EverOS

<div align="center" id="readme-top">

![EverOS banner](https://github.com/EverMind-AI/EverOS/releases/download/v1.0.0/everos-readme-banner-optimized.jpg)

<p align="center">
  <a href="https://x.com/evermind"><img src="https://img.shields.io/badge/EverMind-000000?labelColor=gray&style=for-the-badge&logo=x&logoColor=white" alt="X"></a>
  <a href="https://huggingface.co/EverMind-AI"><img src="https://img.shields.io/badge/🤗_HuggingFace-EverMind-F5C842?labelColor=gray&style=for-the-badge" alt="HuggingFace"></a>
  <a href="https://discord.gg/gYep5nQRZJ"><img src="https://img.shields.io/badge/dynamic/json?url=https%3A%2F%2Fdiscord.com%2Fapi%2Fv10%2Finvites%2FgYep5nQRZJ%3Fwith_counts%3Dtrue&query=%24.approximate_presence_count&suffix=%20online&label=Discord&color=404EED&labelColor=gray&style=for-the-badge&logo=discord&logoColor=white" alt="Discord"></a>
  <a href="https://github.com/EverMind-AI/EverOS/discussions/67"><img src="https://img.shields.io/badge/WeCom-EverMind_社区-07C160?labelColor=gray&style=for-the-badge&logo=wechat&logoColor=white" alt="WeChat"></a>
</p>

[Website](https://evermind.ai) · [Documentation](https://docs.evermind.ai) · [Blog](https://evermind.ai/blogs) · [中文](README.zh-CN.md)

</div>


<br>

<details>
  <summary><kbd>Table of Contents</kbd></summary>

<br>

- [EverOS 1.0.0 Highlights](#everos-100-highlights)
- [Why EverOS](#why-everos)
- [Quick Start](#quick-start)
- [Architecture At A Glance](#architecture-at-a-glance)
- [Storage Layout](#storage-layout)
- [Features](#features)
- [Project Structure](#project-structure)
- [Documentation](#documentation)
- [Use Cases](#use-cases)
- [Watch EverOS](#watch-everos)
- [EverMind Ecosystems](#evermind-ecosystems)
- [Contributing](#contributing)

<br>

</details>


## EverOS 1.0.0 Highlights

> [!IMPORTANT]
>
> **EverOS 1.0.0 is a major release for self-evolving memory.** It brings a
> local-first runtime, Markdown as the source of truth, hybrid retrieval,
> multimodal ingestion, user and agent memory scopes, and modular algorithms
> through [EverAlgo](https://github.com/EverMind-AI/EverAlgo).
>
> **Watch this repository** for the next wave of memory-system work, including
> Wiki-style knowledge layers and Dreaming for deeper offline evolution.

<table>
<tr>
<td width="33%" valign="top">
<strong>Markdown As Source Of Truth</strong><br>
<br>
All memory is persisted as <code>.md</code> files: readable, editable,
grep-able, Git-versioned, and openable directly in Obsidian.
</td>
<td width="33%" valign="top">
<strong>Local Three-Part Stack</strong><br>
<br>
Markdown + SQLite + LanceDB keep vectors, BM25, and scalar filters
local. No MongoDB, Elasticsearch, or Redis required.
</td>
<td width="33%" valign="top">
<strong>Dual-Track Memory</strong><br>
<br>
Agent memory (<code>cases</code> / <code>skills</code>) and user memory
(<code>episodes</code> / <code>profile</code>) are extracted independently.
</td>
</tr>
<tr>
<td width="33%" valign="top">
<strong>Multimodal Ingestion</strong><br>
<br>
Text, images, audio, documents, PDFs, HTML, and email are unified into
searchable memory.
</td>
<td width="33%" valign="top">
<strong>Self-Evolution</strong><br>
<br>
Common skills are extracted from real usage; repeated patterns become
reusable workflows, no retraining required.
</td>
<td width="33%" valign="top">
<strong>Orthogonal Retrieval</strong><br>
<br>
Search independently by <code>user_id</code>, <code>agent_id</code>,
<code>app_id</code>, <code>project_id</code>, and <code>session_id</code>.
</td>
</tr>
</table>

<br>
<div align="right">

[![](https://img.shields.io/badge/-Back_to_top-gray?style=flat-square)](#readme-top)

</div>


## Why EverOS

EverOS is an open-source Python framework for self-evolving long-term
memory across agents and platforms. It gives makers one portable memory
layer for every agent they use - Claude Code, Codex, OpenClaw, Hermes,
and more - so context, decisions, files, and trajectories can follow the
work instead of staying trapped in one tool.

EverOS stores conversations, agent trajectories, and files as readable
Markdown, then syncs local SQLite and LanceDB indexes for fast retrieval.
Agents can reuse past cases and skills, improve from repeated workflows,
and become more proactive over time.

The system is built around three boundaries:

1. **Memory content stays readable** - Markdown is the durable source of truth.
2. **Runtime state stays local** - SQLite tracks state and LanceDB handles vector, BM25, and scalar-filter search.
3. **Algorithms stay modular** - [EverAlgo](https://github.com/EverMind-AI/EverAlgo) owns memory algorithms; EverOS owns runtime, persistence, online flows, and offline evolution.

