Doesn't the agent support setting up memory conversations?

[xiaotang-12-ops]

Self Checks

• I have searched for existing issues search for existing issues, including closed ones.
• I confirm that I am using English to submit this report (我已阅读并同意 Language Policy).
• [FOR CHINESE USERS] 请务必使用英文提交 Issue，否则会被关闭。谢谢！:)
• Please do not modify this template :) and fill in all the required fields.

Content

I want to set up a memory dialogue in the agent, but I found that this option is not available.

[dosubot[bot]]

Agent memory (conversation history) is supported in Dify, but it's managed automatically in the backend and not exposed as a user-configurable option in the UI. If your agent is running within a conversation context (i.e., there's a conversation ID), memory is enabled and the agent will have access to previous messages in that session. There’s no explicit toggle or setting to enable/disable memory for agents—it's handled internally by the system based on whether a conversation context exists. You can see this in the code where memory is initialized and used when a conversation ID is present, for example in the agent runner and memory classes (source 1, source 2).

Currently, there’s no UI option to configure or turn off agent memory, and no documentation or roadmap items mention making this configurable. If you need more control over memory, it would require backend code changes.

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}    

[RiskeyL]

@dosu Does an agent app's memory have a window size limit?

[dosubot[bot]]

Yes, agent memory in Dify has a window size limit. The backend enforces a maximum of 2000 tokens and up to 500 messages for the conversation history used as memory. Even if a plugin or agent node specifies a custom window size, these hardcoded caps always apply. These limits are not user-configurable via the UI or environment variables—they can only be changed by modifying the backend code. You can see this logic in the TokenBufferMemory class and how it's used in the agent node implementation (source, source). If you need to adjust these limits, it would require backend code changes.

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}

   

[wanshichenguang]

As the number of interactions increases, the number of tokens grows exponentially.

[valueLzy]

Is this solved?