Rememoir

Not every memory deserves to be kept. But those that do — deserve to be understood.

Rememoir is a local-first, human-centered cognitive memory system for offline AI agents.
It enables your agent to remember conversations, learn from feedback, and collaborate with you — without ever sending your data to the cloud.

Built for UAssistant and any local LLM agent.

• ✅ 100% offline — no internet required
• ✅ Semantic + contextual recall — finds relevant memories by meaning, not just keywords
• ✅ Feedback-aware learning — adapts based on your corrections and preferences
• ✅ Lightweight & embeddable — powered by LanceDB, zero external dependencies
• ✅ Open source (MIT License) — inspect, modify, redistribute freely
• ✅ Part of the Erabytse ecosystem — tools for intentional digital care

Quick Start

Install:

pip install erabytse-rememoir

Use in your agent:

from erabytse_rememoir import RememoirDB

# Initialize memory for a user (isolated by user_id)
memory = RememoirDB(user_id="alice")

# Add a memory episode
memory.add("I prefer short answers in German.")

# Recall contextually
results = memory.search("How should you answer me?")
print(results[0].content)
# → "I prefer short answers in German."

Philosophy

Rememoir is not a database. It’s a memory companion — designed to forget what’s noise, keep what matters, and always stay under your control.

In a world of surveillance, data extraction, and opaque AI, Rememoir offers a quiet alternative: an intelligent memory that belongs to you, learns from you, and never betrays you.

It embodies Erabytse’s core principle:

Technology should serve attention, not exploit it.

Integration

See examples/integrate_with_uassistant.py for a full walkthrough with UAssistant.

Rememoir works seamlessly with:

• Local LLMs (Ollama, LM Studio, llama.cpp…)
• RAG systems
• Voice or text-based agents
• Personal productivity tools

License

MIT © Erabytse Part of a quiet rebellion against digital waste.