Functions

在OpwenWebUI中显示DeepSeek R1模型的思维链（仅支持0.5.6及以上版本） / Display the reasoning chain of the DeepSeek R1 model in OpwenWebUI(Only supports versions 0.5.6 and above)

Run Python code on open webui

Artifacts V2 inspired by Claude Artifacts. Orignal code is from atgehrhardt.

Parses all HTML, CSS, JavaScript, and SVG code in the LLM output and renders it in real time in your chats.

Artifacts V3 inspired by Claude Artifacts. Orignal code is from atgehrhardt. Updated for openwebui v0.5

Use Deepseek Reasoner & Chat with Realtime Preview of the Thinking Process

OBSOLETE: Please use: https://openwebui.com/f/owndev/azure_ai

A filter that truncates chat history to retain the latest messages while preserving the system prompt for optimal context management.

Run arbitrary Python or Bash code safely in a gVisor sandbox.

The thought process used to enhance LLMs reasoning capabilities. Adapted from https://github.com/richards199999/Thinking-Claude. It supports the soft insertion of your own new system prompt (it is recommended not to be too long), with a maximum of two system prompts coexisting as the constitution of LLM.

Simple formatter that transforms <think> labels into collapsible "💭 Thought for X seconds" panels.

Anthropic API access with all the latest API changes (prompt caching, PDF, vision, etc).

在OpwenWebUI中显示DeepSeek R1模型的思维链，支持官方API和硅基流动的API

Anthropic API access with all the latest API changes (prompt caching, PDF, vision, reasoning etc).

Filter to enhance the prompt given to the llm with good promting practices and grammar.

在OpwenWebUI中显示DeepSeek R1模型的思维链（仅支持0.5.6及以上版本），修复新版本template输出缺少开头think标签问题 / Display the reasoning chain of the DeepSeek R1 model in OpwenWebUI(Only supports versions 0.5.6 and above), Fix the issue of the new version template output missing the starting <think> tag

在OpwenWebUI中显示DeepSeek R1模型的思维链（仅支持0.5.6及以上版本） / Display the reasoning chain of the DeepSeek R1 model in OpwenWebUI(Only supports versions 0.5.6 and above)

DeepSeek-Gemini Vision integrates Deepseek with Gemini for advanced image processing. Gemini handles image analysis, while DeepSeek ensures precise results.

Output formatter for DeepSeek R1 [Deprecated because Open WebUI officially supports think tag as of v0.5.5]

Deepseek v2.5 API both for chat and coding *** NOT SUPPORT IMAGE INPUT ***

Allows users to generate images from text prompts with nano-gpt.com

DeepSeek-Gemini Vision integrates Deepseek with Gemini for advanced image processing. Gemini handles image analysis, while DeepSeek ensures precise results.

DOES NOT WORK WITH OWUI 0.5.X - A streamlined image generation pipeline, handling prompts, periodic status updates, and error management, designed for seamless integration with OpenWebUI's image generation tools.

Automatically recommends tools based on the users prompt.

Manifold for GitHub Models (https://github.com/marketplace/models)

OpenAI o1 Integration for OpenWebUI with developer messages, reasoning control, vision, etc (see "Supports" section in code).

Used to invoke Siliconflow's Flux model, which, when paired with the appropriate prompt, can generate corresponding images.

Unlimited DeepSeek R1 - Easy Install - Feb 3 2025

This function calculates API interaction pricing in Open WebUI and records user model usage for OpenRouter models. Based on maki's edit of bgeneto's cost tracker. See https://github.com/bgeneto/open-webui-cost-tracker for usage.

Connects you to top AI models with smart filtering options for recommended, code-specialized, powerful, and fast models—all through one seamless interface.

Use Deepseek R1 with Realtime Preview of the Thinking Process for Volcengine

https://github.com/Mantonherre/OpenWebUI-Assistant-OpenAI.git Create your assistant in the OpenAI dashboard, generate an API key, and use this code to integrate it into Open WebUI. Ensure that these dependencies are installed in your Python environment before executing the script. pip install --upgrade openai. pip install pydantic, pip install typing-extensions

Anthropic (All 3 including 3.7) API access with all the latest API changes (prompt caching, PDF, vision, etc).

Currently, DeepSeek V3 Pipeline API does not support image file analysis. However, this feature may be added in the future. Please stay tuned for updates from our development team.

