What is graph-enhanced retrieval?

It is a specialized ranking system that boosts the relevance of memories linked to the specific files or components you are currently working on, ensuring high precision in technical recall.

Does it require an internet connection for semantic search?

No, PACT Memory is designed for local-first performance using sqlite-vec and GGUF models for embeddings, keeping your search capabilities and data on your local machine.

How do I save a new memory into the system?

You can use the /pact-memory save command followed by a description. The skill then automatically structures a comprehensive memory object including context, goals, and entities.

How does PACT Memory differ from standard Claude context?

While standard context is lost when a session ends, PACT Memory stores data in a persistent local SQLite database, allowing Claude to recall information across different sessions and projects.

PACT Persistent Memory

Name: PACT Persistent Memory
Author: ProfSynapse

byProfSynapse

•

Productivity & Workflow

Powers Claude Code with a persistent semantic memory system to track context, decisions, and lessons learned across multiple coding sessions.

About

PACT Memory is a specialized capability for Claude Code that enables long-term knowledge retention by storing rich memory objects including goals, architectural decisions, and technical lessons learned. By utilizing semantic search and graph-enhanced retrieval, it allows agents to recall relevant past experiences and context based on current files or natural language queries. This ensures continuity and cumulative learning throughout the software development lifecycle, preventing the loss of critical context when switching between tasks or starting new sessions.

Key Features

Persistent storage for goals, tasks, and complex architectural decisions
40 GitHub stars
Structured 'Lessons Learned' capture to document and prevent recurring issues
Automatic cross-session tracking of context, entities, and project states
Semantic search utilizing local GGUF embedding models for privacy and speed
Graph-based retrieval that prioritizes memories linked to active files

Use Cases

Recalling the rationale behind a complex architectural decision made in a previous session
Documenting edge cases and bugs to ensure they are not reintroduced during future refactoring
Onboarding Claude to a new sub-task by searching for similar implementation patterns

About

Key Features

Persistent storage for goals, tasks, and complex architectural decisions
40 GitHub stars
Structured 'Lessons Learned' capture to document and prevent recurring issues
Automatic cross-session tracking of context, entities, and project states
Semantic search utilizing local GGUF embedding models for privacy and speed
Graph-based retrieval that prioritizes memories linked to active files

Use Cases

Recalling the rationale behind a complex architectural decision made in a previous session
Documenting edge cases and bugs to ensure they are not reintroduced during future refactoring
Onboarding Claude to a new sub-task by searching for similar implementation patterns