Kindroid documents memory as configuration plus recall pathways

Three roles for memory

Kindroid’s current memory documentation divides the system into persistent, cascaded, and retrievable memory. Persistent material includes user-authored configuration such as backstory, key memories, example messages, directives, group context, plus chat history. Cascaded memory is described as a subscriber feature that extends useful medium-term continuity to hundreds or thousands of earlier messages, with lower fidelity at distance. Retrievable material includes automated long-term memories and journal entries selected when relevant.

This is useful because it distinguishes configuration from recall. A detailed backstory is an always-available instruction-like input; a journal entry is activated by specified keyphrases; long-term memory is selected by relevance. The documentation also says users can see recalled retrievable items on a message through a purple-brain icon, and can enable or disable long-term recall and consolidation at Kindroid or group level. Those controls give a user some visibility into a narrow slice of what influenced a response.

Practical reading

Evidence: the help center explicitly names the systems, tiers, editable personality fields, and recall controls. Inference: the most productive troubleshooting question is not ‘does it have memory?’ but which layer should carry the information. Stable character rules belong in configuration; a contingent fact may fit a journal or long-term memory. That does not guarantee correct retrieval. The documentation itself cautions that retrievable memory can be unreliable and does not expose all material considered during selection.

Evidence limit: This is a documentation reading, not a test of subscription tiers or memory accuracy.