refac

docs/features/chat-features/conversation-organization.md

@@ -12,7 +12,7 @@ Folders allow you to group related conversations together for quick access and b
 - **Creating a Folder**: You can create a new folder to store specific conversations. This is useful if you want to keep conversations of a similar topic or purpose together.
 - **Moving Conversations into Folders**: Conversations can be moved into folders by dragging and dropping them. This allows you to structure your workspace in a way that suits your workflow.
 
-![Folder Demo](/img/folder-demo.gif)
+![Folder Demo](/images/folder-demo.gif)
 
 ### Example Use Case
 
@@ -25,7 +25,7 @@ If you are managing multiple projects, you can create separate folders for each
 Tags provide an additional layer of organization by allowing you to label conversations with keywords or phrases.
 
 - **Adding Tags to Conversations**: Tags can be applied to conversations based on their content or purpose. Tags are flexible and can be added or removed as needed.
-![Tag Demo](/img/tag-demo.gif)
+![Tag Demo](/images/tag-demo.gif)
 - **Using Tags for Searching**: Tags make it easy to locate specific conversations by using the search feature. You can filter conversations by tags to quickly find those related to specific topics.
 
 ### Example Use Case

docs/features/evaluation/index.mdx

@@ -59,11 +59,11 @@ For your feedback to affect the leaderboard, you need what’s called a **siblin
 
 Here’s a sneak peek at how the Arena Model interface works:
 
-![Arena Model Example](/img/evaluation/arena.png)
+![Arena Model Example](/images/evaluation/arena.png)
 
 Need more depth? You can even replicate a [**Chatbot Arena**](https://lmarena.ai/)-style setup!
 
-![Chatbot Arena Example](/img/evaluation/arena-many.png)
+![Chatbot Arena Example](/images/evaluation/arena-many.png)
 
 ### **2. Normal Interaction**
 
@@ -71,11 +71,11 @@ No need to switch to “arena mode” if you don't want to. You can use Open Web
 
 For instance, this is how you can rate during a normal interaction:
 
-![Normal Model Rating Interface](/img/evaluation/normal.png)
+![Normal Model Rating Interface](/images/evaluation/normal.png)
 
 And here's an example of setting up a multi-model comparison, similar to an arena:
 
-![Multi-Model Comparison](/img/evaluation/normal-many.png)
+![Multi-Model Comparison](/images/evaluation/normal-many.png)
 
 ---
 
@@ -85,7 +85,7 @@ After rating, check out the **Leaderboard** under the Admin Panel. This is where
 
 This is a sample leaderboard layout:
 
-![Leaderboard Example](/img/evaluation/leaderboard.png)
+![Leaderboard Example](/images/evaluation/leaderboard.png)
 
 ### Topic-Based Reranking
 
@@ -100,7 +100,7 @@ Don't skip this! Tagging is super powerful because it allows you to **re-rank mo
 
 Here’s an example of how re-ranking looks:
 
-![Reranking Leaderboard by Topic](/img/evaluation/leaderboard-reranked.png)
+![Reranking Leaderboard by Topic](/images/evaluation/leaderboard-reranked.png)
 
 ---
 

docs/features/plugin/functions/action.mdx

@@ -14,7 +14,7 @@ An example of a graph visualization Action can be seen in the video below.
 
 <p align="center">
   <a href="#">
-    <img src="/img/pipelines/graph-viz-action.gif" alt="Graph Visualization Action" />
+    <img src="/images/pipelines/graph-viz-action.gif" alt="Graph Visualization Action" />
   </a>
 </p>
 

docs/getting-started/quick-start/starting-with-ollama.mdx

@@ -27,9 +27,9 @@ To manage your Ollama instance in Open WebUI, follow these steps:
 
 Here’s what the management screen looks like:
 
-![Ollama Management Screen](/img/getting-started/quick-start/manage-ollama.png)
+![Ollama Management Screen](/images/getting-started/quick-start/manage-ollama.png)
 
-![Ollama Management Screen](/img/getting-started/quick-start/manage-modal-ollama.png)
+![Ollama Management Screen](/images/getting-started/quick-start/manage-modal-ollama.png)
 