<br>
<div align="right">

[![](https://img.shields.io/badge/-Back_to_top-gray?style=flat-square)](#readme-top)

</div>


## Quick Start

### 1. Install EverOS

```bash
uv pip install everos
# or: pip install everos
```

### 2. Initialize Configuration

Generate a starter `.env` file, then fill the API key fields shown in
the generated comments.

```bash
everos init
```

`everos init` writes `./.env` by default. Use `everos init --xdg` to
write `${XDG_CONFIG_HOME:-~/.config}/everos/.env` instead.

### 3. Start The Server

```bash
everos --help
everos server start
```

`everos server start` searches for `.env` in this order: `--env-file <path>` →
`./.env` (cwd) → `${XDG_CONFIG_HOME:-~/.config}/everos/.env` → `~/.everos/.env`.
The endpoint stack is OpenAI-protocol compatible (OpenAI / OpenRouter / vLLM /
Ollama / DeepInfra) - override `*__BASE_URL` in the generated `.env` to point
at any of them.

For a step-by-step walkthrough (add a conversation, flush, search, then
read the markdown), see [QUICKSTART.md](QUICKSTART.md).

### Optional: Ingest Multimodal Files

To ingest non-text content (image / pdf / audio / office documents)
through `/api/v1/memory/add` `content` items, install the optional
extra:

```bash
uv pip install 'everos[multimodal]'   # or: pip install 'everos[multimodal]'
```

This pulls in `everalgo-parser` (with the `[svg]` bundle for SVG
support via cairosvg) and wires up the multimodal LLM client
(`EVEROS_MULTIMODAL__*` fields in `.env`, defaults to
`google/gemini-3-flash-preview` via OpenRouter).

**Office document support requires LibreOffice as a system dependency.**
The parser shells out to `soffice` (LibreOffice's headless renderer) to
convert `.doc` / `.docx` / `.ppt` / `.pptx` / `.xls` / `.xlsx` to PDF
before feeding the result into the multimodal LLM. Without LibreOffice,
office uploads return HTTP 415 with a clear error message; PDF / image
/ audio / HTML / email parsing is unaffected.

Install on the host before serving office documents:

```bash
brew install --cask libreoffice              # macOS
sudo apt-get install -y libreoffice          # Debian / Ubuntu
```

### For Contributors

```bash
git clone https://github.com/EverMind-AI/EverOS.git
cd EverOS
uv sync                              # creates ./.venv and installs deps
source .venv/bin/activate            # or prefix commands with `uv run`
everos init                          # fill the four API key slots in .env (two distinct keys)

everos --help
make test
```

<br>
<div align="right">

[![](https://img.shields.io/badge/-Back_to_top-gray?style=flat-square)](#readme-top)

</div>

## Architecture At A Glance

```
┌───────────────────────────────────────────────┐
│  entrypoints/  (CLI + HTTP API)                │  presentation
├───────────────────────────────────────────────┤
│  service/      (use cases: memorize/retrieve)  │  application
├───────────────────────────────────────────────┤
│  memory/       (extract + search + cascade)    │  domain
├───────────────────────────────────────────────┤
│  infra/        (markdown / sqlite / lancedb)   │  infrastructure
└───────────────────────────────────────────────┘
        ↑                    ↑
   component/            core/
   (LLM/Embedding)       (observability/lifespan)
```

DDD 5 layers, single-direction dependency. See [docs/architecture.md](docs/architecture.md).

<br>
<div align="right">

[![](https://img.shields.io/badge/-Back_to_top-gray?style=flat-square)](#readme-top)

</div>

## Storage Layout

```
~/.everos/
├── default_app/                  # app_id  ("default" → "default_app" on disk)
│   └── default_project/          # project_id ("default" → "default_project")
│       ├── users/<user_id>/
│       │   ├── user.md           # profile
│       │   ├── episodes/         # daily-log episodes (visible)
│       │   ├── .atomic_facts/    # nested facts (dotfile-hidden)
│       │   └── .foresights/      # predictive memory (dotfile-hidden)
│       └── agents/<agent_id>/
│           ├── agent.md
│           ├── .cases/           # one task case per entry
│           └── skills/           # named procedural memories
├── .index/                       # derived indexes (rebuildable from md)
│   ├── sqlite/system.db          # state + queue + audit
│   └── lancedb/*.lance/          # vector + BM25 + scalar
└── .tmp/                         # transient working files
```

Open any `<app>/<project>/users/<user_id>/` folder in Obsidian — your
agent's brain is just files. The dotfile directories (`.atomic_facts/`,
`.foresights/`, `.cases/`) stay hidden by default so the visible folder
is the user-facing memory surface, while extracted derivatives sit
quietly alongside.

<br>
<div align="right">

[![](https://img.shields.io/badge/-Back_to_top-gray?style=flat-square)](#readme-top)

</div>

## Features

- **Hybrid retrieval**: BM25 + cosine vector ANN + scalar filters, backed by LanceDB
- **Cascade index sync**: edit a `.md` → file watcher → entry-level diff → LanceDB sync, sub-second
- **Multi-source extraction**: conversations / agent trajectories / file knowledge
- **Dual-track memory**: user-track (Episodes / Profiles) + agent-track (Cases / Skills)
- **Async-first**: full asyncio, single event loop
- **Multi-modal**: text + small image / audio inline; large media via S3/OSS reference

<br>
<div alig