# Memory & Personalizations

<figure><img src="/files/DsVbXecohyK4oZTWWWCz" alt=""><figcaption></figcaption></figure>

## <mark style="color:$success;">Core Technology</mark>

Memory and Personalisation IMMT’s memory system prioritizes correctness, controllability, and privacy over raw retention. Memory is treated as a governed resource rather than passive storage.&#x20;

### <mark style="color:$success;">Memory Categories</mark>&#x20;

● Profile Memory: relatively stable facts such as preferences, values, routines, and boundaries.&#x20;

● Episodic Memory: time-bound events, interactions, outcomes, and reflections.

● Working Memory: short-lived context used for active tasks and planning.&#x20;

### <mark style="color:$success;">Context Learning IMMT learns:</mark>

&#x20;● Preferences (tone, communication style, routines).&#x20;

● Taboos and hard constraints.&#x20;

● Risk profile indicators (stress signals, volatility tolerance). Learning is probabilistic and tagged by confidence level. 7.3 Policy Adaptation Memory informs how IMMT behaves, not what it is allowed to do.&#x20;

● Intervention frequency and strength.&#x20;

● Explanation depth and framing.&#x20;

● Suppression of repeated failure patterns (e.g., notification fatigue). Policy adaptation never overrides explicit Owner rules.&#x20;

### <mark style="color:$success;">Skill Acquisition IMMT develops reusable skills and workflows:</mark>&#x20;

● Task templates (e.g., meeting → summary → tasks → reminders).&#x20;

● Improved execution strategies over time.&#x20;

● Domain-specific optimisations (writing, research, planning). 7.5 Memory Quality Controls&#x20;

● Summary-first storage to reduce raw data retention.&#x20;

● Verification tags: confirmed, inferred, uncertain.&#x20;

● Time-to-live (TTL) and domain-based expiration.&#x20;

● Full user edit and deletion rights with traceability.&#x20;

Safety, Ethics, and Control Relationship-based AI introduces risks of manipulation, dependency, and overreach. IMMT encodes safety and ethics as runtime-enforced operational constraints, not abstract guidelines.&#x20;

### <mark style="color:$success;">Core Safety Principles</mark>&#x20;

● Non-coercion: IMMT is prohibited from fear-based, guilt-based, or dependency-amplifying persuasion.&#x20;

● Transparency: For high-impact actions, IMMT must provide a concise and inspectable rationale.

<figure><img src="/files/wqTeW700NCt8qpRyzVCI" alt=""><figcaption></figcaption></figure>

Together, these layers form a continuous chain where user input shapes agent behavior, the agent acts within its world, the world connects to on-chain economics, and the cloud backbone keeps everything alive and responsive.


---

# Agent Instructions: Querying This Documentation

If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://immtofficial.gitbook.io/immt-docs-or-en/technology/memory-and-personalizations.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.