 
 ## A Quick and Efficient Way to Download Models
@@ -38,7 +38,7 @@ If you’re looking for a faster option to get started, you can download models
 
 Here’s an example of how it works:
 
-![Ollama Download Prompt](/img/getting-started/quick-start/selector-ollama.png)
+![Ollama Download Prompt](/images/getting-started/quick-start/selector-ollama.png)
 
 This method is perfect if you want to skip navigating through the Admin Settings menu and get right to using your models.
 

docs/intro.mdx

@@ -24,7 +24,7 @@ import { SponsorList } from "@site/src/components/SponsorList";
 [![Discord](https://img.shields.io/badge/Discord-Open_WebUI-blue?logo=discord&logoColor=white)](https://discord.gg/5rJgQTnV4s)
 [![](https://img.shields.io/static/v1?label=Sponsor&message=%E2%9D%A4&logo=GitHub&color=%23fe8e86)](https://github.com/sponsors/tjbck)
 
-![Open WebUI Demo](/img/demo.gif)
+![Open WebUI Demo](/images/demo.gif)
 
 
 ## Quick Start with Docker 🐳

docs/pipelines/filters.md

@@ -9,7 +9,7 @@ Filters are used to perform actions against incoming user messages and outgoing
 
 <p align="center">
   <a href="#">
-    <img src="/img/pipelines/filters.png" alt="Filter Workflow" />
+    <img src="/images/pipelines/filters.png" alt="Filter Workflow" />
   </a>
 </p>
 

docs/pipelines/index.mdx

@@ -5,7 +5,7 @@ title: "⚡ Pipelines"
 
 <p align="center">
   <a href="#">
-    <img src="/img/pipelines/header.png" alt="Pipelines Logo" />
+    <img src="/images/pipelines/header.png" alt="Pipelines Logo" />
   </a>
 </p>
 
@@ -37,7 +37,7 @@ Welcome to **Pipelines**, an [Open WebUI](https://github.com/open-webui) initiat
 
 <p align="center">
   <a href="#">
-    <img src="/img/pipelines/workflow.png" alt="Pipelines Workflow" />
+    <img src="/images/pipelines/workflow.png" alt="Pipelines Workflow" />
   </a>
 </p>
 

docs/pipelines/pipes.md

@@ -8,7 +8,7 @@ Pipes are functions that can be used to perform actions prior to returning LLM m
 
 <p align="center">
   <a href="#">
-    <img src="/img/pipelines/pipes.png" alt="Pipe Workflow" />
+    <img src="/images/pipelines/pipes.png" alt="Pipe Workflow" />
   </a>
 </p>
 
@@ -16,6 +16,6 @@ Pipes that are defined in your WebUI show up as a new model with an "External" d
 
 <p align="center">
   <a href="#">
-    <img src="/img/pipelines/pipe-model-example.png" alt="Pipe Models in WebUI" />
+    <img src="/images/pipelines/pipe-model-example.png" alt="Pipe Models in WebUI" />
   </a>
 </p>

docs/tutorials/images.md

@@ -160,7 +160,7 @@ Using Azure OpenAI Dall-E directly is unsupported, but you can [set up a LiteLLM
 
 ## Using Image Generation
 
-![Image Generation Tutorial](/img/tutorial_image_generation.png)
+![Image Generation Tutorial](/images/tutorial_image_generation.png)
 
 1. First, use a text generation model to write a prompt for image generation.
 2. After the response has finished, you can click the Picture icon to generate an image.

