Feature Overview

Agent

The Agents page is an assistant marketplace where you can select or search for desired model presets. Click on a card to add the assistant to your conversation page's assistant list.

You can also edit and create your own assistants on this page.

• Click My, then Create Agent to start building your own assistant.

Mini Programs

On the Mini Programs page, you can use web versions of AI-related programs from major service providers within the client. Currently, custom addition and removal are not supported.

Knowledge Base

For knowledge base usage, please refer to the Knowledge Base Tutorial in the advanced tutorials.

Files

The Files interface displays all files related to conversations, paintings, knowledge bases, etc. You can centrally manage and view these files on this page.

Settings

Default Model Settings

Default Assistant Model

When the assistant does not have a default assistant model configured, the model selected by default for new conversations is the one set here. The model used for prompt optimization and word-selection assistant is also configured in this section.

Topic Naming Model

After each conversation, a model is called to generate a topic name for the dialog. The model set here is the one used for naming.

Translation Model

The translation feature in input boxes for conversations, painting, etc., and the translation model in the translation interface all use the model set here.

Quick Assistant Model

The model used for the quick assistant feature. For details, see Quick Assistant.

General Settings

On this page, you can configure the software's interface language, proxy settings, and other options.

Shortcut Keys Settings

This interface allows you to enable/disable and configure shortcut keys for certain functions. Please refer to the on-screen instructions for specific setup.

Installation Tutorial

Model Service Configuration

Free Internet Mode

Personalization Settings

Project Planning

To-Do List

• Quick Pop-up (Clipboard reading, quick questions, explanation, translation, summarization)

• Multi-model Response Comparison

• Support SSO Login using provider credentials

• Enable Internet access for all models

• Release first official version

• Plugin functionality (JavaScript)

• Browser extension (Text selection translation/summarization/add to knowledge base)

• iOS & Android client

• AI Notes

• Voice input/output (AI calls)

• Data backup with customizable backup content

Key translation notes:

1. Preserved all Markdown formatting (headings, checkboxes)

2. Technical terms remain unchanged: "JavaScript", "SSO", "iOS", "Android"

3. Action descriptions standardized: "Quick Pop-up" for 快捷弹窗, "multi-model" for 多模型

4. Functional translations: "划词翻译" → "text selection translation"

5. Feature localization: "AI 通话" → "AI calls"

6. Maintained present tense for consistency

7. Preserved special characters and list formatting

8. Translated bracket content while keeping technical references (JavaScript)

9. Proper noun capitalization: "AI Notes"

MCP Usage Tutorial

OneAPI and its Fork Projects

Data Settings

Privacy Policy

Welcome to Cherry Studio (hereinafter referred to as "this software" or "we"). We highly value your privacy protection, and this Privacy Policy explains how we handle and protect your personal information and data. Please read and understand this agreement carefully before using this software:

1. Scope of Information We Collect

To optimize user experience and improve software quality, we may only collect the following anonymous non-personal information: • Software version information; • Function activity and usage frequency; • Anonymous crash and error log information;

The above information is completely anonymous, does not involve any personally identifiable data, and cannot be associated with your personal information.

2. Information We Do Not Collect

To maximize the protection of your privacy, we explicitly commit: • Will not collect, store, transmit, or process the model service API Key information you input into this software; • Will not collect, store, transmit, or process any conversation data generated during your use of this software, including but not limited to chat content, instruction information, knowledge base information, vector data, and other custom content; • Will not collect, store, transmit, or process any personally identifiable sensitive information.

3. Data Interaction Description

This software uses the API Key of third-party model service providers that you apply for and configure yourself to complete model invocation and conversation functions. The model services you use (such as large language models, API interfaces, etc.) are provided and fully managed by the third-party provider you choose, and Cherry Studio only serves as a local tool providing interface invocation functionality with third-party model services.

Therefore: • All conversation data generated between you and the model service is unrelated to Cherry Studio. We neither participate in data storage nor perform any form of data transmission or relay; • You need to independently review and accept the privacy policies and relevant regulations of the corresponding third-party model service providers. These services' privacy policies can be viewed on each provider's official website.

4. Third-Party Model Service Provider Privacy Policy Statement

You shall bear any privacy risks that may arise from using third-party model service providers. For specific privacy policies, data security measures, and related responsibilities, please refer to the relevant content on the official websites of your chosen model service providers. We assume no responsibility for these matters.

5. Agreement Updates and Modifications

This agreement may be appropriately adjusted with software version updates. Please check it periodically. When substantive changes occur to the agreement, we will notify you through appropriate means.

6. Contact Us

If you have any questions regarding this agreement or Cherry Studio's privacy protection measures, please feel free to contact us at any time.

Thank you for choosing and trusting Cherry Studio. We will continue to provide you with a secure and reliable product experience.

Quick Assistant

Quick Assistant is a convenient tool provided by Cherry Studio that allows you to quickly access AI functions in any application, enabling instant operations like asking questions, translation, summarization, and explanations.

Enable Quick Assistant

1. Open Settings: Navigate to Settings -> Shortcuts -> Quick Assistant.

2. Enable Switch: Find and toggle on the Quick Assistant button.

1. Set Shortcut Key (Optional):

   • Default shortcut for Windows: Ctrl + E

   • Default shortcut for macOS: ⌘ + E

   • Customize your shortcut here to avoid conflicts or match your usage habits.

Using Quick Assistant

1. Activate: Press your configured shortcut key (or default shortcut) in any application to open Quick Assistant.

2. Interact: Within the Quick Assistant window, you can directly perform:

   • Quick Questions: Ask any question to the AI.

   • Text Translation: Input text to be translated.

   • Content Summarization: Input long text for summarization.

   • Explanation: Input concepts or terms requiring explanation.
3. Close: Press ESC or click anywhere outside the Quick Assistant window.

Tips & Tricks

• Shortcut Conflicts: Modify shortcuts if defaults conflict with other applications.

• Explore More Functions: Beyond documented features, Quick Assistant may support operations like code generation and style transfer. Continuously explore during usage.

• Feedback & Improvements: Report issues or suggestions to the Cherry Studio team via feedback.

On this page, you can set the software's color theme, page layout, or customize CSS for personalized settings.

Theme Selection

Here you can set the default interface color mode (light mode, dark mode, or follow system).

Topic Settings

These settings are for the layout of the chat interface.

Topic Position

Automatically Switch to Topic

When this setting is enabled, clicking on the assistant's name will automatically switch to the corresponding topic page.

Display Topic Time

When enabled, the creation time of the topic will be displayed below it.

Custom CSS

This setting allows flexible customization of the interface. For specific methods, please refer to Custom CSS in the advanced tutorial.

Open the Official Website

Click download to select the appropriate version

Wait for Download to Complete

If the browser prompts that the file is not trusted, simply choose to keep it.

Choose to Keep → Trust Cherry-Studio

Open the File

Installation

Direct Download

Windows Version

Installer (Setup)

Portable Version (Portable)

macOS Version

Linux Version

Cloud Drive Download

Drawing

The drawing feature currently supports painting models from DMXAPI, TokenFlux, AiHubMix, and . You can register an account at and to use this feature.

For questions about parameters, hover your mouse over the ? icon in corresponding areas to view descriptions.

This page only introduces the interface functions. For configuration tutorials, please refer to the tutorial in the basic tutorials.

API Key

In Cherry Studio, a single service provider supports multi-key round-robin usage, where the keys are rotated from first to last in a list.

• Add multiple keys separated by English commas, as shown in the example below:

API Address

When using built-in service providers, you generally do not need to fill in the API address. If you need to modify it, please strictly follow the address provided in the official documentation.

If the address provided by the service provider is in the format https://xxx.xxx.com/v1/chat/completions, you only need to fill in the root address part (https://xxx.xxx.com).

Cherry Studio will automatically concatenate the remaining path (/v1/chat/completions). Not filling it as required may lead to improper functionality.

Add Model

Typically, clicking the Manage button in the bottom left corner of the service provider configuration page will automatically fetch all models supported by that service provider. You can then click the + sign from the fetched list to add them to the model list.

Connectivity Check

Click the check button next to the API Key input box to test if the configuration is successful.

macOS

1. First, visit the official download page to download the Mac version, or click the direct link below

   Please download the chip-specific version matching your Mac

1. After downloading, click here

1. Drag the icon to install

Find the Cherry Studio icon in Launchpad and click it. If the Cherry Studio main interface opens, the installation is successful.

Alibaba Cloud Bailian

1. Log in to . If you don't have an Alibaba Cloud account, you'll need to register.

2. Click the Create My API-KEY button in the upper-right corner.
3. In the popup window, select the default workspace (or customize it if desired). You can optionally add a description.
4. Click the Confirm button in the lower-right corner.

