Context Memory in Unique AI Chat — Unique AI Documentation
Context Memory in Unique AI Chat
Context memory lets Unique AI Chat remember your working style, preferences, and recurring context across conversations. Instead of re-explaining who you are, how you like answers formatted, or what you are currently working on in every new chat, the assistant maintains a compact memory profile of you that is loaded into every conversation where context memory is enabled.
For example, if you tell the assistant once that you prefer concise answers with concrete examples, or that you work in the risk department on a specific reporting project, future chats can take this into account automatically — in every space that has context memory enabled.
What Context Memory Is
Context memory is a small, structured profile about you, distilled from your conversations with the assistant. It captures durable facts such as:
- Who you are (role, team, language, timezone).
- How you like to communicate (tone, format, level of detail).
- Your current work context (projects, focus areas, goals).
- Your skills and expertise level.
- Recently discussed topics and open follow-ups.
The memory is deliberately not a transcript of your chats. It is rewritten after each conversation turn as a short profile.
Example Memory
Identity
- Works as a credit risk analyst at a Swiss private bank, based in Zurich (CET).
- Native German speaker; communicates with the assistant in English.
Communication Preferences
- Prefers concise answers with concrete examples.
- Wants tables for numeric comparisons instead of prose.
Work Context
- Preparing the quarterly IFRS 9 impairment report.
- Migrating team reporting from Excel to a Python-based pipeline.
Skills & Expertise
- Advanced Excel and SQL.
Recent Topics
- 2026-06-17 11:40 UTC: Discussed expected-credit-loss staging rules.
Open Questions / Follow-ups
- Wants to revisit the scenario-weighting approach.
How It Works
| Step | What happens |
|---|---|
| Start of a turn | If context memory is enabled for the space and for you, your memory profile is loaded and made available to the assistant as part of its instructions, under a section titled “What I know about you”. |
| During the answer | The assistant can read the memory and adapt its tone, examples, and depth to you. It cannot edit the memory while answering. |
| After the answer | A silent post-processing step reviews the latest exchange. If it contains durable new facts about you, your memory profile is updated. If nothing worth remembering happened (small talk, factual questions, code requests), the memory is left unchanged. |
| Next chat | The updated memory is loaded again — in this chat and in any other chat or space where context memory is enabled. |
One Memory Per User — Private to You
You have exactly one context memory. It is shared across all your chats and all spaces that have context memory enabled, so the assistant behaves consistently wherever you work.
Your memory is strictly private:
- Each user's memory is scoped to them. No other user can access your memory, and you cannot access anyone else's.
- The memory is stored in a private, per-user folder in the Knowledge Base that only you can read and write.
- Spaces do not have their own memory — every space with context memory enabled loads the same, single memory profile of you.
Seeing What the Assistant Remembers
You can simply ask the assistant what it knows or remembers about you. Your memory file (memory.md) is also visible to you in the Knowledge Base under the user-memory folder, in a private subfolder named after your user ID.
Managing Context Memory
Open your Settings from the user menu and select the Context Memory tab. A single toggle controls the feature for your account:
- Enabled (default): your memory is loaded into chats and updated after each turn, in every space where context memory is active.
- Disabled: your memory is no longer loaded into any chat and is no longer updated. Per the product behavior described in the settings dialog, disabling deletes your memory.
This personal setting applies across all spaces: even if a space has context memory enabled, your account-level opt-out always wins.
Context Memory and Web Search
By default, context memory is unavailable in spaces where web search is enabled. This keeps your personal context out of external search queries. In such spaces, the assistant simply behaves as if you had no memory; your memory is neither loaded nor updated there. Your administrator can lift this restriction for your organization if the combination is acceptable in your environment.
For Administrators
Context memory must be enabled per space before it is active for users. Administrators can find setup instructions, the memory profile format, configuration details, and feature flags in Context Memory.
Personal settings for context memory