widemem_addLABSCO SUMMARY
Confidence on every retrieval is the part other memory layers skip.
Search returns a confidence level rather than just the best available match, so an agent can say it does not know instead of guessing from something loosely related — and there are modes for how strictly you want that enforced. The other design choice worth noting is contradiction handling: adding a fact that conflicts with a stored one triggers a single resolution call and updates the memory, rather than leaving both versions in place to be retrieved at random.
What it is
A local-first memory layer for LLMs, usable as a Python library or as an MCP server. Facts are extracted, rated for importance, decayed over time, and searched with a confidence level attached — so an agent knows when memory has nothing for it.
What you get
widemem_add— extracts facts from text, resolves contradictions with what is already stored, and saves themwidemem_search— semantic search across memories, ranked by similarity, importance and recencywidemem_pin— store a fact at elevated importance so it cannot be forgottenwidemem_delete,widemem_count,widemem_exportfor JSON export, andwidemem_health- Importance ratings from 1 to 10 with a choice of decay functions, including one that never decays
- Health, legal, financial and safety facts get an importance floor, immunity from decay, and forced contradiction detection, classified first by pattern and then by the model for implied cases
Requirements
Python 3.10+. Install widemem-ai with the mcp extra and run the server module; storage is SQLite plus FAISS by default, with Qdrant as an alternative. Defaults are fully local — Ollama for the LLM and sentence-transformers for embeddings — configured through WIDEMEM_LLM_PROVIDER, WIDEMEM_LLM_MODEL and WIDEMEM_EMBEDDING_PROVIDER, with data under WIDEMEM_DATA_PATH. OpenAI and Anthropic are supported providers if you prefer them.
Setup effort
One command — pip install "widemem-ai[mcp,faiss,ollama,sentence-transformers]"
USE CASES
01
Remembering a user across sessions
Add what they tell you as it comes up, and the extraction and conflict resolution keep the record current.
02
Making sure something critical never fades
Pin the fact at elevated importance instead of hoping the decay function is kind to it.
03
Letting the agent admit it does not know
Use the confidence level on the search result to decide between answering and abstaining.
04
Keeping everything on your own machine
Run with the local LLM and local embedding providers, so no memory content leaves the device.
05
Moving or backing up what was learned
Export the stored memories as JSON, optionally scoped to one user.
7 tools
Listed in project docs. Your client may show a different list.
widemem_searchwidemem_deletewidemem_countwidemem_pinwidemem_exportwidemem_healthBEFORE YOU INSTALL
What to know first
Tools you get
7 tools named in the project's documentation.
Requires
- Python 3.10+
- Ollama running locally with the llama3.2 model, for the documented configuration
Where it runs
The documented client runs the server on your machine.
INSTALLATION
Connect your client
Server setup
TERMINAL
pip install "widemem-ai[mcp,faiss,ollama,sentence-transformers]"MCP Clients
Claude Desktop
CLAUDE DESKTOP · JSON
{
"mcpServers": {
"widemem": {
"command": "python",
"args": [
"-m",
"widemem.mcp_server"
],
"env": {
"WIDEMEM_LLM_PROVIDER": "ollama",
"WIDEMEM_LLM_MODEL": "llama3.2",
"WIDEMEM_EMBEDDING_PROVIDER": "sentence-transformers"
}
}
}
}Use the python you installed widemem-ai into (give its full path if that is a virtual environment). This configuration uses Ollama at http://localhost:11434 with the llama3.2 model, so have Ollama running with that model pulled.