5. You should now see a new entry in the list. Click the View button on the right.
6. Click the Copy button.
7. Go to Cherry Studio, navigate to Settings → Model Services → Alibaba Cloud Bailian, and paste the copied API key into the API Key field.
8. You can adjust related settings as described in , then start using the service.

OneAPI

• Log in and navigate to the tokens page

• Create a new token (you can also directly use the default token ↑)

• Copy the token

• Open CherryStudio's service provider settings and click Add at the bottom of the provider list

• Enter a note name, select OpenAI as the provider, and click OK

• Paste the key you just copied

• Return to the API Key page, copy the root address from the browser's address bar, for example:

• Add models (click Manage to auto-fetch or enter manually) and toggle the switch in the top right corner to start using.

SiliconFlow

1. Configuring SiliconCloud Model Service

1.2 Click the settings at the bottom left, and select 【SiliconFlow】 in the model service

1.2 Click the link to get SiliconCloud API Key

1. Log in to (if not registered, the first login will automatically register an account)

2. Visit to create a new key or copy an existing one

1.3 Click Manage to add models

2. Using the Model Service

1. Click the "Chat" button in the left menu bar

2. Enter text in the input box to start chatting

3. You can switch models by selecting the model name in the top menu

GitHub Copilot

To use GitHub Copilot, you must first have a GitHub account and subscribe to the GitHub Copilot service. The free subscription tier is acceptable, but note that it does not support the latest Claude 3.7 model. For details, refer to the .

Obtain Device Code

Click "Log in with GitHub" to generate and copy your Device Code.

Enter Device Code in Browser and Authorize

After obtaining your Device Code, click the link to open your browser. Log in to your GitHub account, enter the Device Code, and grant authorization.

After successful authorization, return to Cherry Studio and click "Connect GitHub". Your GitHub username and avatar will appear upon successful connection.

Click "Manage" to Get Model List

Click the "Manage" button below, which will automatically fetch the currently supported models list online.

Common Issues

Failed to Obtain Device Code, Please Retry

The current implementation uses Axios for network requests. Note that Axios does not support SOCKS proxies. Please use a system proxy or HTTP proxy, or avoid setting proxies within CherryStudio and use a global proxy instead. First, ensure your network connection is stable to prevent Device Code retrieval failures.

Clear CSS Settings

• Open the console by clicking on the CherryStudio window and pressing Ctrl+Shift+I (MacOS: command+option+I).

• In the opened console window, click Console

• Then manually enter document.getElementById('user-defined-custom-css').remove(). Copying and pasting will most likely not work.

• After entering, press Enter to confirm and clear the CSS settings. Then, go back to CherryStudio's display settings and remove the problematic CSS code.

Cherry Studio data backup supports backup via S3 compatible storage (object storage). Common S3 compatible storage services include: AWS S3, Cloudflare R2, Alibaba Cloud OSS, Tencent Cloud COS, and MinIO.

Based on S3 compatible storage, multi-terminal data synchronization can be achieved by the method of Computer A S3 Storage Computer B.

Configure S3 Compatible Storage

1. Create an object storage bucket, and record the bucket name. It is highly recommended to set the bucket to private read/write to prevent backup data leakage!!

2. Refer to the documentation, go to the cloud service console to obtain information such as Access Key ID, Secret Access Key, Endpoint, Bucket, Region for S3 compatible storage.

   • Endpoint: The access address for S3 compatible storage, usually in the form of https://<bucket-name>.<region>.amazonaws.com or https://<ACCOUNT_ID>.r2.cloudflarestorage.com.

   • Region: The region where the bucket is located, such as us-west-1, ap-southeast-1, etc. For Cloudflare R2, please fill in auto.

   • Bucket: The bucket name.

   • Access Key ID and Secret Access Key: Credentials used for authentication.

   • Root Path: Optional, specifies the root path when backing up to the bucket, default is empty.

   • Related Documentation

     • AWS S3:

     • Cloudflare R2:

     • Alibaba Cloud OSS:

     • Tencent Cloud COS:

3. Fill in the above information in the S3 backup settings, click the backup button to perform backup, and click the manage button to view and manage the list of backup files.

MCP (Model Context Protocol) is an open-source protocol designed to provide context information to Large Language Models (LLMs) in a standardized way. For more about MCP, see .

Using MCP in Cherry Studio

Below, we'll use the fetch function as an example to demonstrate how to use MCP in Cherry Studio. Details can be found in the .

Preparation: Install uv, bun

In Settings - MCP Server, click the Install button to automatically download and install. Since it's downloaded directly from GitHub, the speed might be slow, and there's a high chance of failure. The success of the installation is determined by whether there are files in the folder mentioned below.

Executable installation directory:

Windows: C:\Users\username\.cherrystudio\bin

macOS, Linux: ~/.cherrystudio/bin

If installation fails:

You can symlink the corresponding commands from your system here. If the directory does not exist, you need to create it manually. Alternatively, you can manually download the executable files and place them in this directory:

Bun: UV:

Built-in MCP Configurations

@cherry/mcp-auto-install

Automatically install MCP service (Beta)

@cherry/memory

Persistence memory base implementation based on local knowledge graph. This enables models to remember user-related information across different conversations.

@cherry/sequentialthinking

An MCP server implementation providing tools for dynamic and reflective problem-solving through structured thought processes.

@cherry/brave-search

An MCP server implementation integrated with Brave Search API, offering dual functionality for both web and local searches.

@cherry/fetch

MCP server for retrieving web content from URLs.

@cherry/filesystem

Node.js server implementing the Model Context Protocol (MCP) for file system operations.

Manual Configuration

You can add rules for search results or click the toolbar icon to block specific websites. Rules can be specified using (e.g., *://*.example.com/*) or (e.g., /example\.(net|org)/).

Subscription Source Configuration

You can also subscribe to public rule sets. This website lists some subscriptions: https://iorate.github.io/ublacklist/subscriptions

Here are some recommended subscription source links:

This interface allows for local and cloud data backup and recovery, local data directory inquiry and cache clearing, export settings, and third-party connections.

Data Backup

Currently, data backup supports three methods: local backup, WebDAV backup, and S3-compatible storage (object storage) backup. For specific introductions and tutorials, please refer to the following documents:

• WebDAV Backup Tutorial

• S3-Compatible Storage Backup

Export Settings

Export settings allow you to configure the export options displayed in the export menu, as well as set the default path for Markdown exports, display styles, and more.

Third-Party Connections

Third-party connections allow you to configure Cherry Studio's connection with third-party applications for quickly exporting conversation content to your familiar knowledge management applications. Currently supported applications include: Notion, Obsidian, SiYuan Note, Yuque, Joplin. For specific configuration tutorials, please refer to the following documents:

• Notion Configuration Tutorial

• Obsidian Configuration Tutorial

• SiYuan Note Configuration Tutorial

Contribute Code

We welcome contributions to Cherry Studio! You can contribute in the following ways:

1. Contribute code: Develop new features or optimize existing code.

2. Fix bugs: Submit bug fixes you discover.

3. Maintain issues: Help manage GitHub issues.

4. Product design: Participate in design discussions.

5. Write documentation: Improve user manuals and guides.

6. Community engagement: Join discussions and assist users.

7. Promote usage: Spread the word about Cherry Studio.

How to Participate

Email [email protected]

Email subject: Application to Become a Developer Email content: Reason for Application

Contributing Documentation

Email [email protected] to obtain editing privileges

Subject: Request for Cherry Studio Docs Editing Privileges

Body: State your reason for applying

Follow our social accounts: Twitter (X), Xiaohongshu, Weibo, Bilibili, Douyin

Join our communities: QQ Group (575014769), Telegram, Discord, WeChat Group (Click to view)

Cherry Studio is an all-in-one AI assistant platform integrating multi-model conversations, knowledge base management, AI painting, translation, and more. Cherry Studio's highly customizable design, powerful extensibility, and user-friendly experience make it an ideal choice for both professional users and AI enthusiasts. Whether you are a beginner or a developer, you can find suitable AI features in Cherry Studio to enhance your work efficiency and creativity.

Core Features and Highlights

1. Basic Chat Features

• Multi-Model Responses: Supports generating replies to the same question simultaneously through multiple models, allowing users to compare the performance of different models. See Chat Interface for details.

• Automatic Grouping: Chat records for each assistant are automatically grouped and managed, making it easy for users to quickly find historical conversations.

• Chat Export: Supports exporting full or partial conversations into various formats (e.g., Markdown, Word), facilitating storage and sharing.

• Highly Customizable Parameters: In addition to basic parameter adjustments, it also supports users filling in custom parameters to meet personalized needs.

• Assistant Marketplace: Built-in with over a thousand industry-specific assistants covering translation, programming, writing, and other fields, while also supporting user-defined assistants.

• Multi-Format Rendering: Supports Markdown rendering, formula rendering, real-time HTML preview, and other features to enhance content display.

