How Conversation Memory Works on SoulrisAI
What "memory" means here
Within a single SoulrisAI conversation thread, the AI has access to everything said earlier in that thread — your messages and its own replies. That's what lets you say "make that shorter" or "what about the second option you mentioned" without re-explaining the whole context each time. This is sometimes called context, and it's what makes a conversation feel coherent rather than like a series of disconnected questions.
What doesn't carry over
A new conversation starts with no memory of your other conversations. If you ask something in one chat and open a fresh one later, the AI in the new thread doesn't know what was discussed in the old one unless you tell it. This is deliberate — it keeps each conversation's context relevant to what you're actually doing in it, rather than mixing unrelated topics together.
Your account-level settings — response style and display name — are the exception: those apply to every conversation regardless, because they're profile settings, not conversation content.
Why this matters for how you chat
Two practical implications:
- Keep related follow-ups in the same thread. If you're working through something over several messages — refining a document, debugging an approach, exploring an idea — stay in one conversation rather than starting over. The AI's ability to build on what you've already discussed depends on it.
- Start a new conversation for a genuinely new topic. Carrying unrelated context into a new question doesn't help and can occasionally confuse the response. See starting a new conversation vs. continuing one for more on making that call.
Does attaching a document or image add to memory?
Yes, within that conversation. Once you've sent an image or document for the AI to analyze (see analyzing images and uploading documents), the AI can refer back to it in later messages in the same thread without you re-attaching it.
Is there a limit to how much a conversation can remember?
Very long conversations do have practical limits — extremely long threads can eventually exceed what a model can take into account at once. If a conversation has grown very large and responses start feeling like they're losing track of earlier context, starting a fresh thread (and briefly restating what matters) usually resolves it.
The bottom line
Think of each conversation as its own self-contained context: everything inside it is available to the AI, nothing outside it is, except your personal settings. Use that to decide when to continue a thread and when to start fresh.