[KongchangAI]
ProductDeterministic Unit-Record Memory Engine

DUMEdb

A deterministic unit-record memory engine for long-running AI agents that manages context through structured state records rather than replaying chat logs. It uses LMDB for underlying storage and natively supports MCP bridging.

Timeline (last 90 days)

Oct 20

DUMEdb (Deterministic Unit-Record Memory Engine) is a solution shared by a developer on Reddit, used to solve the session amnesia problem of long-running agents.

Unverified50%
Oct 20

DUMEdb's inspiration comes from the 1959 IBM 1401 punch card system, which relied on immutable unit-record accounting rather than replaying history.

Unverified50%
Oct 20

DUMEdb's zero-turn wake handshake (cabinet_wake) consumes less than 220 tokens while preserving 99.8% of the clean context runway.

Unverified50%
Oct 20

DUMEdb uses memory-mapped LMDB storage (dumedb.mdb) underneath, achieving sub-microsecond zero-copy register reads.

Unverified50%
Oct 20

DUMEdb directly audits the transcript.jsonl file on disk, computes the real token consumption after compaction, and triggers a clean session handoff before automatic compaction causes amnesia.

Unverified50%
Oct 20

DUMEdb natively supports an MCP bridge, allowing seamless integration with Google Antigravity, Claude Desktop, Cursor, and Python agent loops.

Unverified50%
Oct 20

DUMEdb offers an 88-second deadpan comedy video, an online studio and interactive calculator, and a GitHub repository, and supports quick installation via pip install dumedb.

Unverified50%

All Facts (7)

Source Articles