---
title: Mnemoria
description: "The local memory platform for your own data — overview of the product line and where MnemoVi fits."
canonical: https://giglabo.com/products/mnemoria
locale: en
---

# Mnemoria

> Markdown twin of https://giglabo.com/products/mnemoria
> Fetch this instead of the HTML page: same content, a fraction of the bytes.
> Site structure and the full page list for agents: https://giglabo.com/llms.txt

The local memory platform for your own data — overview of the product line and where MnemoVi fits.

- Type: page

## What Mnemoria is

Mnemoria is a Rust-based semantic memory engine — one store for everything a team reads, hears and sees. It ingests text, email, documents, audio and video frames, derives embeddings, reranks the candidates and serves retrieval back. It is multi-tenant and access-controlled from the first commit, accelerated on Metal and CUDA, and runs on a zero-dollar stack with no managed-SaaS lock-in.

Mnemoria is the **engine**. The desktop face of it is [MnemoVi — Local Video Knowledge Base](https://giglabo.com/products/mnemoria/applications/mnemovi-local-video-knowledge-base), which runs the same retrieval brains in-process on one machine.

## Capabilities

- **Multi-modal in one store** — text, email, documents, audio (STT) and visual (OCR + embedding) ingestion and retrieval, unified in a single index.
- **Detachable cube architecture** — embedding, reranking, STT and OCR run as independent, hot-swappable serving units, each with its own admission control. Scale the bottleneck without touching the rest.
- **Local and accelerated** — Metal on Apple hardware, CUDA on NVIDIA. Qwen3-Embedding-0.6B and Qwen3-Reranker-0.6B run where your hardware is.
- **Streaming transcription** — real-time SSE transcription keeps audio search responsive; results stream as the model hears them.
- **Multi-tenant and access-controlled** — Keycloak auth, RBAC and Postgres row-level security. Built for teams, isolated by default.
- **Zero-dollar stack** — production-grade retrieval without managed-SaaS lock-in. You own the whole pipeline, top to bottom.

## How it is deployed

- As a **server**, Mnemoria scales across a team: stand it up on hardware you already own, point ingestion at your sources, query the unified index.
- As an **in-process engine**, the same retrieval powers a desktop app — that is what MnemoVi does, with no server, no account and no upload.
- Source and documentation live on [GitHub](https://github.com/giglabo).

## Where to go next

- [MnemoVi — Local Video Knowledge Base](https://giglabo.com/products/mnemoria/applications/mnemovi-local-video-knowledge-base) — the desktop application built on this engine.
- [Connect MnemoVi to an AI agent (MCP)](https://giglabo.com/products/mnemoria/applications/mnemovi-local-video-knowledge-base/mcp) — the local MCP server that lets Claude Desktop, Claude Code, Codex or any MCP agent query your own video archive.

## Related

- HTML version of this page: https://giglabo.com/products/mnemoria
- Site map for agents: https://giglabo.com/llms.txt