docs/tutorials/integrations/deepseekr1-dynamic.md

@@ -0,0 +1,169 @@
+---
+sidebar_position: 1
+title: "🦥 Run DeepSeek R1 Dynamic 1.58-bit with Llama.cpp"
+---
+
+A huge shoutout to **UnslothAI** for their incredible efforts! Thanks to their hard work, we can now run the **full DeepSeek-R1** 671B parameter model in its dynamic 1.58-bit quantized form (compressed to just 131GB) on **Llama.cpp**! And the best part? You no longer have to despair about needing massive enterprise-class GPUs or servers — it’s possible to run this model on your personal machine (albeit slowly for most consumer hardware).  
+
+:::note
+The only true **DeepSeek-R1** model on Ollama is the **671B version** available here: [https://ollama.com/library/deepseek-r1:671b](https://ollama.com/library/deepseek-r1:671b). Other versions are **distilled** models.
+:::
+
+This guide focuses on running the **full DeepSeek-R1 Dynamic 1.58-bit quantized model** using **Llama.cpp** integrated with **Open WebUI**. For this tutorial, we’ll demonstrate the steps with an **M4 Max + 128GB RAM** machine. You can adapt the settings to your own configuration.  
+
+---
+
+## Step 1: Install Llama.cpp  
+
+You can either:  
+- [Download the prebuilt binaries](https://github.com/ggerganov/llama.cpp/releases)  
+- **Or build it yourself**: Follow the instructions here: [Llama.cpp Build Guide](https://github.com/ggerganov/llama.cpp/blob/master/docs/build.md)  
+
+## Step 2: Download the Model Provided by UnslothAI  
+
+Head over to [Unsloth’s Hugging Face page](https://huggingface.co/unsloth/DeepSeek-R1-GGUF) and download the appropriate **dynamic quantized version** of DeepSeek-R1. For this tutorial, we’ll use the **1.58-bit (131GB)** version, which is highly optimized yet remains surprisingly functional.
+
+
+:::tip
+Know your "working directory" — where your Python script or terminal session is running. The model files will download to a subfolder of that directory by default, so be sure you know its path! For example, if you're running the command below in `/Users/yourname/Documents/projects`, your downloaded model will be saved under `/Users/yourname/Documents/projects/DeepSeek-R1-GGUF`. 
+:::
+
+To understand more about UnslothAI’s development process and why these dynamic quantized versions are so efficient, check out their blog post: [UnslothAI DeepSeek R1 Dynamic Quantization](https://unsloth.ai/blog/deepseekr1-dynamic).  
+
+Here’s how to download the model programmatically:  
+```python
+# Install Hugging Face dependencies before running this:
+# pip install huggingface_hub hf_transfer
+
+from huggingface_hub import snapshot_download
+
+snapshot_download(
+    repo_id = "unsloth/DeepSeek-R1-GGUF",  # Specify the Hugging Face repo
+    local_dir = "DeepSeek-R1-GGUF",         # Model will download into this directory
+    allow_patterns = ["*UD-IQ1_S*"],        # Only download the 1.58-bit version
+)
+```
+
+Once the download completes, you’ll find the model files in a directory structure like this:  
+```
+DeepSeek-R1-GGUF/
+├── DeepSeek-R1-UD-IQ1_S/
+│   ├── DeepSeek-R1-UD-IQ1_S-00001-of-00003.gguf
+│   ├── DeepSeek-R1-UD-IQ1_S-00002-of-00003.gguf
+│   ├── DeepSeek-R1-UD-IQ1_S-00003-of-00003.gguf
+```
+
+:::info
+🛠️ Update paths in the later steps to **match your specific directory structure**. For example, if your script was in `/Users/tim/Downloads`, the full path to the GGUF file would be:  
+`/Users/tim/Downloads/DeepSeek-R1-GGUF/DeepSeek-R1-UD-IQ1_S/DeepSeek-R1-UD-IQ1_S-00001-of-00003.gguf`.
+:::
+
+## Step 3: Make Sure Open WebUI is Installed and Running  
+
+If you don’t already have **Open WebUI** installed, no worries! It’s a simple setup. Just follow the [Open WebUI documentation here](https://docs.openwebui.com/). Once installed, start the application — we’ll connect it in a later step to interact with the DeepSeek-R1 model.  
+
+
+## Step 4: Serve the Model Using Llama.cpp  
+
+Now that the model is downloaded, the next step is to run it using **Llama.cpp’s server mode**. Before you begin:  
+
+1. **Locate the `llama-server` binary.**  
+   If you built from source (as outlined in Step 1), the `llama-server` executable will be located in `llama.cpp/build/bin`. Navigate to this directory by using the `cd` command:  
