Widemem FAQs

Question 1

Is Widemem easy to integrate and customize?

Accepted Answer

Widemem is a local-first Python library designed for quick integration, often in just a few lines of code. It's batteries-included with SQLite and FAISS by default, but offers flexible configuration for various LLM (OpenAI, Anthropic, Ollama), embedding (OpenAI, sentence-transformers), and vector store providers (FAISS, Qdrant).

Question 2

What is Widemem and how does it improve AI memory?

Accepted Answer

Widemem is an open-source AI memory layer designed to give Large Language Models (LLMs) and agents a robust, intelligent long-term memory. It moves beyond basic context windows by implementing importance scoring, temporal decay, and hierarchical organization to remember what truly matters, ensuring critical facts aren't forgotten.

Question 3

How does Widemem prioritize important information?

Accepted Answer

Widemem employs intelligent importance scoring (1-10) and temporal decay, ensuring critical facts (like 'peanut allergy') are retained longer than transient information ('had pizza for lunch'). It also features YMYL (Your Money Your Life) prioritization for health, legal, and financial facts, making them resistant to decay and subject to forced contradiction detection.

Question 4

Can Widemem handle conflicting information?

Accepted Answer

Yes, Widemem includes advanced batch conflict resolution. When new information contradicts existing memories, it detects the conflict and resolves it efficiently with a single LLM call, updating the memory like a reasonable adult would, rather than silently appending contradictions.

Question 5

What kind of memory organization does Widemem offer?

Accepted Answer

Widemem uses a three-layered hierarchical memory system: facts roll up into summaries, and summaries roll up into themes. This allows your AI to 'zoom in' for specific details or 'zoom out' for the big picture without redundant API calls, providing a more nuanced understanding.

Widemem

Widemem

Key Features

Use Cases

Key Features

Use Cases