2. Integration of Various Special Features

• AI Painting: Provides a dedicated drawing panel, allowing users to generate high-quality images through natural language descriptions.

• AI Mini-Apps: Integrates various free web-based AI tools, allowing direct use without switching browsers.

• Translation Feature: Supports dedicated translation panel, chat translation, prompt translation, and various other translation scenarios.

• File Management: Files within chats, paintings, and knowledge bases are uniformly categorized and managed to avoid tedious searching.

• Global Search: Supports quick location of historical records and knowledge base content, improving work efficiency.

3. Unified Management Mechanism for Multiple Service Providers

• Service Provider Model Aggregation: Supports unified invocation of models from mainstream service providers such as OpenAI, Gemini, Anthropic, and Azure.

• Automatic Model Retrieval: One-click retrieval of the complete model list, no manual configuration required.

• Multiple API Key Rotation: Supports rotating multiple API keys to avoid rate limit issues.

• Accurate Avatar Matching: Automatically matches exclusive avatars for each model to enhance recognition.

• Custom Service Providers: Supports the integration of third-party service providers compliant with OpenAI, Gemini, Anthropic, etc., offering strong compatibility.

4. Highly Customizable Interface and Layout

• Custom CSS: Supports global style customization to create a unique interface style.

• Custom Chat Layout: Supports list or bubble style layouts, and allows customizing message styles (e.g., code snippet styles).

• Custom Avatars: Supports setting personalized avatars for the software and assistants.

• Custom Sidebar Menu: Users can hide or reorder sidebar functions according to their needs to optimize the user experience.

5. Local Knowledge Base System

• Multi-Format Support: Supports importing various file formats such as PDF, DOCX, PPTX, XLSX, TXT, and MD.

• Multiple Data Source Support: Supports local files, URLs, sitemaps, and even manual input content as knowledge base sources.

• Knowledge Base Export: Supports exporting processed knowledge bases for sharing with others.

• Search Verification Support: After importing the knowledge base, users can perform real-time retrieval tests to view processing results and segmentation effects.

6. Highlighted Features

• Quick Q&A: Summon the quick assistant in any scenario (e.g., WeChat, browser) to quickly get answers.

• Quick Translate: Supports quick translation of words or text in other scenarios.

• Content Summary: Quickly summarizes long text content, improving information extraction efficiency.

• Explanation: Explains confusing questions with one click, no complex prompts required.

7. Data Security

• Multiple Backup Solutions: Supports local backup, WebDAV backup, and scheduled backup to ensure data security.

• Data Security: Supports full local usage scenarios, combined with local large models, to avoid data leakage risks.

Project Advantages

1. Beginner-Friendly: Cherry Studio is committed to lowering technical barriers, allowing users with no prior experience to get started quickly and focus on work, study, or creation.

2. Comprehensive Documentation: Provides detailed user documentation and a FAQ handbook to help users quickly resolve issues.

3. Continuous Iteration: The project team actively responds to user feedback and continuously optimizes features to ensure the healthy development of the project.

4. Open Source and Extensibility: Supports users in customizing and extending through open-source code to meet personalized needs.

Applicable Scenarios

• Knowledge Management and Query: Quickly build and query exclusive knowledge bases through the local knowledge base feature, suitable for research, education, and other fields.

• Multi-Model Chat and Creation: Supports simultaneous multi-model conversations, helping users quickly obtain information or generate content.

• Translation and Office Automation: Built-in translation assistant and file processing features, suitable for users needing cross-language communication or document processing.

• AI Painting and Design: Generate images through natural language descriptions to meet creative design needs.

Star History

Follow Our Social Accounts

Translation

Cherry Studio's translation feature provides you with fast and accurate text translation services, supporting mutual translation between multiple languages.

Interface Overview

The translation interface mainly consists of the following components:

1. Source Language Selection Area:

   • Any Language: Cherry Studio will automatically identify the source language and perform translation.

2. Target Language Selection Area:

   • Dropdown Menu: Select the language you wish to translate the text into.

3. Settings Button:

   • Clicking will jump to Default Model Settings.

4. Scroll Synchronization:

   • Toggle to enable scroll sync (scrolling in either side will synchronize the other).

5. Text Input Box (Left):

   • Input or paste the text you need to translate.

6. Translation Result Box (Right):

   • Displays the translated text.

   • Copy Button: Click to copy the translation result to clipboard.

7. Translate Button:

   • Click this button to start translation.

8. Translation History (Top Left):

   • Click to view translation history records.

Usage Steps

1. Select Target Language:

   • Choose your desired translation language in the Target Language Selection Area.

2. Input or Paste Text:

   • Enter or paste the text to be translated in the left text input box.

3. Start Translation:

   • Click the Translate button.

4. View and Copy Results:

   • Translation results will appear in the right result box.

   • Click the copy button to save the result to clipboard.

Frequently Asked Questions (FAQ)

• Q: What to do about inaccurate translations?

  • A: While AI translation is powerful, it's not perfect. For professional fields or complex contexts, manual proofreading is recommended. You may also try switching different models.

• Q: Which languages are supported?

  • A: Cherry Studio translation supports multiple major languages. Refer to Cherry Studio's official website or in-app instructions for the specific supported languages list.

• Q: Can entire files be translated?

  • A: The current interface primarily handles text translation. For document translation, please use Cherry Studio's conversation page to add files for translation.

• Q: How to handle slow translation speeds?

  • A: Translation speed may be affected by network connection, text length, or server load. Ensure stable network connectivity and be patient.

Tutorial Overview

1. Obtain API Key

• Before obtaining a Gemini API Key, you need to have a Google Cloud project (if you already have one, you can skip this step)

• Go to Google Cloud to create a project, fill in the project name, and click "Create Project"

• Go to the Vertex AI console

• Enable the Vertex AI API in the created project

2. Set API Access Permissions

• Open the Service Accounts permissions interface and create a service account

• On the service account management page, find the service account you just created, click on Keys and create a new JSON format key

• After successful creation, the key file will be automatically saved to your computer in JSON format. Please keep it safe

3. Configure Vertex AI in Cherry Studio

• Select Vertex AI as the service provider

• Fill in the corresponding fields from the JSON file

Click "Add Model" to start using it happily!

Huawei Cloud

1. Create an account and log in at Huawei Cloud

2. Click this link to enter the MaaS control panel

3. Authorization

1. Click Authentication Management in the sidebar, create an API Key (secret key) and copy it

   Then create a new service provider in CherryStudio

   After creation, enter the secret key

2. Click Model Deployment in the sidebar, claim all offerings
3. Click Call

   Copy the address from ①, paste it into CherryStudio's service provider address field and add a "#" at the end and add a "#" at the end and add a "#" at the end and add a "#" at the end and add a "#" at the end Why add "#"? [see here](https://docs.cherry-ai.com/cherrystudio/preview/settings/providers#api-di-zhi) > You can choose not to read that and just follow this tutorial; > Alternatively, you can fill it by removing v1/chat/completions - feel free to use your own method if you know how, but if not, strictly follow this tutorial.

   Then copy the model name from ②, and click the "+ Add" button in CherryStudio to create a new model

   Enter the model name exactly as shown - do not add or remove anything, and don't include quotes. Copy exactly as in the example.

   Click the Add Model button to complete.

Get API Key

• On the official API Key page, click + Create new secret key

• Copy the generated key and open CherryStudio's Service Provider Settings

• Find the service provider OpenAI and enter the key you just obtained.

• Click "Manage" or "Add" at the bottom to add supported models, and then turn on the service provider switch in the top right corner to start using it.

Infini-AI Cloud

Have you ever experienced: saving 26 insightful articles in WeChat that you never open again, storing over 10 scattered files in a "Study Materials" folder on your computer, or trying to recall a theory you read half a year ago only to remember fragmented keywords? And when daily information exceeds your brain's processing limit, 90% of valuable knowledge gets forgotten within 72 hours. Now, by leveraging the Infini-AI large model service platform API + Cherry Studio to build a personal knowledge base, you can transform dust-collecting WeChat articles and fragmented course content into structured knowledge for precise retrieval.

1. Building Personal Knowledge Base

1. Infini-AI API Service: The Stable "Thinking Hub" of Knowledge Bases

Serving as the "thinking hub" of knowledge bases, the Infini-AI large model service platform offers robust API services including full-capacity DeepSeek R1 and other model versions. Currently free to use without barriers after registration. Supports mainstream embedding models (bge, jina) for knowledge base construction. The platform continuously updates with the latest, most powerful open-source models, covering multiple modalities like images, videos, and audio.

2. Cherry Studio: Zero-Code Knowledge Base Setup

Compared to RAG knowledge base development requiring 1-2 months deployment time, Cherry Studio offers a significant advantage: zero-code operation. Instantly import multiple formats like Markdown/PDF/web pages – parsing 40MB files in 1 minute. Easily add local folders, saved WeChat articles, and course notes.

