
SpicyChat AI memory: context, forgetting and Memory Manager
Understand SpicyChat AI context windows, why characters forget, how Memory Manager and Semantic Memory 2.0 work, and what each tier includes.
Why older details disappear
The model can only see a limited context window. Character definition, Persona, instructions, recent messages and replies all share that space. When it fills, the oldest messages leave the active context; the character can no longer use them unless the information is repeated or stored in memory.
Context window by tier
| Tier | Official help-center figure | Memory tools |
|---|---|---|
| Free | 4,096 tokens | None |
| Get A Taste | 4,096 tokens | Manual Memory Manager |
| True Supporter | 8,192 tokens | Memory Manager + Semantic Memory 2.0 |
| I’m All In | 16,384 tokens in support article | Memory Manager + Semantic Memory 2.0 |
How to keep important facts available
- Open the three dots in the chat.
- Enable memory if the menu offers that action.
- Open Memory Manager and save the facts that matter.
- Pin only the most important current-story details so they do not crowd out other memories.
Semantic Memory 2.0 can create memories automatically on True Supporter and above. Starting a new chat still begins with a separate conversation context.
Practical fixes for long roleplays
Keep the core character definition focused, move background detail into Lorebooks, repeat a crucial fact naturally, delete unneeded messages, or clone the conversation from a point where it worked well. More context helps, but clean definitions and deliberate memories also matter.