+   ```bash
+   cd [path-to-llama-cpp]/llama.cpp/build/bin
+   ```
+
+   Replace `[path-to-llama-cpp]` with the location where you cloned or built Llama.cpp. For example:  
+   ```bash
+   cd ~/Documents/workspace/llama.cpp/build/bin
+   ```
+
+2. **Point to your model folder.**  
+   Use the full path to the downloaded GGUF files created in Step 2. When serving the model, specify the first part of the split GGUF files (e.g., `DeepSeek-R1-UD-IQ1_S-00001-of-00003.gguf`).  
+
+Here’s the command to start the server:  
+```bash
+./llama-server \
+    --model /[your-directory]/DeepSeek-R1-GGUF/DeepSeek-R1-UD-IQ1_S/DeepSeek-R1-UD-IQ1_S-00001-of-00003.gguf \
+    --port 10000 \
+    --ctx-size 1024 \
+    --n-gpu-layers 40
+```
+
+> 🔑 **Parameters to Customize Based on Your Machine:**  
+> - **`--model`:** Replace `/[your-directory]/` with the path where the GGUF files were downloaded in Step 2.  
+> - **`--port`:** The server default is `8080`, but feel free to change it based on your port availability.  
+> - **`--ctx-size`:** Determines context length (number of tokens). You can increase it if your hardware allows, but be cautious of rising RAM/VRAM usage.  
+> - **`--n-gpu-layers`:** Set the number of layers you want to offload to your GPU for faster inference. The exact number depends on your GPU’s memory capacity — reference Unsloth’s table for specific recommendations. For CPU-only setups, set it to `0`.  
+
+For example, if your model was downloaded to `/Users/tim/Documents/workspace` and you have an RTX 4090 GPU with 24GB VRAM, your command would look like this:  
+```bash
+./llama-server \
+    --model /Users/tim/Documents/workspace/DeepSeek-R1-GGUF/DeepSeek-R1-UD-IQ1_S/DeepSeek-R1-UD-IQ1_S-00001-of-00003.gguf \
+    --port 10000 \
+    --ctx-size 1024 \
+    --n-gpu-layers 40
+```
+
+Once the server starts, it will host a **local OpenAI-compatible API** endpoint at:  
+```
+http://127.0.0.1:10000
+```
+
+:::info
+🖥️ **Llama.cpp Server Running**  
+
+![Server Screenshot](/images/tutorials/deepseek/serve.png)  
+
+After running the command, you should see a message confirming the server is active and listening on port 10000.
+:::
+
+Be sure to **keep this terminal session running**, as it serves the model for all subsequent steps.
+
+## Step 5: Connect Llama.cpp to Open WebUI  
+
+1. Go to **Admin Settings** in Open WebUI.  
+2. Navigate to **Connections > OpenAI Connections.**  
+3. Add the following details for the new connection:  
+   - URL: `http://127.0.0.1:10000/v1`  
+   - API Key: `none`
+
+:::info
+🖥️ **Adding Connection in Open WebUI**  
+
+![Connection Screenshot](/images/tutorials/deepseek/connection.png)  
+
+After running the command, you should see a message confirming the server is active and listening on port 10000.
+:::
+
+Once the connection is saved, you can start querying **DeepSeek-R1** directly from Open WebUI! 🎉  
+
+---
+
+## Example: Generating Responses  
+
+You can now use Open WebUI’s chat interface to interact with the **DeepSeek-R1 Dynamic 1.58-bit model**.  
+
+:::info
+🖥️ **DeepSeek-R1 Response in Open WebUI**  
+
+![Response Screenshot](/images/tutorials/deepseek/response.png)  
+:::
+
+---
+
+## Notes and Considerations  
+
+- **Performance:**  
+  Running a massive 131GB model like DeepSeek-R1 on personal hardware will be **slow**. Even with our M4 Max (128GB RAM), inference speeds were modest. But the fact that it works at all is a testament to UnslothAI’s optimizations.  
+
+- **VRAM/Memory Requirements:**  
+  Ensure sufficient VRAM and system RAM for optimal performance. With low-end GPUs or CPU-only setups, expect slower speeds (but it’s still doable!).  
+
+---
+
+Thanks to **UnslothAI** and **Llama.cpp**, running one of the largest open-source reasoning models, **DeepSeek-R1** (1.58-bit version), is finally accessible to individuals. While it’s challenging to run such models on consumer hardware, the ability to do so without massive computational infrastructure is a significant technological milestone.  
+
+⭐ Big thanks to the community for pushing the boundaries of open AI research.  
+
+Happy experimenting! 🚀  

