0:
+        await get_all_models()
+    else:
+        pass
-class UrlUpdateForm(BaseModel):
+    response = await call_next(request)
-    url: str
+    return response
-class KeyUpdateForm(BaseModel):
+class UrlsUpdateForm(BaseModel):
-    key: str
+    urls: List[str]
-@app.get("/url")
+class KeysUpdateForm(BaseModel):
-async def get_openai_url(user=Depends(get_admin_user)):
+    keys: List[str]
-    return {"OPENAI_API_BASE_URL": app.state.OPENAI_API_BASE_URL}
-@app.post("/url/update")
+@app.get("/urls")
-async def update_openai_url(form_data: UrlUpdateForm, user=Depends(get_admin_user)):
+async def get_openai_urls(user=Depends(get_admin_user)):
-    app.state.OPENAI_API_BASE_URL = form_data.url
+    return {"OPENAI_API_BASE_URLS": app.state.OPENAI_API_BASE_URLS}
-    return {"OPENAI_API_BASE_URL": app.state.OPENAI_API_BASE_URL}
-@app.get("/key")
+@app.post("/urls/update")
-async def get_openai_key(user=Depends(get_admin_user)):
+async def update_openai_urls(form_data: UrlsUpdateForm, user=Depends(get_admin_user)):
-    return {"OPENAI_API_KEY": app.state.OPENAI_API_KEY}
+    app.state.OPENAI_API_BASE_URLS = form_data.urls
+    return {"OPENAI_API_BASE_URLS": app.state.OPENAI_API_BASE_URLS}
-@app.post("/key/update")
+@app.get("/keys")
-async def update_openai_key(form_data: KeyUpdateForm, user=Depends(get_admin_user)):
+async def get_openai_keys(user=Depends(get_admin_user)):
-    app.state.OPENAI_API_KEY = form_data.key
+    return {"OPENAI_API_KEYS": app.state.OPENAI_API_KEYS}
-    return {"OPENAI_API_KEY": app.state.OPENAI_API_KEY}
+@app.post("/keys/update")
+async def update_openai_key(form_data: KeysUpdateForm, user=Depends(get_admin_user)):
+    app.state.OPENAI_API_KEYS = form_data.keys
+    return {"OPENAI_API_KEYS": app.state.OPENAI_API_KEYS}
 @app.post("/audio/speech")
 async def speech(request: Request, user=Depends(get_verified_user)):
-    target_url = f"{app.state.OPENAI_API_BASE_URL}/audio/speech"
+    idx = None
+    try:
+        idx = app.state.OPENAI_API_BASE_URLS.index("https://api.openai.com/v1")
+        body = await request.body()
+        name = hashlib.sha256(body).hexdigest()
+        SPEECH_CACHE_DIR = Path(CACHE_DIR).joinpath("./audio/speech/")
+        SPEECH_CACHE_DIR.mkdir(parents=True, exist_ok=True)
+        file_path = SPEECH_CACHE_DIR.joinpath(f"{name}.mp3")
+        file_body_path = SPEECH_CACHE_DIR.joinpath(f"{name}.json")
+        # Check if the file already exists in the cache
+        if file_path.is_file():
+            return FileResponse(file_path)
+        headers = {}
+        headers["Authorization"] = f"Bearer {app.state.OPENAI_API_KEYS[idx]}"
+        headers["Content-Type"] = "application/json"
+        try:
+            r = requests.post(
+                url=f"{app.state.OPENAI_API_BASE_URLS[idx]}/audio/speech",
+                data=body,
+                headers=headers,
+                stream=True,
+            )
-    if app.state.OPENAI_API_KEY == "":
+            r.raise_for_status()
-        raise HTTPException(status_code=401, detail=ERROR_MESSAGES.API_KEY_NOT_FOUND)
-    body = await request.body()
+            # Save the streaming content to a file
+            with open(file_path, "wb") as f:
+                for chunk in r.iter_content(chunk_size=8192):
+                    f.write(chunk)
-    name = hashlib.sha256(body).hexdigest()
+            with open(file_body_path, "w") as f:
+                json.dump(json.loads(body.decode("utf-8")), f)
-    SPEECH_CACHE_DIR = Path(CACHE_DIR).joinpath("./audio/speech/")
+            # Return the saved file
-    SPEECH_CACHE_DIR.mkdir(parents=True, exist_ok=True)
+            return FileResponse(file_path)
-    file_path = SPEECH_CACHE_DIR.joinpath(f"{name}.mp3")
-    file_body_path = SPEECH_CACHE_DIR.joinpath(f"{name}.json")
-    # Check if the file already exists in the cache
+        except Exception as e:
-    if file_path.is_file():
+            print(e)
-        return FileResponse(file_path)
+            error_detail = "Open WebUI: Server Connection Error"
+            if r is not None:
+                try:
+                    res = r.json()
+                    if "error" in res:
+                        error_detail = f"External: {res['error']}"
+                except:
+                    error_detail = f"External: {e}"
+            raise HTTPException(status_code=r.status_code, detail=error_detail)
+    except ValueError:
+        raise HTTPException(status_code=401, detail=ERROR_MESSAGES.OPENAI_NOT_FOUND)
-    headers = {}
-    headers["Authorization"] = f"Bearer {app.state.OPENAI_API_KEY}"
-    headers["Content-Type"] = "application/json"
+async def fetch_url(url, key):
    try:
-        print("openai")
+        headers = {"Authorization": f"Bearer {key}"}
-        r = requests.post(
+        async with aiohttp.ClientSession() as session:
-            url=target_url,
+            async with session.get(url, headers=headers) as response:
-            data=body,
+                return await response.json()
-            headers=headers,
