> For the complete documentation index, see [llms.txt](https://docs.cherryai.com.cn/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.cherryai.com.cn/docs/en-us/advanced-basic/memory.md). # Global Memory {% hint style="warning" %} **Feature Description**: * **Currently only Assistant is supported; Cherry Agent does not yet support global memory.** * This feature is still in the **Beta** phase; overall effectiveness is limited, and it may be adjusted or removed in future versions, so please rely on it with caution. {% endhint %} AI in regular conversations has no long-term memory — every time you start a new topic, it has to introduce itself again. **Global Memory**provides AI with**long-term memory capability**: after enabling, AI will**across conversations**remember key information about you (occupation, preferences, common tone, long-term facts, etc.), which can be directly used in new conversations. > It is recommended to first read [Concept Basics](/docs/en-us/advanced-basic/concepts-101.md) Clarify the relationship between memory and other features. ### Two preparations before enabling Global memory involves two types of work: * **Understanding and organizing memory**: requires a conversation model (the same kind used for daily chats) * **Storing and retrieving memory**: requires an "embedding model" **The role of the embedding model**is**to convert each memory into a numeric vector**, and AI quickly finds relevant memories by comparing vector similarity. Such specialized small models are compact, fast, and low-cost to call. First, in `Settings → Model Services` configure at least one Provider, and **add both a conversation model and an embedding model**. Taking CherryIN as an example: on the Provider details page, click **Get model list**, switch the top tab in the pop-up to the "Embedding" category, and add at least 1 (recommended `bge-m3` or `text-embedding-3-small`). {% hint style="warning" %} If you only add a conversation model and do not add an embedding model, you will not be able to complete memory setup (the embedding model dropdown will show "No data yet"). {% endhint %} ### Enable Global Memory 1. Open `Settings → Global Memory` 2. Click the upper-right corner **Global Memory** toggle

Disabled state: gray toggle in the upper right + empty state "No memories yet"

When the toggle is turned on for the first time, it will automatically pop up the **Memory Settings** dialog, requiring you to select models:

Required items: LLM model + embedding model

* **LLM model**: choose any conversation model you have configured * **The role of the embedding model**: choose any Embedding model you have added * **Embedding dimension**: can be left blank; it will be automatically determined by the default dimension of the Embedding model After filling in, click **OK**. System-level global memory is now enabled. ### You also need to enable it once in each assistant {% hint style="warning" %} **This is where many users get stuck**: after completing all the above steps, AI still seems unable to remember things in new conversations — because global memory has**two layers of toggles**: 1. system level (what you just completed) 2. assistant level (**disabled by default**) Each assistant needs to be enabled separately before memory can be used. {% endhint %} Steps: 1. Go to the chat page, click the target assistant in the assistant list → Edit (or when creating a new assistant in the assistant library) 2. In the assistant settings, find the **Global Memory** toggle, and turn it on 3. After that, the assistant will automatically read and update the memory database in new conversations Enable it once for your commonly used "default assistant," and then you're set for good. ### What should I do if the embedding model dropdown shows "No data yet" The most common sticking point: after you select an LLM model, the embedding model dropdown shows "No data yet".

Typical sticking point: LLM selected (CherryIN), embedding dropdown shows "No data yet"

Reason: there is no Embedding model available in your Provider. Solution: 1. Close the current dialog (click **Cancel**) 2. Go to `Settings → Model Services → your Provider` 3. Click **Get model list**, and switch the tab at the top of the pop-up to **Embedding** 4. Add 1–2 embedding models (such as `bge-m3`,`text-embedding-3-small` etc.) 5. Go back to `Settings → Global Memory`, reopen the toggle ### User Management Global memory is grouped by "user." A `default user`is provided by default, and you can also create separate memory databases for family or team members: * Click **User Management** Right-side +👤 icon → enter a new user ID to create * After switching different users, the following **memory list** and statistics are displayed independently ### Add, view, and delete memories After enabling: * Click **Add your first memory** or **+ Add Memory**, enter the content in the pop-up and save * via **Search memories...** box to filter by keyword * Each memory can be edited or deleted individually * The More actions menu offers **Reset Memory** / **Reset User Memory**, which clears all memories for the current user (irreversible, use with caution) ### Token consumption After enabling global memory, each conversation will consume additional tokens: * Extract question vectors, retrieve candidate memories * Let the LLM evaluate whether to write a new memory + write operation If you are sensitive to cost, you can enable global memory only for the 1–2 most commonly used assistants and keep it off for others. {% hint style="info" %} Global memory is currently in Beta; future versions may adjust the entry point and field names, so please refer to what is actually shown in the app on this page. {% endhint %} ### Tips and Tricks * After enabling it for the first time, you can `Add your first memory` write in **a profile about yourself**(occupation, interests, preferred tone, etc.), and subsequent conversations will automatically refer to it * If you do not want certain assistants to use memory, go to that assistant's settings and disable "Global Memory" separately * For long-term use, it is recommended to regularly clear outdated memories to avoid polluting the context *** ### 💡 Get help and submit feedback If you encounter any questions, bugs, or have suggestions for feature improvements during configuration or use, please refer to [Feedback and Suggestions](/docs/en-us/question-contact/suggestions.md) the official channels provided there. --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter: ``` GET https://docs.cherryai.com.cn/docs/en-us/advanced-basic/memory.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.