Janitor AI guide

Janitor AI memory: context, character definitions, and long roleplays

Janitor AI character definitions, memory fields, saved chat history, and the model’s active context are different things. Keep baseline character information in its definition, evolving story facts in a checked continuity note, and the original transcript as a separate record.

Get 1 free summary

Definitions, lore, and chat context do different jobs

A character definition describes the character and scenario; any lore or memory fields you configure can supply selected facts to the prompt. Saved chat history is the conversation record, while active context is the material actually sent for a reply. These layers do not automatically preserve every event from a long, evolving roleplay.

Context size and prompt construction depend on the connected model and settings. Check the current configuration and visible character or memory fields; do not rely on unofficial fixed-limit numbers or assume that inaccessible/hidden character text can be recovered.

Signs it is time to make a continuity note

  • The bot follows its definition but loses a recent alliance, injury, or plot consequence.
  • Facts that exist only in old dialogue stop being reflected in replies.
  • A long roleplay needs a cleaner chapter break.

A practical Janitor AI preservation workflow

Save the source chat text you can access. Supply only the relevant text to a summarizer, or write the continuity note yourself.

Record current story facts, character changes, relationships, known locations, unresolved scenes, and the exact next beat. Mark uncertainty instead of filling gaps.

Review the result against the original before using it as an opening note or updating a memory field. Keep baseline character definition separate, and retain the original chat as its own record.

thredly is independent of Janitor AI. It does not read Janitor AI chats automatically. Copy a conversation section or use a native export when available, then paste or upload the text through thredly's supported web workflow.

What a useful handover should retain

  • Character identity, personality, voice, and any user persona details that still matter.
  • Names, relationships, boundaries, promises, preferences, and corrections that supersede earlier information.
  • Important story events, locations, dates, running jokes, unresolved plot points, and the current objective.
  • A short instruction for the next session that states where the conversation should resume.

Related guides

Frequently asked questions

Is Janitor AI chat history permanent memory?

Chat history and prompt memory are different. A model only receives the context selected for its prompt, while definitions and lore can provide separate durable instructions.

What belongs in a Janitor AI continuity summary?

Put current story facts, character changes, relationships, known locations, unresolved scenes, and the next objective in the summary. Leave baseline character design in the character definition.