A function for evaluating prompt quality using NLP techniques, with language detection for pt-br and en-us.

开箱即用，无配置项，修复 deepseek-r1 深度思考

适配nvidia官方推出的nmi微模型中的r1模型深度思考过程展示。

Multi-API reasoning effort pipeline for deepseek-r1 models with OpenAI/Ollama support. Compatible with open-webui +v0.5.x

DeepSeek Pipeline

Inspired by Claude Artifacts. Original code: atgehrhardt & tokyohouseparty.

In OpenWebUI, displays the thought chain of the DeepSeek R1 model and Bocha searchs (version 0.5.6 or higher)

In OpenWebUI, displays the thought chain of the DeepSeek R1 model and searchs (version 0.5.6 or higher)

支持 DeepSeek API

Fix miss rendering "Thinking" block on SiliconFlow API. <think> block healing, add additional '\n' for middleware processing of Open WebUI. Inspired by zgccrui/deepseek_r1.

在OpwenWebUI中显示DeepSeek R1模型的思维链（仅支持0.5.6及以上版本），修复新版本template输出缺少开头think标签问题，修复联网搜索和生成标题出错的问题，通过添加处理非流式传输时不添加<think>标签来实现。 / Display the reasoning chain of the DeepSeek R1 model in OpwenWebUI(Only supports versions 0.5.6 and above), Fix the issue of the new version template output missing the starting <think> tag, resolve errors in online search and title generation, and achieve this by adding processing to not include the <think> tag during non-streaming transmission.

使用目前最强大的 Perplexity 的api进行联网查询，有着丝滑顺畅的搜索体验，v1.0.4版本增加网页显示的折叠以及最新的 `r1-1776` 和 `sonar-reasoning-pro` 模型，Use the most powerful Perplexity API for online queries, providing a smooth and seamless search experience. Version 1.0.4 adds webpage display folding as well as the latest `r1-1776` and `sonar-reasoning-pro` models.

removes the think tag

If self.toggle is True in the code, prepends /no-think to the system prompt. Otherwise, makes no changes related to /no-think.

Problems -> code, to help solving problems.

在OpwenWebUI中显示DeepSeek R1模型的思维链（仅支持0.5.6及以上版本） / Display the reasoning chain of the DeepSeek R1 model in OpwenWebUI(Only supports versions 0.5.6 and above)

Adapted code from @zgccrui

You can use DeepSeek R1 or QwQ 32B for cheap and fast thinking, and use stronger and more expensive models like Claude 3.7 Sonnet for final summarization output, to achieve a better balance between inference cost and performance.

Manifold for GitHub Models (https://github.com/marketplace/models)

Automatically saves important User or Assistant text as memories, this code is ment to work with dockerized deployments of open webui (v0.5.20), you must have Ollama installed and running local models for this to work. Update the valves with necessary ports, api's if you choose, ect: if you get errors this function has really good debugging so just look at your docker logs.

Unleash precision in coding and problem-solving with the CodeMaster Reasoning Pipe. This advanced manifold excels in both coding tasks and general queries through a powerful three-phase process: initial reasoning breaks down the problem, a configurable chain-of-thought refines solutions with iterative depth, and a final synthesis delivers polished code or concise answers. Optimized for OpenAI and Ollama APIs, it supports multiple reasoning models, adjustable iteration counts, and optional reasoning trace visibility. Perfect for developers and thinkers seeking top-tier AI-driven accuracy and clarity.

The pipeline implementation for interacting with the Google Gemini API. It allows you to easily send prompts, receive responses, and manage your API keys. Crucially, it includes support for API key rotation, enabling you to seamlessly switch between multiple Gemini API keys for improved reliability, rate limiting management, and key security. This is a robust and feature-rich solution for developers building applications that utilize Google's Gemini models.

Route any attached image to describe and use its own prompt to answer

You can use DeepSeek R1 or QwQ 32B for cheap and fast thinking, and use stronger and more expensive models like Claude 3.7 Sonnet for final summarization output, to achieve a better balance between inference cost and performance.

A manifold for reasoning using the Hypothesis to Theories prompting approach, with RL scoring - Updated for 0.5

This code is a Pipeline class that integrates OpenWebUI with Dify. It efficiently manages Dify's conversation_id by mapping the user's user_id and chat_id.