Show HN: TenseMem – deterministic memory for AI agents

Category: ai-ml

Tags: ai-agents, memory, sqlite, llm, deterministic, knowledge-graph

Score: 7.0/10 (Innovation: 7, Technical: 7, Documentation: 8, Utility: 6)

TenseMem is an embedded, deterministic memory layer for AI agents built on a single SQLite file. It distinguishes facts by tense (current vs. superseded), enforces a single current truth per slot, reports freshness on every read, prevents self-contamination from recalled context, and lets the LLM propose typed ops while the engine resolves them deterministically. It ships with a benchmark (TenseBench), CLI, REST, MCP, and a Studio inspector.

Target audience: backend devs and AI engineers building LLM agents and assistants

Repository: https://github.com/tensemem/tensemem · Python · Apache-2.0 · 5 stars

View on Hacker News