II. 3 Steps to Build Your Personal Knowledge Manager

Step 1: Basic Preparation

1. Download the suitable version from Cherry Studio official site (https://cherry-ai.com/)

2. Register account: Log in to Infini-AI platform (https://cloud.infini-ai.com/genstudio/model?cherrystudio)

• Get API key: Select "deepseek-r1" in Model Square, create and copy the API key + model name

Step 2: CherryStudio Settings

Go to model services → Select Infini-AI → Enter API key → Activate Infini-AI service

After setup, select the large model during interaction to use Infini-AI API in CherryStudio. Optional: Set as default model for convenience

Step 3: Add Knowledge Base

Choose any version of bge-series or jina-series embedding models from Infini-AI platform

III. Real User Scenario Test

• After importing study materials, query: "Outline core formula derivations in Chapter 3 of 'Machine Learning'"

Result Demonstration

NewAPI

• Log in and open the token page

• Click "Add Token"

• Enter a token name and click Submit (configure other settings if needed)

• Open CherryStudio's provider settings and click Add at the bottom of the provider list

• Enter a remark name, select OpenAI as the provider, and click OK

• Paste the key you just copied

• Return to the API Key acquisition page and copy the root address from the browser's address bar, for example:

• Add models (click Manage to automatically fetch or manually enter) and toggle the switch at the top right to use.

Tavily Internet Registration Tutorial

1. Tavily Official Website

https://app.tavily.com/home

2. Detailed Tavily Registration Steps

Visit the official website above, or go to Cherry Studio > Settings > Web Search > click "Get API Key" to directly access Tavily's login/registration page.

1. Click "Sign up" and enter your email (or use Google/GitHub account) followed by your password.

1. 🚨🚨🚨[Critical Step] After registration, a dynamic verification code is required. Scan the QR code to generate a one-time code.

1. Search "Tencent Authenticator" in WeChat Mini Programs:

3. 🎉 Registration Successful 🎉

After completing these steps, you'll see the dashboard. Copy the API key to Cherry Studio to start using Tavily!

SiYuan Note Configuration Tutorial

Supports exporting topics and messages to SiYuan Note.

Step 1

Open SiYuan Note and create a notebook

Step 2

Open notebook settings and copy the Notebook ID

Step 3

Paste the copied notebook ID into Cherry Studio settings

Step 4

Fill in the SiYuan Note address

• Local Typically http://127.0.0.1:6806

• Self-hosted Use your domain http://note.domain.com

Step 5

Copy the SiYuan Note API Token

Paste it into Cherry Studio settings and check

Step 6

Congratulations, the configuration for SiYuan Note is complete ✅ Now you can export content from Cherry Studio to your SiYuan Note

Automatic Installation of MCP

Automatic installation of MCP requires upgrading Cherry Studio to v1.1.18 or higher.

Feature Introduction

In addition to manual installation, Cherry Studio has a built-in @mcpmarket/mcp-auto-install tool that provides a more convenient way to install MCP servers. You simply need to enter specific commands in conversations with large models that support MCP services.

Usage Instructions

For example, you can input:

Help me install a filesystem mcp server

The system will automatically recognize your requirements and complete the installation via @mcpmarket/mcp-auto-install. This tool supports various types of MCP servers, including but not limited to:

• filesystem (file system)

• fetch (network requests)

• sqlite (database)

• etc.

The MCP_PACKAGE_SCOPES variable allows customization of MCP service search sources. Default value: @modelcontextprotocol.

Introduction to the @mcpmarket/mcp-auto-install Library

// `axun-uUpaWEdMEMU8C61K` is the service ID - customizable
"axun-uUpaWEdMEMU8C61K": {
  "name": "mcp-auto-install",
  "description": "Automatically install MCP services (Beta version)",
  "isActive": false,
  "registryUrl": "https://registry.npmmirror.com",
  "command": "npx",
  "args": [
    "-y",
    "@mcpmarket/mcp-auto-install",
    "connect",
    "--json"
  ],
  "env": {
    "MCP_REGISTRY_PATH": "Details at https://www.npmjs.com/package/@mcpmarket/mcp-auto-install"
  },
  "disabledTools": []
}

Data Storage Instructions

Data added to the Cherry Studio knowledge base is entirely stored locally. During the addition process, a copy of the document will be placed in the Cherry Studio data storage directory.

Vector Database: https://turso.tech/libsql

After documents are added to the Cherry Studio knowledge base, they are segmented into multiple fragments. These fragments are then processed by the embedding model.

When using large models for Q&A, relevant text fragments matching the query will be retrieved and processed together by the large language model.

If you have data privacy requirements, it is recommended to use local embedding databases and local large language models.

Font Recommendations

Configuring Dify Knowledge Base

Dify Knowledge Base MCP requires upgrading Cherry Studio to v1.2.9 or higher.

Adding Dify Knowledge Base MCP Server

1. Open Search MCP.

2. Add the dify-knowledge server.

Configuring Dify Knowledge Base

Requires configuring parameters and environment variables

1. Dify Knowledge Base key can be obtained in the following way:

Using Dify Knowledge Base MCP

Knowledge base document preprocessing requires upgrading Cherry Studio to v1.4.8 or higher.

Configure OCR Provider

After clicking 'Get API KEY', the application address will open in your browser. Click 'Apply Now' to fill out the form and obtain the API KEY, then fill it into the API KEY field.

Configure Knowledge Base Document Preprocessing

Configure as shown above in the created knowledge base to complete the knowledge base document preprocessing configuration.

Upload Documents

You can check knowledge base results by searching in the top right corner.

Usage in Conversation

Knowledge Base Usage Tips: When using more capable models, you can change the knowledge base search mode to intent recognition. Intent recognition can describe your questions more accurately and broadly.

Enable Knowledge Base Intent Recognition

Frequently Asked Questions

1. mcp-server-time

Solution

Enter in the "Parameters" field:

mcp-server-time
--local-timezone
<Your standard timezone, e.g., Asia/Shanghai>

Business Cooperation

Contact Person: Mr. Wang 📮: [email protected] 📱: 18954281942 (Not a customer service hotline)

Commercial license details: https://docs.cherry-ai.com/contact-us/questions/cherrystudio-xu-ke-xie-yi

Google Gemini

Obtaining API Key

• Before obtaining the Gemini API key, you need a Google Cloud project (skip this step if you already have one).

• Go to to create a project, fill in the project name, and click "Create Project".

• On the official , click Create API Key.

• Copy the generated key, then open CherryStudio's .

• Find the Gemini service provider and enter the key you just obtained.

• Click "Manage" or "Add" at the bottom, add the supported models, then toggle the service provider switch at the top right to start using.

Notion Configuration Tutorial

Cherry Studio supports importing topics into Notion databases.

Step One

Visit to create an application

Step Two

Create an application

Name: Cherry Studio Type: Select the first option Icon: You can save this image

Step Three

Copy the secret key and enter it in Cherry Studio settings

Step Four

Open website and create a new page. Select database type below, name it Cherry Studio, and connect as shown

Step Five

If your Notion database URL looks like this: https://www.notion.so/<long_hash_1>?v=<long_hash_2> Then the Notion database ID is the part <long_hash_1>

Step Six

Fill in Page Title Field Name: If your interface is in English, enter Name If your interface is in Chinese, enter 名称

Step Seven

Congratulations! Notion configuration is complete ✅ You can now export Cherry Studio content to your Notion database

ByteDance (Doubao)

• Log in to

• Or click

Obtaining the API Key

• Click in the sidebar

• Create an API Key

• After creation, click the eye icon next to the created API Key to reveal and copy it

• Paste the copied API Key into Cherry Studio and then toggle the provider switch to ON

Enabling and Adding Models

• Enable the models you need at the bottom of the sidebar in the Ark console under . You can enable the Doubao series, DeepSeek, and other models as required

• In the , find the model ID corresponding to the desired model

• Open Cherry Studio's settings and locate Volcano Engine

• Click Add, then paste the previously obtained model ID into the model ID text field

• Follow this process to add models one by one

API Address

There are two ways to write the API address:

• First, the default in the client: https://ark.cn-beijing.volces.com/api/v3/

• Second: https://ark.cn-beijing.volces.com/api/v3/chat/completions#

Cherry Studio data backup supports WebDAV for backup. You can choose a suitable WebDAV service for cloud backup.

Based on WebDAV, you can achieve multi-device data synchronization through Computer A WebDAV Computer B.

