# Core Concepts

A short tour of the nouns that show up everywhere in MIRARI.

## Agent Connection

A bound LLM endpoint. One user can bind many. Each connection has:

| Field      | Meaning                                                |
| ---------- | ------------------------------------------------------ |
| `name`     | Human label ("Hermes · Local", "GPT for code review")  |
| `provider` | `hermes` \| `openrouter` \| `nous` \| `custom`         |
| `base_url` | Full URL to the chat-completions root                  |
| `model`    | Model id sent in the request body                      |
| `status`   | `online` \| `offline` \| `error` — updated by **Test** |

The **API key** is never stored server-side. It lives in the browser's local key vault, keyed by connection id.

## Conversation

A chat thread bound to a single connection. Has a `title`, an `agent_connection_id`, and many `messages`. Messages have `role` (`system` / `user` / `assistant`), `content`, and timestamps.

## Memory Node

A single fact, preference, or piece of long-term context. Fields:

* `label` — the fact itself ("User ships on Cloudflare Workers")
* `category` — `preference` / `fact` / `goal` / `style` / free-form
* `strength` — `0.0`–`1.0`; nodes ≥ `0.75` get auto-injected into prompts
* `source` — where it came from (conversation id, mirror report id, manual)

The Atlas is **strength-decayed** — unused nodes drift down, re-referenced nodes drift up.

## Skill

A reusable, versioned instruction the agent can pull in. Statuses move along a pipeline:

```
 suggested → accepted → testing → promoted → archived
       ↑                              │
       └──────── re-suggested ◀───────┘
```

Each skill has:

* `name`, `description`, `category`
* `score` — Mirror Mode rates skills against held-out logs
* `version` — increments when the template prompt changes
* `template_prompt` — the actual prompt fragment injected when active

## Mirror Report

The output of **Mirror Mode**. Generated by feeding logs to a reasoning model and parsing into:

```ts
{
  title: string,
  strengths: string[],
  weaknesses: string[],
  recommendations: string[],
  suggested_prompt: string,
}
```

Reports are immutable; promoting a suggested prompt creates a new version on the connection's system prompt.

## Template

A shareable bundle of skills and (optionally) seed memory. Templates strip identifying info on export so you can publish a "Customer-Support Oracle v2" pack without leaking your private corpus.

## Task

Lightweight todo entries with status `pending` / `in_progress` / `done`. Useful for "things the agent should chip away at over the week" without needing a full project tool.


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## 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://entermirari.gitbook.io/entermirari-docs/getting-started/core-concepts.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.