agent-memory

Agent memory and knowledge management with 4-layer cognitive architecture

Part of the AumOS open-source agent infrastructure portfolio.

Features

• Four-layer cognitive memory model — WORKING (active context), EPISODIC (events), SEMANTIC (facts and concepts), and PROCEDURAL (how-to knowledge) — each with its own storage, retrieval, and eviction semantics
• Importance scoring with five qualitative tiers (CRITICAL through TRIVIAL) and a numeric [0, 1] score, plus time-based decay that reduces importance automatically as memories age
• Contradiction detection and resolution: ContradictionDetector identifies conflicting MemoryEntry pairs and ContradictionResolver applies configurable resolution strategies
• Provenance tracking records the source and trust level of every memory item, with reliability scores that factor into retrieval ranking
• Three storage backends — in-memory, SQLite, and Redis — all behind the MemoryStore ABC; UnifiedMemory provides a single facade across all four layers
• AutoMemorizeMiddleware and ContextBuilder work together to automatically memorize tool outputs and construct retrieval-augmented context windows for LLM calls
• Freshness validation and forced refresh policies prevent stale semantic memories from being surfaced in time-sensitive retrieval queries

Quick Start

Install from PyPI:

pip install agent-memory

Verify the installation:

agent-memory version

Basic usage:

import agent_memory

# See examples/01_quickstart.py for a working example

Documentation

• Architecture
• Contributing
• Changelog
• Examples

Enterprise Upgrade

For production deployments requiring SLA-backed support and advanced integrations, contact the maintainers or see the commercial extensions documentation.

Contributing

Contributions are welcome. Please read CONTRIBUTING.md before opening a pull request.

License

Apache 2.0 — see LICENSE for full terms.

Part of AumOS — open-source agent infrastructure.