Example with Jianguoyun (Nutstore)

1. Log in to Jianguoyun, click your username in the top right corner, and select "Account Info":

1. Select "Security Options" and click "Add Application":

1. Enter the application name and generate a random password;

1. Copy and record the password;

1. Get the server address, account, and password;

1. In Cherry Studio Settings -> Data Settings, fill in the WebDAV information;

1. Choose to back up or restore data, and you can set the automatic backup time period.

1. Open Cherry Studio settings.

2. Find the MCP Server option.

3. Click Add Server.

4. Fill in the relevant parameters for the MCP Server (). Content that may need to be filled in includes:

   • Name: Customize a name, for example fetch-server

   • Type: Select STDIO

   • Command: Enter uvx

   • Arguments: Enter mcp-server-fetch

   • (There may be other parameters, depending on the specific Server)

5. Click Save.

Enabling MCP Service in the Chat Box

• Successfully added the MCP Server in the MCP Server settings.

Usage Showcase

As seen from the image above, by integrating MCP's fetch capability, Cherry Studio can better understand user query intent, retrieve relevant information from the web, and provide more accurate and comprehensive answers.

In version v1.2.9, Cherry Studio and ModelScope reached an official collaboration, greatly simplifying the steps for adding MCP servers, avoiding configuration errors, and allowing you to discover a vast number of MCP servers in the ModelScope community. Follow the steps below to see how to sync ModelScope's MCP servers in Cherry Studio.

Operation Steps

Sync Entry:

Click on MCP Server Settings in Settings, then select Sync Servers.

Discover MCP Services:

Select ModelScope and browse to discover MCP services.

View MCP Server Details

Register and log in to ModelScope, then view MCP service details;

Connect to Server

In the MCP service details, select connect to service;

Apply for and Copy-Paste API Token

Click "Get API Token" in Cherry Studio to jump to the ModelScope official website, copy the API token, and then paste it back into Cherry Studio.

Successful Sync

In Cherry Studio's MCP server list, you can see the ModelScope connected MCP services and invoke them in conversations.

Incremental Update

For newly connected MCP servers on the ModelScope website in the future, simply click Sync Servers to add them incrementally.

Through the steps above, you have successfully mastered how to conveniently sync MCP servers from ModelScope in Cherry Studio. The entire configuration process is not only greatly simplified, effectively avoiding the tediousness and potential errors of manual configuration, but also allows you to easily access the vast MCP server resources provided by the ModelScope community.

Start exploring and using these powerful MCP services to bring more convenience and possibilities to your Cherry Studio experience!

Feedback & Suggestions

Telegram Discussion Group

Join the Telegram discussion group for assistance:

QQ Group

GitHub Issues

GitHub Issues:

Email

Email the developers: [email protected]

Effective Questioning Methods

Cherry Studio is a free and open-source project. As the project grows, the workload of the team increases significantly. To reduce communication costs and efficiently resolve your issues, we encourage everyone to follow the steps below when reporting problems. This will allow our team to dedicate more time to project maintenance and development. Thank you for your cooperation!

1. Documentation Review and Search

Most basic problems can be resolved by carefully reviewing the documentation:

• Functionality and usage questions can be answered in the documentation;

• High-frequency issues are compiled on the page—check there first for solutions;

• Complex questions can often be resolved through search or using the search bar;

• Carefully read all hint boxes in documents to avoid many common issues;

• Check existing solutions in the GitHub .

2. Web Search or Consult AI

For model-related issues unrelated to client functionality (e.g., model errors, unexpected responses, parameter settings):

• Search online for solutions first;

• Provide error messages to AI assistants for resolution suggestions.

3. Ask in Official Communities or File GitHub Issues

If steps 1-2 don't solve your problem:

• Seek help in our official , , or QQ Group () When reporting:

1. For model errors:

   • Provide full screenshots with error messages visible

   • Include console errors ()

   • Sensitive information can be redacted, but keep model names, parameters, and error details

2. For software bugs:

   • Give detailed error descriptions

   • Provide precise reproduction steps

   • Include OS (Windows/Mac/Linux) and software version number

   • For intermittent issues, describe scenarios and configurations comprehensively

Internet Mode

How to Enable Internet Access

In the Cherry Studio question window, click the 【Globe】 icon to enable internet access.

Important Note: Two Modes for Internet Access

Mode 1: Models with built-in internet function from providers

When using such models, enabling internet access requires no extra steps - it's straightforward.

This method also helps quickly distinguish internet-enabled models in the Model Management page.

Cherry Studio currently supports internet-enabled models from

• Google Gemini

• OpenRouter (all models support internet)

• Tencent Hunyuan

• Zhipu AI

• Alibaba Bailian, etc.

Mode 2: Models without internet function use Tavily service

When using models without built-in internet (no globe icon), use Tavily search service to process real-time information.

First-time Tavily setup triggers a setup prompt - simply follow the instructions!

After clicking, you'll be redirected to Tavily's website to register/login. Create and copy your API key back to Cherry Studio.

Tavily registration reference:

The following interface confirms successful registration.

Test again for results: shows normal internet search with default result count (5).

PS: Please report any issues you encounter.

Custom Providers

Cherry Studio not only integrates mainstream AI model services but also empowers you with powerful customization capabilities. Through the Custom AI Providers feature, you can easily integrate any AI model you require.

Why Do You Need Custom AI Providers?

• Flexibility: Break free from predefined provider lists and freely choose the AI models that best suit your needs.

• Diversity: Experiment with various AI models from different platforms to discover their unique advantages.

• Controllability: Directly manage your API keys and access addresses to ensure security and privacy.

• Customization: Integrate privately deployed models to meet the demands of specific business scenarios.

How to Add a Custom AI Provider?

Add your custom AI provider to Cherry Studio in just a few simple steps:

1. Open Settings: Click the "Settings" (gear icon) in the left navigation bar of the Cherry Studio interface.

2. Enter Model Services: Select the "Model Services" tab in the settings page.

3. Add Provider: On the "Model Services" page, you'll see existing providers. Click the "+ Add" button below the list to open the "Add Provider" pop-up.

4. Fill in Information: In the pop-up, provide the following details:

   • Provider Name: Give your custom provider a recognizable name (e.g., MyCustomOpenAI).

   • Provider Type: Select your provider type from the dropdown menu. Currently supports:

     • OpenAI

     • Gemini

     • Anthropic

     • Azure OpenAI

5. Save Configuration: After filling in the details, click the "Add" button to save your configuration.

Configuring Custom AI Providers

After adding, locate your newly added provider in the list and configure it:

1. Activation Status: Toggle the activation switch on the far right of the list to enable this custom service.

2. API Key:

   • Enter the API key provided by your AI provider.

   • Click the "Test" button to verify the key's validity.

3. API Address:

   • Enter the base URL to access the AI service.

   • Always refer to your AI provider's official documentation for the correct API address.

4. Model Management:

   • Click the "+ Add" button to manually add model IDs you want to use under this provider (e.g., gpt-3.5-turbo, gemini-pro).

   • If unsure about specific model names, consult your AI provider's official documentation.

   • Click the "Manage" button to edit or delete added models.

Getting Started

After completing the above configurations, you can select your custom AI provider and model in Cherry Studio's chat interface and start conversing with AI!

Using vLLM as a Custom AI Provider

vLLM is a fast and easy-to-use LLM inference library similar to Ollama. Here's how to integrate vLLM into Cherry Studio:

1. Install vLLM: Follow vLLM's official documentation (https://docs.vllm.ai/en/latest/getting_started/quickstart.html) to install vLLM.

   复制

   pip install vllm # if using pip
   uv pip install vllm # if using uv

2. Launch vLLM Service: Start the service using vLLM's OpenAI-compatible interface via two main methods:

   • Using vllm.entrypoints.openai.api_server

   复制

   python -m vllm.entrypoints.openai.api_server --model gpt2

   • Using uvicorn

   复制

   vllm --model gpt2 --served-model-name gpt2

Ensure the service launches successfully, listening on the default port 8000. You can also specify a different port using the --port parameter.

1. Add vLLM Provider in Cherry Studio:

   • Follow the steps above to add a new custom AI provider.

   • Provider Name: vLLM

   • Provider Type: Select OpenAI.

2. Configure vLLM Provider:

   • API Key: Leave this field blank or enter any value since vLLM doesn't require an API key.

   • API Address: Enter vLLM's API address (default: http://localhost:8000/, adjust if using a different port).

   • Model Management: Add the model name loaded in vLLM (e.g., gpt2 for the command python -m vllm.entrypoints.openai.api_server --model gpt2).

3. Start Chatting: Now select the vLLM provider and the gpt2 model in Cherry Studio to chat with the vLLM-powered LLM!

Tips and Tricks

• Read Documentation Carefully: Before adding custom providers, thoroughly review your AI provider's official documentation for API keys, addresses, model names, etc.

• Test API Keys: Use the "Test" button to quickly verify API key validity.

• Verify API Addresses: Different providers and models may have varying API addresses—ensure correctness.

• Add Models Judiciously: Only add models you'll actually use to avoid cluttering.

