You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
When using Redis as memory storage in Flowise and enabling a context window, the memory returned to the agent does not contain the latest messages.
Instead, Flowise is returning the oldest messages stored in Redis.
This breaks agent continuity because the conversation is rebuilt with outdated context.
Expected behavior:
When context window is enabled, Flowise should discard the oldest messages, keeping only the lastest X messages (according to the k setted surface window) Example of correct behavior:
Return the last messages first so the agent gets the most recent conversation state
Current behavior:
Flowise appears to apply the slice logic incorrectly, keeping the messages from the start of the array (oldest ones).
As a result, the agent receives outdated context and responds incorrectly, because:
The final messages of the conversation are missing
The model thinks the conversation is still in an old state
Multi-agent workflows break (they depend on recent context)
Steps to reproduce:
Enable Redis as memory storage
Set a context window size (e.g. 10 messages)
Have a conversation with more than 10 interactions
The agent will only receive the first 10 stored messages — not the latest 10
Impact:
This affects:
Multi-agent systems
Chatflows using Supervisor Agents
Agents that make decisions based on recent context
It makes Redis memory unreliable for production usage with Flowise.
reacted with thumbs up emoji reacted with thumbs down emoji reacted with laugh emoji reacted with hooray emoji reacted with confused emoji reacted with heart emoji reacted with rocket emoji reacted with eyes emoji
Uh oh!
There was an error while loading. Please reload this page.
-
Issue:
When using Redis as memory storage in Flowise and enabling a context window, the memory returned to the agent does not contain the latest messages.
Instead, Flowise is returning the oldest messages stored in Redis.
This breaks agent continuity because the conversation is rebuilt with outdated context.
Expected behavior:
When context window is enabled, Flowise should discard the oldest messages, keeping only the lastest X messages (according to the k setted surface window) Example of correct behavior:
const recentMessages = allMessages.slice(-windowSize);
Return the last messages first so the agent gets the most recent conversation state
Current behavior:
Flowise appears to apply the slice logic incorrectly, keeping the messages from the start of the array (oldest ones).
As a result, the agent receives outdated context and responds incorrectly, because:
The final messages of the conversation are missing
The model thinks the conversation is still in an old state
Multi-agent workflows break (they depend on recent context)
Steps to reproduce:
Enable Redis as memory storage
Set a context window size (e.g. 10 messages)
Have a conversation with more than 10 interactions
The agent will only receive the first 10 stored messages — not the latest 10
Impact:
This affects:
Multi-agent systems
Chatflows using Supervisor Agents
Agents that make decisions based on recent context
It makes Redis memory unreliable for production usage with Flowise.
Beta Was this translation helpful? Give feedback.
All reactions