Retold

Long-term memory for AI agents, with evidence, isolation, and permission to return nothing.

Retold stores durable facts, decisions, and experiences in SQLite. Local embedding, lexical, and entity retrieval can return an explained result or nothing; source quotes, lifecycle rules, and scope checks keep the store useful as it grows.

Try it

python -m pip install "retold[local-models] @ https://github.com/adimyth/retold/releases/download/v1.2.1/retold-1.2.1-py3-none-any.whl"

from retold import Retold

with Retold.open("memory.sqlite", profile="lite") as retold:
    with retold.session(user_id="aditya") as memory:
        memory.remember(
            "I prefer concise answers.",
            evidence="I prefer concise answers.",
            attribute="answer_style",
        )
        print(memory.search("What kind of answers do I prefer?").text)

The first run downloads about 640 MB of local model files. It needs no API key. Start with the usage guide, then read the architecture or API reference when you need the lower-level controls.

What Retold protects

• Record quality

  A direct claim keeps the transcript quote that supports it. Inferences expire unless later evidence reinforces them.

• Retrieval quality

  Dense, lexical, and entity channels feed a relevance gate that can return nothing and explain why.

• Isolation

  Private agent-user memory needs no setup. Shared user, project, and organization scopes require grants before ranking begins.

• Prompt stability

  The default mode exposes memory as tools. The opt-in utility-aware mode admits conditional memory only when it would change a draft.

Current boundary

Retold 1.1 is a Python 3.12 or 3.13 library backed by one SQLite file. Background extraction and utility-aware admission require hosted model clients. The utility-aware measurements come from controlled offline fixtures, so production adopters should start that path in shadow mode.