EverMind-AI/EverOS

EverMind‑AI / EverOS – Quick, jargon‑light overview

What it is – EverOS is a Python library that gives AI agents a local‑first, durable memory that lives as ordinary Markdown files. The library watches those files, syncs them into a tiny SQLite database and a LanceDB vector index, and exposes a simple HTTP API for adding, flushing and searching memories.

Why it matters – Most agent‑memory tools store state in opaque databases or remote services, making it hard to version, edit, or back up. EverOS keeps the source of truth in human‑readable .md files that can be diffed, committed to Git, and edited by hand, while still offering fast keyword or vector search.

Core components

• Markdown store – every conversation, file, or “episode” is persisted as a readable Markdown document under ~/.everos.
• SQLite – lightweight relational store for metadata and quick look‑ups.
• LanceDB – optional vector index for embedding‑based (hybrid) search.
• HTTP server – /api/v2/memory/* endpoints let any app add messages, flush a session, or query memories.

Getting started

1. Install with pip install everos (or uv pip install everos).
2. Run everos init and drop an OpenRouter API key into the generated everos.toml.
3. Start the server: everos server start.
4. Use the demo (everos demo) or curl calls to add a conversation, flush it, and search it back.

Features by tier

Tier	What you get
Tier 1 (LLM only)	Markdown persistence, cascade indexing, keyword search.
Add [embedding]	Vector‑based hybrid search, memory reflection, skill extraction.
Add [rerank]	Agentic search, Knowledge Wiki capabilities.
Add [multimodal]	Ingest images, PDFs, audio, Office docs (requires everalgo-parser and LibreOffice).

Use‑case landscape – The README lists dozens of community projects that plug EverOS into smart‑glasses agents, coding assistants, multi‑agent orchestration platforms, memory‑aware games, and personal health aides. All rely on the same durable, editable memory layer.

How to contribute – Clone, run uv sync, try the demo, then make test. The repo follows a typical Python packaging workflow.

Bottom line – EverOS is a genuine, open‑source tool for giving LLM‑driven agents a persistent, version‑controlled memory that works locally without heavyweight services. It fits squarely in the AI‑agent / RAG ecosystem.