# Symbolic Memory

> Stateless symbolic memory effect for LLM agents using SQL facts + canonical semantics, activated via symbols and JIT meaning (PostgreSQL + Ollama).

- Skill: `modbender/symbolic-memory` (Agent Skill, multi-file: 3 files)
- Install (CLI): `npx skillmds@latest add modbender/symbolic-memory`
- Raw SKILL.md: https://api.skillmd.com/api/skills/modbender/symbolic-memory/raw
- Safety review: pending (external: skill-scanner PASS, skillspector PASS)
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Data & Analytics
- Author: modbender (https://skillmd.com/u/modbender)
- Updated: 2026-09-09
- Page: https://skillmd.com/skills/modbender/symbolic-memory

---


# symbolic-memory

## Purpose

Provide a stateless symbolic memory workflow:
- Store facts + canonical semantics in PostgreSQL
- Expose references as symbols
- Activate meaning just-in-time (budgeted)
- Send only activated facts to the LLM (Ollama)

Rule:
Store semantics. Compute meaning. Never confuse the two.

NOTE

This skill does not rely on agent-local persistent memory for long-term state.

Durable knowledge is stored in a shared, versioned symbolic memory substrate
(as defined in the JIT Symbolic Memory design pattern).

The skill may operate without symbolic memory, but long-term coherence
and multi-agent consistency are only guaranteed when it is present.

## Design pattern

https://github.com/Th3Hypn0tist/random/blob/main/jit-symbolic-memory-design-pattern