How does the memory windows size work in chat flow?

[mosakashaka]

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 set the windows size to 1, but when I ask the third question, I see the previous two questions are still sent both as context. I assume it should only contain the latest question I've asked?

[dosubot[bot]]

The window size represents conversation rounds (user + assistant message pairs), not individual messages. So a window size of 1 should include 2 prompt messages (1 user question + 1 assistant response).

However, what you're experiencing is a known bug. If you're using an agent node, there's an issue where the code doesn't properly check if window.enabled is True before applying the message limit, causing your window size settings to be ignored.

LLM nodes handle this correctly by checking both window.enabled and the size value, but agent nodes currently skip that validation.

Workaround: If you're self-hosting, you can manually edit api/dify_graph/nodes/agent/agent_node.py (around line 332-334) to add the proper check:

prompt_messages = memory.get_history_prompt_messages(
    message_limit=node_data.memory.window.size if (
        node_data.memory.window.enabled and 
        node_data.memory.window.size is not None and 
        node_data.memory.window.size > 0
    ) else None
)

Then restart the backend service. This isn't configurable via environment variables, so the code change is the only reliable fix for now.

To reply, just mention @dosu.

---

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