+    except Exception as e:
-            stream=True,
+        # Handle connection error here
+        print(f"Connection error: {e}")
+        return None
+def merge_models_lists(model_lists):
+    merged_list = []
+    for idx, models in enumerate(model_lists):
+        merged_list.extend(
+            [
+                {**model, "urlIdx": idx}
+                for model in models
+                if "api.openai.com" not in app.state.OPENAI_API_BASE_URLS[idx]
+                or "gpt" in model["id"]
+            ]
        )
-        r.raise_for_status()
+    return merged_list
-        # Save the streaming content to a file
-        with open(file_path, "wb") as f:
-            for chunk in r.iter_content(chunk_size=8192):
-                f.write(chunk)
-        with open(file_body_path, "w") as f:
+async def get_all_models():
-            json.dump(json.loads(body.decode("utf-8")), f)
+    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 the saved file
+    return models
-        return FileResponse(file_path)
-    except Exception as e:
-        print(e)
-        error_detail = "Open WebUI: Server Connection Error"
-        if r is not None:
-            try:
-                res = r.json()
-                if "error" in res:
-                    error_detail = f"External: {res['error']}"
-            except:
-                error_detail = f"External: {e}"
-        raise HTTPException(status_code=r.status_code, detail=error_detail)
+# , user=Depends(get_current_user)
+@app.get("/models")
+@app.get("/models/{url_idx}")
+async def get_models(url_idx: Optional[int] = None):
+    if url_idx == None:
+        return await get_all_models()
+    else:
+        url = app.state.OPENAI_API_BASE_URLS[url_idx]
+        try:
+            r = requests.request(method="GET", url=f"{url}/models")
+            r.raise_for_status()
+            response_data = r.json()
+            if "api.openai.com" in url:
+                response_data["data"] = list(
+                    filter(lambda model: "gpt" in model["id"], response_data["data"])
+                )
+            return response_data
+        except Exception as e:
+            print(e)
+            error_detail = "Open WebUI: Server Connection Error"
+            if r is not None:
+                try:
+                    res = r.json()
+                    if "error" in res:
+                        error_detail = f"External: {res['error']}"
+                except:
+                    error_detail = f"External: {e}"
+            raise HTTPException(
+                status_code=r.status_code if r else 500,
+                detail=error_detail,
+            )
 @app.api_route("/{path:path}", methods=["GET", "POST", "PUT", "DELETE"])
 async def proxy(path: str, request: Request, user=Depends(get_verified_user)):
-    target_url = f"{app.state.OPENAI_API_BASE_URL}/{path}"
+    idx = 0
-    print(target_url, app.state.OPENAI_API_KEY)
-    if app.state.OPENAI_API_KEY == "":
-        raise HTTPException(status_code=401, detail=ERROR_MESSAGES.API_KEY_NOT_FOUND)
    body = await request.body()
    # TODO: Remove below after gpt-4-vision fix from Open AI
    # Try to decode the body of the request from bytes to a UTF-8 string (Require add max_token to fix gpt-4-vision)
    try:
        body = body.decode("utf-8")
        body = json.loads(body)
+        idx = app.state.MODELS[body.get("model")]["urlIdx"]
        # Check if the model is "gpt-4-vision-preview" and set "max_tokens" to 4000
        # This is a workaround until OpenAI fixes the issue with this model
        if body.get("model") == "gpt-4-vision-preview":
@@ -158,8 +254,16 @@ async def proxy(path: str, request: Request, user=Depends(get_verified_user)):
    except json.JSONDecodeError as e:
        print("Error loading request body into a dictionary:", e)
+    url = app.state.OPENAI_API_BASE_URLS[idx]
+    key = app.state.OPENAI_API_KEYS[idx]
+    target_url = f"{url}/{path}"
+    if key == "":
+        raise HTTPException(status_code=401, detail=ERROR_MESSAGES.API_KEY_NOT_FOUND)
    headers = {}
-    headers["Authorization"] = f"Bearer {app.state.OPENAI_API_KEY}"
+    headers["Authorization"] = f"Bearer {key}"
    headers["Content-Type"] = "application/json"
    try:
@@ -181,21 +285,7 @@ async def proxy(path: str, request: Request, user=Depends(get_verified_user)):
                headers=dict(r.headers),
            )
        else:
-            # For non-SSE, read the response and return it
-            # response_data = (
-            #     r.json()
-            #     if r.headers.get("Content-Type", "")
-            #     == "application/json"
-            #     else r.text
-            # )
            response_data = r.json()
-            if "api.openai.com" in app.state.OPENAI_API_BASE_URL and path == "models":
-                response_data["data"] = list(
-                    filter(lambda model: "gpt" in model["id"], response_data["data"])
-                )
            return response_data
    except Exception as e:
        print(e)