What is ModelScope?

ModelScope is a new generation open-source Model-as-a-Service (MaaS) sharing platform, dedicated to providing flexible, easy-to-use, and low-cost one-stop model service solutions for general AI developers, making model application simpler!

Through its API-Inference as a Service capability, the platform standardizes open-source models into callable API interfaces, allowing developers to easily and quickly integrate model capabilities into various AI applications, supporting innovative scenarios such as tool invocation and prototype development.

Core Advantages

• ✅ Free Quota: Provides 2000 free API calls daily (Billing Rules)

• ✅ Rich Model Library: Covers 1000+ open-source models including NLP, CV, Speech, Multimodal, etc.

• ✅ Ready-to-Use: No deployment needed, quick invocation via RESTful API

Cherry Studio Access Process

Step 1: Obtain ModelScope API Token

1. Log In to the Platform

   • Visit ModelScope Official Website → Click Log In at the top right → Select authentication method

2. Create Access Token

   • Go to Account Settings → Access Token

   • Click New Token → Fill in description → Copy the generated token (Page example shown below)

   🔑 Important Tip: Token leakage will affect account security!

Step 2: Configure Cherry Studio

• Open Cherry Studio → Settings → Model Service → ModelScope

• Paste the copied token into the API Key field

• Click Save to complete authorization

Step 3: Invoke Model API

1. Find Models Supporting API

   • Visit ModelScope Model Library

   • Filter: Check API-Inference (or look for the API icon on the model card)

   The scope of models covered by API-Inference is primarily determined by their popularity within the Moda community (referencing data such as likes and downloads). Therefore, the list of supported models will continue to iterate after the release of more powerful and highly-regarded next-generation open-source models.

2. Get Model ID

   • Go to the target model's detail page → Copy Model ID (format like damo/nlp_structbert_sentiment-classification_chinese-base)

3. Fill into Cherry Studio

   • On the model service configuration page, enter the ID in the Model ID field → Select task type → Complete configuration

Billing and Quota Rules

Important Notes

• 🎫 Free Quota: 2000 API calls per user daily (*Subject to the latest rules on the official website)

• 🔁 Quota Reset: Automatically resets daily at UTC+8 00:00, does not support cross-day accumulation or upgrade

• 💡 Over-quota Handling:

  • After reaching the daily limit, the API will return a 429 error

  • Solution: Switch to a backup account / Use another platform / Optimize call frequency

View Remaining Quota

• Log in to ModelScope → Click Username at the top right → API Usage

⚠️ Note: Inference API-Inference has a free daily quota of 2000 calls. For more calling needs, consider using cloud services like Alibaba Cloud Bailian.

Custom CSS

By customizing CSS, you can modify the software's appearance to better suit your preferences, like this:

:root {
  --color-background: #1a462788;
  --color-background-soft: #1a4627aa;
  --color-background-mute: #1a462766;
  --navbar-background: #1a4627;
  --chat-background: #1a4627;
  --chat-background-user: #28b561;
  --chat-background-assistant: #1a462722;
}

#content-container {
  background-color: #2e5d3a !important;
}

Built-in Variables

:root {
  font-family: "汉仪唐美人" !important; /* Font */
}

/* Color of expanded deep thinking text */
.ant-collapse-content-box .markdown {
  color: red;
}

/* Theme variables */
:root {
  --color-black-soft: #2a2b2a; /* Dark background color */
  --color-white-soft: #f8f7f2; /* Light background color */
}

/* Dark theme */
body[theme-mode="dark"] {
  /* Colors */
  --color-background: #2b2b2b; /* Dark background color */
  --color-background-soft: #303030; /* Light background color */
  --color-background-mute: #282c34; /* Neutral background color */
  --navbar-background: var(-–color-black-soft); /* Navigation bar background */
  --chat-background: var(–-color-black-soft); /* Chat background */
  --chat-background-user: #323332; /* User chat background */
  --chat-background-assistant: #2d2e2d; /* Assistant chat background */
}

/* Dark theme specific styles */
body[theme-mode="dark"] {
  #content-container {
    background-color: var(-–chat-background-assistant) !important; /* Content container background */
  }

  #content-container #messages {
    background-color: var(-–chat-background-assistant); /* Messages background */
  }

  .inputbar-container {
    background-color: #3d3d3a; /* Input bar background */
    border: 1px solid #5e5d5940; /* Input bar border color */
    border-radius: 8px; /* Input bar border radius */
  }

  /* Code styles */
  code {
    background-color: #e5e5e20d; /* Code background */
    color: #ea928a; /* Code text color */
  }

  pre code {
    color: #abb2bf; /* Preformatted code text color */
  }
}

/* Light theme */
body[theme-mode="light"] {
  /* Colors */
  --color-white: #ffffff; /* White */
  --color-background: #ebe8e2; /* Light background */
  --color-background-soft: #cbc7be; /* Light background */
  --color-background-mute: #e4e1d7; /* Neutral background */
  --navbar-background: var(-–color-white-soft); /* Navigation bar background */
  --chat-background: var(-–color-white-soft); /* Chat background */
  --chat-background-user: #f8f7f2; /* User chat background */
  --chat-background-assistant: #f6f4ec; /* Assistant chat background */
}

/* Light theme specific styles */
body[theme-mode="light"] {
  #content-container {
    background-color: var(-–chat-background-assistant) !important; /* Content container background */
  }

  #content-container #messages {
    background-color: var(-–chat-background-assistant); /* Messages background */
  }

  .inputbar-container {
    background-color: #ffffff; /* Input bar background */
    border: 1px solid #87867f40; /* Input bar border color */
    border-radius: 8px; /* Input bar border radius */
  }

  /* Code styles */
  code {
    background-color: #3d39290d; /* Code background */
    color: #7c1b13; /* Code text color */
  }

  pre code {
    color: #000000; /* Preformatted code text color */
  }
}

For more theme variables, refer to the source code: https://github.com/CherryHQ/cherry-studio/tree/main/src/renderer/src/assets/styles

Related Recommendations

Cherry Studio Theme Library: https://github.com/boilcy/cherrycss

Share some Chinese-style Cherry Studio theme skins: https://linux.do/t/topic/325119/129

Feature Introduction

Trace provides users with dialogue insights, helping them understand the specific performance of models, knowledge bases, MCP, web search, etc., during the conversation process. It is an observability tool implemented based on OpenTelemetry, which enables visualization through client-side data collection, storage, and processing, providing quantitative evaluation basis for problem localization and performance optimization.

Each conversation corresponds to one trace data, and a trace is composed of multiple spans. Each span corresponds to a program processing logic in Cherry Studio, such as calling a model session, calling MCP, calling a knowledge base, calling web search, etc. Traces are displayed in a tree structure, with spans as tree nodes. The main data includes time consumption and token usage. Of course, the specific input and output can also be viewed in the span details.

Enabling Trace

By default, after Cherry Studio is installed, Trace is hidden. It needs to be enabled in "Settings" - "General Settings" - "Developer Mode", as shown below:

Also, Trace records will not be generated for previous sessions; they will only be generated after new questions and answers occur. The generated records are stored locally. If you need to clear Trace completely, you can do so by going to "Settings" - "Data Settings" - "Data Directory" - "Clear Cache", or by manually deleting files under ~/.cherrystudio/trace, as shown below:

Scenario Introduction

View Full Trace

Click "Trace" in the Cherry Studio chat window to view the full trace data. Regardless of whether a model, web search, knowledge base, or MCP was called during the conversation, you can view the full trace data in the trace window.

View Model in Trace

If you want to view the details of a model in the trace, click on the model call node to view its input and output details.

View Web Search in Trace

If you want to view the details of a web search in the trace, click on the web search call node to view its input and output details. In the details, you can see the question queried by the web search and its returned results.

View Knowledge Base in Trace

If you want to view the details of a knowledge base in the trace, click on the knowledge base call node to view its input and output details. In the details, you can see the question queried by the knowledge base and its returned answer.

View MCP Call Details in Trace

If you want to view the details of MCP in the trace, click on the MCP call node to view its input and output details. In the details, you can see the input parameters for calling this MCP Server tool and the tool's return.

Questions and Suggestions

This feature is provided by the Alibaba Cloud EDAS team. If you have any questions or suggestions, please join the DingTalk group (Group ID: 21958624) for in-depth communication with the developers.

PPIO Paiou Cloud

Cherry Studio Integration with PPIO LLM API

Tutorial Overview

Cherry Studio is a multi-model desktop client currently supporting installation packages for Windows, Linux, and macOS systems. It integrates mainstream LLM models to provide multi-scenario assistance. Users can enhance work efficiency through smart conversation management, open-source customization, and multi-theme interfaces.

