Skip to content
JuicyChat AI

Chat Memory and Models on JuicyChat

A transcript records old dialogue. Model context is the material available for the next reply. You can still read an old turn even when the model no longer has it in context.

Read History and Recall Separately

In Recent Chats, we keep the opener and the lines you enter on that browser. In model chat, what the next reply can use also depends on the selected model and the information supplied to it. Finding a line in the transcript isn't a test of whether it used that line.

Pro, Max and Alternate Outcomes

Check Chat Model in Chat Settings. Pro models support up to 64K tokens of context. Diamond adds access to Max models with up to 1 million tokens. Capacity matters in a long story, but does not guarantee that a character will bring up the detail you expect.

Changing temperature varies replies, not the context window. If you want to see two outcomes of the same decision, keep one in a separate branch. See the plan limit for parallel chats.

Recap the Fact the Next Turn Needs

We stopped at the west gate. Tess has the crate but not the slip. The registrar is still waiting to see a name.

That's the useful recap if the next exchange is about entering the museum. Leave out details that don't affect the decision.