memory-layer

Store and retrieve decisions, bugfixes, and patterns across sessions using SQLite.

37|8|Updated May 1, 2026
One-click install
npx skills add https://github.com/GeneGulanesJr/LaPis --skill memory-layer
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: memory-layer
Source: https://github.com/GeneGulanesJr/LaPis/tree/main/skills/memory-layer
Command: npx skills add https://github.com/GeneGulanesJr/LaPis --skill memory-layer

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

It solves the problem of losing valuable decisions, bugfixes, patterns, and project context between Pi sessions, forcing you to repeat research and re-diagnose issues from scratch.

Core Features & Use Cases

  • Remembers across sessions: Persistently stores observations such as decisions, bugfixes, constraints, and discoveries in a local SQLite database.
  • Smart recall and context injection: Searches with hybrid ranking (FTS relevance × recency × trust × recall history) and injects the most relevant memories into new sessions while prioritizing architecture and key work.
  • Code and doc intelligence: Indexes code symbols and Markdown docs (tree-sitter AST via WASM and native doc section/link/glossary extraction) so you can retrieve both memory and the underlying source context.
  • Trust scoring and dedup: Deduplicates similar saves, updates in place by ID, soft-deletes stale memories, and adjusts trust based on code changes.

Quick Start

Use memory-layer to search for what you previously decided about a specific component by running: memory search for "how we handle session recovery" in your project.

Frequently Asked Questions about memory-layer

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I persist decisions and bugfixes across coding sessions so I don't repeat research?

Persistent memory stores decisions, bugfixes, and project context in a local SQLite database, enabling smart recall and context injection so you avoid re-diagnosing issues. It uses hybrid ranking combining FTS relevance, recency, trust, and recall history.

How does code indexing work for retrieving source context alongside saved memories?

Code indexing uses web-tree-sitter WASM to parse AST symbols and native extraction for Markdown docs, allowing you to retrieve both memories and underlying source context. This enables hybrid search over memories plus code symbols.

Can I use SQLite-backed persistent memory for project-scoped and personal-scope workflows?

Yes, persistent memory works for both project-scoped and personal-scope workflows, including session lifecycle management, deduplication, trust recovery, and hybrid search. CLI operations support save, search, update, delete, and context retrieval.

What's the best way to search for what I previously decided about a specific component?

Use memory search with a query like "how we handle session recovery" in your project scope. Smart recall applies hybrid ranking that weighs FTS relevance, recency, trust scoring, and recall history to surface the most relevant memories.

How does trust scoring and deduplication handle stale or duplicate memories?

Trust scoring deduplicates similar saves, updates entries in place by ID, soft-deletes stale memories, and adjusts trust based on code changes. This keeps the memory store clean and prioritizes architecture and key work during context injection.

Do I need a local SQLite database to store and retrieve memories across sessions?

Yes, a local SQLite-backed memory store is required for saving and retrieving decisions, bugfixes, and indexed code context. It serves as the foundation for confidence-aware ranking, deduplication, and session lifecycle management.