--- a/backend/apps/rag/main.py
+++ b/backend/apps/rag/main.py
@@ -425,7 +425,7 @@ def get_loader(filename: str, file_content_type: str, file_path: str):
    ]
    if file_ext == "pdf":
-        loader = PyPDFLoader(file_path)
+        loader = PyPDFLoader(file_path, extract_images=True)
    elif file_ext == "csv":
        loader = CSVLoader(file_path)
    elif file_ext == "rst":

--- a/backend/apps/web/routers/utils.py
+++ b/backend/apps/web/routers/utils.py
@@ -14,7 +14,7 @@ import json
 from utils.utils import get_admin_user
 from utils.misc import calculate_sha256, get_gravatar_url
-from config import OLLAMA_API_BASE_URL, DATA_DIR, UPLOAD_DIR
+from config import OLLAMA_BASE_URLS, DATA_DIR, UPLOAD_DIR
 from constants import ERROR_MESSAGES
@@ -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_API_BASE_URL}/blobs/sha256:{hashed}"
+                    url = f"{OLLAMA_BASE_URLS[0]}/blobs/sha256:{hashed}"
                    response = requests.post(url, data=file)
                    if response.ok:
@@ -147,7 +147,7 @@ def upload(file: UploadFile = File(...)):
                    hashed = calculate_sha256(f)
                    f.seek(0)
-                    url = f"{OLLAMA_API_BASE_URL}/blobs/sha256:{hashed}"
+                    url = f"{OLLAMA_BASE_URLS[0]}/blobs/sha256:{hashed}"
                    response = requests.post(url, data=f)
                    if response.ok:

--- a/backend/config.py
+++ b/backend/config.py
@@ -200,27 +200,32 @@ if not os.path.exists(LITELLM_CONFIG_PATH):
 ####################################
-# OLLAMA_API_BASE_URL
+# OLLAMA_BASE_URL
 ####################################
 OLLAMA_API_BASE_URL = os.environ.get(
    "OLLAMA_API_BASE_URL", "http://localhost:11434/api"
 )
-if ENV == "prod":
+OLLAMA_BASE_URL = os.environ.get("OLLAMA_BASE_URL", "")
-    if OLLAMA_API_BASE_URL == "/ollama/api":
-        OLLAMA_API_BASE_URL = "http://host.docker.internal:11434/api"
+if ENV == "prod":
+    if OLLAMA_BASE_URL == "/ollama":
+        OLLAMA_BASE_URL = "http://host.docker.internal:11434"
-OLLAMA_BASE_URL = os.environ.get("OLLAMA_BASE_URL", "")
-if OLLAMA_BASE_URL == "":
+if OLLAMA_BASE_URL == "" and OLLAMA_API_BASE_URL != "":
    OLLAMA_BASE_URL = (
        OLLAMA_API_BASE_URL[:-4]
        if OLLAMA_API_BASE_URL.endswith("/api")
        else OLLAMA_API_BASE_URL
    )
+OLLAMA_BASE_URLS = os.environ.get("OLLAMA_BASE_URLS", "")
+OLLAMA_BASE_URLS = OLLAMA_BASE_URLS if OLLAMA_BASE_URLS != "" else OLLAMA_BASE_URL
+OLLAMA_BASE_URLS = [url.strip() for url in OLLAMA_BASE_URLS.split(";")]
 ####################################
 # OPENAI_API
@@ -229,9 +234,25 @@ if OLLAMA_BASE_URL == "":
 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"
+OPENAI_API_KEYS = os.environ.get("OPENAI_API_KEYS", "")
+OPENAI_API_KEYS = OPENAI_API_KEYS if OPENAI_API_KEYS != "" else OPENAI_API_KEY
+OPENAI_API_KEYS = [url.strip() for url in OPENAI_API_KEYS.split(";")]
+OPENAI_API_BASE_URLS = os.environ.get("OPENAI_API_BASE_URLS", "")
+OPENAI_API_BASE_URLS = (
+    OPENAI_API_BASE_URLS if OPENAI_API_BASE_URLS != "" else OPENAI_API_BASE_URL
+)
+OPENAI_API_BASE_URLS = [url.strip() for url in OPENAI_API_BASE_URL.split(";")]
 ####################################
 # WEBUI

--- a/backend/constants.py
+++ b/backend/constants.py
@@ -41,6 +41,7 @@ class ERROR_MESSAGES(str, Enum):
    NOT_FOUND = "We could not find what you're looking for :/"
    USER_NOT_FOUND = "We could not find what you're looking for :/"
    API_KEY_NOT_FOUND = "Oops! It looks like there's a hiccup. The API key is missing. Please make sure to provide a valid API key to access this feature."
    MALICIOUS = "Unusual activities detected, please try again in a few minutes."
    PANDOC_NOT_INSTALLED = "Pandoc is not installed on the server. Please contact your administrator for assistance."
@@ -50,3 +51,4 @@ class ERROR_MESSAGES(str, Enum):
    RATE_LIMIT_EXCEEDED = "API rate limit exceeded"
    MODEL_NOT_FOUND = lambda name="": f"Model '{name}' was not found"
+    OPENAI_NOT_FOUND = lambda name="": f"OpenAI API was not found"
--- a/backend/requirements.txt
+++ b/backend/requirements.txt
@@ -35,6 +35,9 @@ openpyxl
 pyxlsb
 xlrd
+opencv-python-headless
+rapidocr-onnxruntime
 faster-whisper
 PyJWT

