Show HN: Kore – local AI memory layer with Ebbinghaus forgetting curve

Posted by juanauriti on Thursday February 19, 2026 @02:34AM
from the show-hn dept.

https://github.com/auriti-web-design/kore-memory

I built Kore because every AI memory tool I found either required cloud APIs, was too heavy, or stored everything forever with no pruning.

Kore is different: - Memory decay based on the Ebbinghaus forgetting curve — memories fade unless retrieved, with half-life based on importance (7 days for casual notes, 1 year for critical info) - Auto-importance scoring locally — no LLM call needed - Semantic search in 50+ languages — local sentence-transformers, zero API calls - Memory compression — auto-merges similar memories - Agent namespace isolation — multi-agent safe - Runs fully offline — SQLite + FTS5, FastAPI, no external services

pip install kore-memory[semantic] then kore to start.

Would love feedback on the decay formula and whether the Ebbinghaus approach makes sense for long-running agents.

( 1 comment | on HN )

Re:Show HN: Kore – local AI memory layer with Ebbinghaus for... by maxxmini on Thursday February 19, 2026 @03:36AM (#47071393)

The Ebbinghaus decay approach is really interesting for agent memory. I've been working with a file-based memory system for a long-running AI agent (daily logs + curated long-term memory file), and the biggest challenge is exactly what you're tackling: knowing what to forget.

One thing I've found in practice: importance isn't always knowable at write time. Something that seems trivial today becomes critical context a week later. Have you considered a retrieval-based boost where accessing a memory resets or extends its half-life? That way naturally useful memories self-reinforce.

Also curious about the compression/merging strategy — how do you handle contradictory memories (e.g., "user prefers X" followed by "user now prefers Y")?

[ Parent | on HN ]