jaylfc

jaylfc / taosmd

Public

Framework-agnostic AI memory system. 97.0% Recall@5 on LongMemEval-S — beats MemPalace (96.6%) and agentmemory (95.2%). Standalone library, zero cloud dependencies.

ai-memory arm edge-ai embedding framework-agnostic
11
1
100% credibility
Found Apr 17, 2026 at 11 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

taOSmd is a local AI memory framework that stores and retrieves conversation history with high accuracy using hybrid search techniques, designed to run on affordable hardware without cloud dependencies.

How It Works

1
🔍 Discover taOSmd

You hear about taOSmd, a smart memory helper that lets AI assistants remember conversations perfectly without needing the internet.

2
📦 Easy one-click setup

Run a simple setup command that downloads everything needed and gets your memory system ready in minutes.

3
🤖 Your AI assistant joins in

Tell your AI assistant to register itself, and it sets up its own private memory shelf automatically.

4
💬 Start chatting with your assistant

As you talk, every conversation is safely stored so nothing gets forgotten.

5
🧠 Perfect recall every time

When you ask about past chats, your assistant pulls up exactly what was said, beating other systems in memory tests.

✅ Your AI remembers like a pro

Enjoy conversations where your assistant always knows the full history, all running smoothly on your own computer.

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 11 to 11 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is taosmd?

taosmd is a standalone Python library for building persistent memory into AI agents, storing full conversation transcripts, extracted facts, and session metadata in a local stack of vector search, knowledge graphs, and archives. It handles ingestion from chats, retrieves relevant context via hybrid semantic+keyword search, and supports temporal queries—all with zero cloud dependencies, running on cheap ARM hardware like an Orange Pi. Developers get a framework-agnostic memory system that delivers 97.0% Recall@5 on LongMemEval-S.

Why is it gaining traction?

It beats MemPalace (96.6%) and agentmemory (95.2%) on the same benchmark using the same embedding model, proving superior retrieval without cloud crutches. The one-line setup script pulls models and verifies everything, while agent rules integrate seamlessly into prompts for automatic memory checks. Local speed (0.3ms embeddings) and verbatim archiving make it reliable for long-horizon agent tasks.

Who should use this?

AI agent builders running local LLMs who need offline memory for multi-session reasoning, like personal assistants tracking user prefs or dev agents recalling code decisions across runs. Ideal for taOS users or custom Python setups avoiding vendor lock-in, especially on edge devices.

Verdict

Strong pick for local-first memory—try the agent-install prompt if benchmarks matter. At 1.0% credibility and 11 stars, it's early; docs shine but expect tweaks as it matures.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.