--- a/docker-compose.yaml
+++ b/docker-compose.yaml
@@ -14,7 +14,7 @@ services:
    build:
      context: .
      args:
-        OLLAMA_API_BASE_URL: '/ollama/api'
+        OLLAMA_BASE_URL: '/ollama'
      dockerfile: Dockerfile
    image: ghcr.io/open-webui/open-webui:main
    container_name: open-webui
@@ -25,7 +25,7 @@ services:
    ports:
      - ${OPEN_WEBUI_PORT-3000}:8080
    environment:
-      - 'OLLAMA_API_BASE_URL=http://ollama:11434/api'
+      - 'OLLAMA_BASE_URL=http://ollama:11434'
      - 'WEBUI_SECRET_KEY='
    extra_hosts:
      - host.docker.internal:host-gateway

--- a/i18next-parser.config.ts
+++ b/i18next-parser.config.ts
+// i18next-parser.config.ts
+import { getLanguages } from './src/lib/i18n/index.ts';
+const getLangCodes = async () => {
+	const languages = await getLanguages();
+	return languages.map((l) => l.code);
+};
+export default {
+	contextSeparator: '_',
+	createOldCatalogs: false,
+	defaultNamespace: 'translation',
+	defaultValue: '',
+	indentation: 2,
+	keepRemoved: false,
+	keySeparator: false,
+	lexers: {
+		svelte: ['JavascriptLexer'],
+		js: ['JavascriptLexer'], // if you're writing jsx inside .js files, change this to JsxLexer
+		ts: ['JavascriptLexer'],
+		default: ['JavascriptLexer']
+	},
+	lineEnding: 'auto',
+	locales: await getLangCodes(),
+	namespaceSeparator: false,
+	output: 'src/lib/i18n/locales/$LOCALE/$NAMESPACE.json',
+	pluralSeparator: '_',
+	input: 'src/**/*.{js,svelte}',
+	sort: true,
+	verbose: true,
+	failOnWarnings: false,
+	failOnUpdate: false,
+	customValueTemplate: null,
+	resetDefaultValueLocale: null,
+	i18nextOptions: null,
+	yamlOptions: null
+};
--- a/kubernetes/helm/templates/webui-deployment.yaml
+++ b/kubernetes/helm/templates/webui-deployment.yaml
@@ -40,7 +40,7 @@ spec:
        - name: data
          mountPath: /app/backend/data
        env:
-        - name: OLLAMA_API_BASE_URL
+        - name: OLLAMA_BASE_URL
          value: {{ include "ollama.url" . | quote }}
        tty: true
      {{- with .Values.webui.nodeSelector }}

--- a/kubernetes/manifest/base/webui-deployment.yaml
+++ b/kubernetes/manifest/base/webui-deployment.yaml
@@ -26,8 +26,8 @@ spec:
            cpu: "1000m"
            memory: "1Gi"
        env:
-        - name: OLLAMA_API_BASE_URL
+        - name: OLLAMA_BASE_URL
-          value: "http://ollama-service.open-webui.svc.cluster.local:11434/api"
+          value: "http://ollama-service.open-webui.svc.cluster.local:11434"
        tty: true
        volumeMounts:
        - name: webui-volume

--- a/package.json
+++ b/package.json
 {
 	"name": "open-webui",
-	"version": "0.1.108",
+	"version": "0.1.110",
 	"private": true,
 	"scripts": {
 		"dev": "vite dev --host",
@@ -27,6 +27,7 @@
 		"eslint": "^8.56.0",
 		"eslint-config-prettier": "^8.5.0",
 		"eslint-plugin-svelte": "^2.30.0",
+		"i18next-parser": "^8.13.0",
 		"postcss": "^8.4.31",
 		"prettier": "^2.8.0",
 		"prettier-plugin-svelte": "^2.10.1",

--- a/src/app.css
+++ b/src/app.css
@@ -43,6 +43,10 @@ ol > li {
 	font-weight: 400;
 }
+li p {
+	display: inline;
+}
 ::-webkit-scrollbar-thumb {
 	--tw-border-opacity: 1;
 	background-color: rgba(217, 217, 227, 0.8);