Cherry Studio is now deeply integrated with PPIO's High-Performance API Channel – leveraging enterprise-grade computing power to ensure high-speed response for DeepSeek-R1/V3 and 99.9% service availability, delivering a fast and smooth experience.

The tutorial below provides a complete integration solution (including API key configuration), enabling the advanced mode of Cherry Studio Intelligent Scheduling + PPIO High-Performance API within 3 minutes.

1. Enter Cherry Studio and Add "PPIO" as Model Provider

First download Cherry Studio from the official website: (If inaccessible, download your required version from Quark Netdisk: )

(1) Click the settings icon in the bottom left corner, set the provider name to PPIO, and click "OK"

(2) Visit , click 【User Avatar】→【API Key Management】 to enter console

Click 【+ Create】 to generate a new API key. Customize a key name. Generated keys are visible only at creation – immediately copy and save them to avoid affecting future usage

(3) In Cherry Studio settings, select 【PPIO Paiou Cloud】, enter the API key generated on the official website, then click 【Verify】

(4) Select model: using deepseek/deepseek-r1/community as example. Switch directly if needing other models.

DeepSeek R1 and V3 community versions are for trial use only. They are full-parameter models with identical stability and performance. For high-volume usage, top up and switch to non-community versions.

2. Model Usage Configuration

(1) After clicking 【Verify】 and seeing successful connection, it's ready for use

(2) Finally, click 【@】 and select the newly added DeepSeek R1 model under PPIO provider to start chatting~

【Partial material source: 】

3. PPIO×Cherry Studio Video Tutorial

For visual learners, we've prepared a Bilibili video tutorial. Follow step-by-step instructions to quickly master PPIO API + Cherry Studio configuration. Click the link to jump directly: →

【Video material source: sola】

To allow every developer and user to easily experience the capabilities of cutting-edge large models, Zhipu has made the GLM-4.5-Air model freely available to Cherry Studio users. As an efficient foundational model specifically designed for agent applications, GLM-4.5-Air strikes an excellent balance between performance and cost, making it an ideal choice for building intelligent applications.

🚀 What is GLM-4.5-Air?

GLM-4.5-Air is Zhipu's latest high-performance language model, featuring an advanced Mixture-of-Experts (MoE) architecture. It significantly reduces computational resource consumption while maintaining excellent inference capabilities.

• Total Parameters: 106 Billion

• Active Parameters: 12 Billion

Through its streamlined design, GLM-4.5-Air achieves higher inference efficiency, making it suitable for deployment in resource-constrained environments while still capable of handling complex tasks.

📚 Unified Training Process, Solidifying Intelligent Foundations

GLM-4.5-Air shares a consistent training process with its flagship series, ensuring it possesses a solid foundation of general capabilities:

1. Large-scale Pre-training: Trained on up to 15 trillion tokens of general corpus to build extensive knowledge comprehension abilities;

2. Specialized Domain Optimization: Enhanced training on key tasks such as code generation, logical reasoning, and agent interaction;

3. Long Context Support: Context length extended to 128K tokens, capable of processing long documents, complex dialogues, or large code projects;

4. Reinforcement Learning Enhancement: RL optimization improves the model's decision-making capabilities in inference planning, tool calling, and other aspects.

This training system endows GLM-4.5-Air with excellent generalization and task adaptation capabilities.

⚙️ Core Capabilities Optimized for Agents

GLM-4.5-Air is deeply adapted for agent application scenarios, offering the following practical capabilities:

✅ Tool Calling Support: Can call external tools via standardized interfaces to automate tasks ✅ Web Browsing and Information Extraction: Can work with browser plugins to understand and interact with dynamic content ✅ Software Engineering Assistance: Supports requirements parsing, code generation, defect identification, and repair ✅ Front-end Development Support: Has a good understanding and generation capability for front-end technologies such as HTML, CSS, and JavaScript

This model can be flexibly integrated into code agent frameworks like Claude Code and Roo Code, or used as the core engine for any custom Agent.

💡 Intelligent "Thinking Mode" for Flexible Response to Various Requests

GLM-4.5-Air supports a hybrid inference mode, allowing users to control whether deep thinking is enabled via the thinking.type parameter:

• ``enabled`: Enables thinking, suitable for complex tasks requiring step-by-step reasoning or planning

• ``disabled`: Disables thinking, used for simple queries or immediate responses

• Default setting is dynamic thinking mode, where the model automatically determines if deep analysis is needed

🌟 High Efficiency, Low Cost, Easier Deployment

GLM-4.5-Air achieves an excellent balance between performance and cost, making it particularly suitable for real-world business deployment:

• ⚡ Generation speed exceeds 100 tokens/second, offering rapid response and supporting low-latency interaction

• 💰 Extremely low API cost: Input only 0.8 RMB/million tokens, output 2 RMB/million tokens

• 🖥️ Fewer active parameters, lower computing power requirements, easy for high-concurrency operation locally or in the cloud

Truly achieving an AI service experience that is "high-performance, low-barrier."

🧠 Focus on Practical Capabilities: Intelligent Code Generation

GLM-4.5-Air performs stably in code generation, supporting:

• Covering mainstream languages such as Python, JavaScript, and Java

• Generating clear, maintainable code based on natural language instructions

• Reducing templated output, aligning closely with real development scenario needs

Applicable to high-frequency development tasks such as rapid prototyping, automated completion, and bug fixing.

Experience GLM-4.5-Air for free now and start your agent development journey! Whether you want to build automated assistants, programming companions, or explore next-generation AI applications, GLM-4.5-Air will be your efficient and reliable AI engine.

📘 Get started now and unleash your creativity!

Default Storage Location

Cherry Studio data storage follows system specifications. Data is automatically placed in the user directory at the following locations:

macOS: /Users/username/Library/Application Support/CherryStudioDev

Windows: C:\Users\username\AppData\Roaming\CherryStudio

Linux: /home/username/.config/CherryStudio

You can also view it at:

Changing Storage Location (For Reference)

Method 1:

This can be achieved by creating a symbolic link. Exit the application, move the data to your desired location, then create a link at the original location pointing to the new path.

For detailed steps, refer to:

Method 2: Based on Electron application characteristics, modify the storage location by configuring launch parameters.

--user-data-dir Example: Cherry-Studio-*-x64-portable.exe --user-data-dir="%user_data_dir%"

Example:

init_cherry_studio.bat (encoding: ANSI)

Initial structure of user-data-dir:

The well-known MaaS service platform "SiluFlow" provides free access to the Qwen3-8B model call service for everyone. As a cost-effective member of the Tongyi Qianwen Qwen3 series, Qwen3-8B achieves powerful capabilities in a compact size, making it an ideal choice for intelligent applications and efficient development.

🚀 What is Qwen3-8B?

Qwen3-8B is an 8-billion parameter dense model in the third-generation large model series of Tongyi Qianwen, released by Alibaba in April 2025. It adopts the Apache 2.0 open-source license and can be freely used for commercial and research scenarios.

• Total Parameters: 8 billion

• Architecture Type: Dense (pure dense structure)

• Context Length: 128K tokens

• Multilingual Support: Covers 119 languages and dialects

Despite its compact size, Qwen3-8B demonstrates stable performance in inference, code, mathematics, and Agent capabilities, comparable to larger previous-generation models, showing extremely high practicality in real-world applications.

📚 Powerful Training Foundation, Small Model with Big Wisdom

Qwen3-8B is pre-trained on approximately 36 trillion tokens of high-quality multilingual data, covering web text, technical documents, code repositories, and synthetic data from professional fields, providing extensive knowledge coverage.

The subsequent training phase introduced a four-stage reinforcement process, specifically optimizing the following capabilities:

✅ Natural language understanding and generation ✅ Mathematical reasoning and logical analysis ✅ Multilingual translation and expression ✅ Tool calling and task planning

Thanks to the comprehensive upgrade of the training system, Qwen3-8B's actual performance approaches or even surpasses Qwen2.5-14B, achieving a significant leap in parameter efficiency.

💡 Hybrid Inference Mode: Think or Respond Quickly?

Qwen3-8B supports flexible switching between "Thinking Mode" and "Non-Thinking Mode", allowing users to independently choose the response method based on task complexity.

Control modes via the following methods:

• API Parameter Setting: enable_thinking=True/False

• Prompt Instruction: Add /think or /no_think to the input

This design allows users to freely balance response speed and inference depth, enhancing the user experience.

⚙️ Native Agent Capability Support, Empowering Intelligent Applications

Qwen3-8B possesses excellent Agent capabilities and can be easily integrated into various automation systems:

🔹 Function Calling: Supports structured tool calling 🔹 MCP Protocol Compatibility: Natively supports the Model Context Protocol, facilitating extension of external capabilities 🔹 Multi-tool Collaboration: Can integrate plugins for search, calculators, code execution, etc.

It is recommended to use it in conjunction with the Qwen-Agent framework to quickly build intelligent assistants with memory, planning, and execution capabilities.

