# widemem.ai

> Memory infrastructure for LLM agents

widemem is an open-source Python library that gives AI agents persistent memory with importance scoring, temporal decay, hierarchical organization, and YMYL (Your Money or Your Life) prioritization.

## Install

pip install widemem-ai

## Key Features

- Batch conflict resolution: One LLM call for N facts, not N calls
- Importance scoring (1-10) with exponential/linear/step decay
- Hierarchical memory: facts → summaries → themes, auto-routed by query type
- Active retrieval: contradiction detection + clarifying questions
- YMYL prioritization: health, legal, financial facts get immunity from decay
- Self-supervised extraction: collect training data, distill to local model
- Full audit trail: every add, update, delete logged

## Providers

- LLM: OpenAI, Anthropic Claude, Ollama (local)
- Embeddings: OpenAI, Sentence Transformers (local)
- Vector stores: FAISS (local, default), Qdrant

## Quick Start

```python
from widemem import WideMemory

memory = WideMemory()
memory.add("I live in San Francisco and work as an engineer", user_id="alice")
results = memory.search("where does alice live", user_id="alice")
print(results[0].memory.content)
```

## Links

- Website: https://widemem.ai
- GitHub: https://github.com/remete618/widemem-ai
- PyPI: https://pypi.org/project/widemem-ai/
- Whitepaper: https://github.com/remete618/llm-memory-whitepaper
- License: Apache 2.0

## Stats

- 140 tests passing
- Python 3.10+
- Zero external services required (SQLite + FAISS by default)