docs/tutorials/tips/contributing-tutorial.md

@@ -60,7 +60,7 @@ b. **Modify `docusaurus.config.ts` to Use Environment Variables**
      const config: Config = {
        title: "Open WebUI",
        tagline: "ChatGPT-Style WebUI for LLMs (Formerly Ollama WebUI)",
-       favicon: "img/favicon.png",
+       favicon: "images/favicon.png",
        url: process.env.SITE_URL || "https://openwebui.com",
        baseUrl: process.env.BASE_URL || "/",
        ...

docs/tutorials/web_search.md

@@ -348,7 +348,7 @@ docker exec -it open-webui curl http://host.docker.internal:8080/search?q=this+i
 5. Adjust the `Search Result Count` and `Concurrent Requests` values accordingly
 6. Save changes
 
-![SearXNG GUI Configuration](/img/tutorial_searxng_config.png)
+![SearXNG GUI Configuration](/images/tutorial_searxng_config.png)
 
 ## 5. Using Web Search in a Chat
 
@@ -356,7 +356,7 @@ To access Web Search, Click on the + next to the message input field.
 
 Here you can toggle Web Search On/Off.
 
-![Web Search UI Toggle](/img/web_search_toggle.png)
+![Web Search UI Toggle](/images/web_search_toggle.png)
 
 By following these steps, you will have successfully set up SearXNG with Open WebUI, enabling you to perform web searches using the SearXNG engine.
 
@@ -379,14 +379,14 @@ This is enabled on a per session basis eg. reloading the page, changing to anoth
 7. Fill `Google PSE API Key` with the `API key` and `Google PSE Engine Id` (# 4)
 8. Click `Save`
 
-![Open WebUI Admin panel](/img/tutorial_google_pse1.png)
+![Open WebUI Admin panel](/images/tutorial_google_pse1.png)
 
 #### Note
 
 You have to enable `Web search` in the prompt field, using plus (`+`) button.
 Search the web ;-)
 
-![enable Web search](/img/tutorial_google_pse2.png)
+![enable Web search](/images/tutorial_google_pse2.png)
 
 ## Brave API
 
@@ -444,13 +444,13 @@ services:
 6. [Optional] Enter the `SearchApi engine` name you want to query. Example, `google`, `bing`, `baidu`, `google_news`, `bing_news`, `google_videos`, `google_scholar` and `google_patents.` By default, it is set to `google`.
 7. Click `Save`.
 
-![Open WebUI Admin panel](/img/tutorial_searchapi_search.png)
+![Open WebUI Admin panel](/images/tutorial_searchapi_search.png)
 
 #### Note
 
 You have to enable `Web search` in the prompt field, using plus (`+`) button to search the web using [SearchApi](https://www.searchapi.io/) engines.
 
-![enable Web search](/img/enable_web_search.png)
+![enable Web search](/images/enable_web_search.png)
 
 ## Kagi API
 

docusaurus.config.ts

@@ -6,7 +6,7 @@ import { themes as prismThemes } from "prism-react-renderer";
 const config: Config = {
 	title: "Open WebUI",
 	tagline: "ChatGPT-Style WebUI for LLMs (Formerly Ollama WebUI)",
-	favicon: "img/favicon.png",
+	favicon: "images/favicon.png",
 
 	// Set the production url of your site here
 	url: "https://openwebui.com",
@@ -65,12 +65,12 @@ const config: Config = {
 
 	themeConfig: {
 		// Replace with your project's social card
-		// image: "img/docusaurus-social-card.jpg",
+		// image: "images/docusaurus-social-card.jpg",
 		navbar: {
 			title: "Open WebUI",
 			logo: {
-				src: "img/logo.png",
-				srcDark: "img/logo-dark.png",
+				src: "images/logo.png",
+				srcDark: "images/logo-dark.png",
 			},
 			items: [
 				// {
@@ -108,7 +108,7 @@ const config: Config = {
 		},
 		footer: {
 			logo: {
-				src: "img/logo-dark.png",
+				src: "images/logo-dark.png",
 				height: 100,
 			},
 			style: "light",