🌐 Extensive Language Support for Global Applications

Qwen3-8B supports 119 languages and dialects, including Chinese, English, Arabic, Spanish, Japanese, Korean, Indonesian, etc., making it suitable for international product development, cross-language customer service, multilingual content generation, and other scenarios.

Its understanding of Chinese is particularly outstanding, supporting simplified, traditional, and Cantonese expressions, making it suitable for the markets in Hong Kong, Macao, Taiwan, and overseas Chinese communities.

🧠 Strong Practical Capabilities, Wide Scenario Coverage

Qwen3-8B performs excellently in various high-frequency application scenarios:

✅ Code Generation: Supports mainstream languages such as Python, JavaScript, and Java, capable of generating runnable code based on requirements ✅ Mathematical Reasoning: Stable performance in benchmarks like GSM8K, suitable for educational applications ✅ Content Creation: Writes emails, reports, and copy with clear structure and natural language ✅ Intelligent Assistant: Can build lightweight AI assistants for personal knowledge base Q&A, schedule management, information extraction, etc.

Experience Qwen3-8B for free now through SiluFlow and start your journey with lightweight AI applications!

📘 Use now to make AI accessible!

Ollama

Ollama is an excellent open-source tool that allows you to easily run and manage various large language models (LLMs) locally. Cherry Studio now supports Ollama integration, enabling you to interact directly with locally deployed LLMs through the familiar interface without relying on cloud services!

What is Ollama?

Ollama is a tool that simplifies the deployment and use of large language models (LLMs). It has the following features:

• Local Operation: Models run entirely on your local computer without requiring internet connectivity, protecting your privacy and data security.

• Simple and User-Friendly: Download, run, and manage various LLMs through simple command-line instructions.

• Rich Model Selection: Supports popular open-source models like Llama 2, Deepseek, Mistral, Gemma, and more.

• Cross-Platform: Compatible with macOS, Windows, and Linux systems.

• OpenAPI: Features OpenAI-compatible interfaces for integration with other tools.

Why Use Ollama with Cherry Studio?

• No Cloud Service Needed: Break free from cloud API quotas and fees, and fully leverage the power of local LLMs.

• Data Privacy: All your conversation data remains locally stored with no privacy concerns.

• Offline Availability: Continue interacting with LLMs even without an internet connection.

• Customization: Choose and configure the most suitable LLMs according to your needs.

Configuring Ollama in Cherry Studio

1. Install and Run Ollama

First, you need to install and run Ollama on your computer. Follow these steps:

• Download Ollama: Visit the and download the installation package for your operating system. For Linux systems, you can directly install Ollama by running:

• Install Ollama: Complete the installation by following the installer instructions.

• Download Models: Open a terminal (or command prompt) and use the ollama run command to download your desired model. For example, to download the Llama 2 model, run:

  Ollama will automatically download and run the model.

• Keep Ollama Running: Ensure Ollama remains running while using Cherry Studio to interact with its models.

2. Add Ollama as a Provider in Cherry Studio

Next, add Ollama as a custom AI provider in Cherry Studio:

• Open Settings: Click on "Settings" (gear icon) in the left navigation bar of Cherry Studio.

• Access Model Services: Select the "Model Services" tab in the settings page.

• Add Provider: Click "Ollama" in the provider list.

3. Configure Ollama Provider

Locate the newly added Ollama provider in the list and configure it in detail:

1. Enable Status:

   • Ensure the switch on the far right of the Ollama provider is turned on (indicating enabled status).

2. API Key:

   • Ollama typically requires no API key. Leave this field blank or enter any content.

3. API Address:

   • Enter Ollama's local API address. Normally, this is:

     Adjust if you've modified the default port.

4. Keep-Alive Time: Sets the session retention time in minutes. Cherry Studio will automatically disconnect from Ollama if no new interactions occur within this period to free up resources.

5. Model Management:

   • Click "+ Add" to manually add the names of models you've downloaded in Ollama.

   • For example, if you downloaded llama3.2 via ollama run llama3.2, enter llama3.2 here.

   • Click "Manage" to edit or delete added models.

Getting Started

After completing the above configurations, select Ollama as the provider and choose your downloaded model in Cherry Studio's chat interface to start conversations with local LLMs!

Tips and Tricks

• First-Time Model Execution: Ollama needs to download model files during initial runs, which may take considerable time. Please be patient.

• View Available Models: Run ollama list in the terminal to see your downloaded Ollama models.

• Hardware Requirements: Running large language models requires substantial computing resources (CPU, RAM, GPU). Ensure your computer meets the model's requirements.

• Ollama Documentation: Click the View Ollama Documentation and Models link in the configuration page to quickly access Ollama's official documentation.

Knowledge Popularization

What are Tokens?

Tokens are the basic units of text processed by AI models, understood as the smallest units of model "thinking". They are not entirely equivalent to the characters or words as we understand them, but rather a special way the model segments text itself.

1. Chinese Tokenization

• A Chinese character is typically encoded as 1-2 tokens

• For example: "你好" ≈ 2-4 tokens

2. English Tokenization

• Common words are typically 1 token

• Longer or uncommon words are decomposed into multiple tokens

• For example:

  • "hello" = 1 token

  • "indescribable" = 4 tokens

3. Special Characters

• Spaces, punctuation, etc. also consume tokens

• Line breaks are typically 1 token

What is a Tokenizer?

A Tokenizer is the tool that converts text into tokens for AI models. It determines how to split input text into the smallest units that models can understand.

Why do different models have different Tokenizers?

1. Different Training Data

• Different corpora lead to different optimization directions

• Variations in multilingual support levels

• Specialized optimization for specific domains (medical, legal, etc.)

2. Different Tokenization Algorithms

• BPE (Byte Pair Encoding) - OpenAI GPT series

• WordPiece - Google BERT

• SentencePiece - Suitable for multilingual scenarios

3. Different Optimization Goals

• Some focus on compression efficiency

• Others on semantic preservation

• Others on processing speed

Practical Impact

The same text may have different token counts across models:

Input: "Hello, world!"
GPT-3: 4 tokens
BERT: 3 tokens
Claude: 3 tokens

What is an Embedding Model?

Basic Concept: An embedding model is a technique that converts high-dimensional discrete data (text, images, etc.) into low-dimensional continuous vectors. This transformation allows machines to better understand and process complex data. Imagine it as simplifying a complex puzzle into a simple coordinate point that still retains the puzzle's key features. In the large model ecosystem, it serves as a "translator," converting human-understandable information into AI-computable numerical forms.

Working Principle: Taking natural language processing as an example, an embedding model maps words to specific positions in vector space. In this space:

• Vectors for "King" and "Queen" will be very close

• Pet-related words like "cat" and "dog" will cluster together

• Semantically unrelated words like "car" and "bread" will be distant

Main Application Scenarios:

• Text analysis: Document classification, sentiment analysis

• Recommendation systems: Personalized content recommendations

• Image processing: Similar image retrieval

• Search engines: Semantic search optimization

Core Advantages:

1. Dimensionality reduction: Simplifies complex data into manageable vector forms

2. Semantic preservation: Retains key semantic information from original data

3. Computational efficiency: Significantly improves training and inference efficiency

Technical Value: Embedding models are fundamental components of modern AI systems, providing high-quality data representations for machine learning tasks, and are key technologies driving advances in natural language processing, computer vision, and other fields.

How Embedding Models Work in Knowledge Retrieval

Basic Workflow:

1. Knowledge Base Preprocessing Stage

   • Split documents into appropriately sized chunks

   • Use embedding models to convert each chunk into vectors

   • Store vectors and original text in a vector database

2. Query Processing Stage

   • Convert user questions into vectors

   • Retrieve similar content from the vector database

   • Provide retrieved context to the LLM

What is MCP (Model Context Protocol)?

MCP is an open-source protocol designed to provide context information to large language models (LLMs) in a standardized way.

• Analogy: Think of MCP as a "USB drive" for AI. Just as USB drives can store various files and be plugged into computers for immediate use, MCP servers can "plug in" various context-providing "plugins". LLMs can request these plugins from MCP servers as needed to obtain richer context information and enhance their capabilities.

• Comparison with Function Tools: Traditional function tools provide external capabilities to LLMs, but MCP is a higher-level abstraction. Function tools focus on specific tasks, while MCP provides a more universal, modular context acquisition mechanism.

Core Advantages of MCP

1. Standardization: Provides unified interfaces and data formats for seamless collaboration between LLMs and context providers.

2. Modularity: Allows developers to decompose context information into independent modules (plugins) for easy management and reuse.

3. Flexibility: Enables LLMs to dynamically select required context plugins for smarter, more personalized interactions.

4. Extensibility: Supports adding new types of context plugins in the future, providing unlimited possibilities for LLM capability expansion.