Can I use this to find specific code I wrote weeks ago?

Yes, by querying a topic like 'authentication', you can identify relevant TOC nodes and navigate directly to the specific conversation segments where that code was discussed.

How does time-decayed importance scoring work?

It uses an exponential decay formula (default 30-day half-life) to reduce the weight of older topics, ensuring that your current project priorities surface at the top of results.

What is a Topic Graph in the context of agent memory?

A Topic Graph is a structured map of semantic themes extracted from your AI interactions, showing how different concepts relate to one another through hierarchies and similarities.

Does the Topic Graph work across different projects?

Currently, the topic graph is generated on a per-project basis to maintain strict contextual boundaries and ensure relevant relationship mapping.

How do I update the topics after a long session?

While extraction is scheduled automatically, you can force a refresh using the 'memory-daemon admin extract-topics' command to capture the latest conversation patterns.

Topic Graph Explorer

Name: Topic Graph Explorer
Author: SpillwaveSolutions

bySpillwaveSolutions

•

数据科学与机器学习

Maps and analyzes semantic themes within agent memory using time-decayed importance scoring to discover conversational patterns.

This skill enables deep semantic exploration of agent-memory by constructing a dynamic topic graph. It automatically extracts recurring themes, calculates their relevance using an exponential time-decay algorithm, and identifies relationships such as parent-child hierarchies or co-occurrences. Ideal for developers and AI researchers, it transforms unstructured conversation logs into a navigable map of concepts, allowing users to track how topics evolve over time and retrieve specific context nodes based on high-level thematic queries rather than simple keyword matches.

主要功能

01Semantic topic extraction and clustering from conversation history

02Topic-guided navigation to jump directly to relevant memory segments

034 GitHub stars

04Multi-dimensional relationship mapping including parent, child, and similar topics

05Automated lifecycle management for active, dormant, and resurrected topics

06Time-decayed importance scoring to prioritize recent relevant themes

使用场景

01Identifying recurring architectural patterns across long-term development sessions

02Navigating complex project documentation via high-level thematic connections

03Tracking the evolution of specific technical discussions over weeks or months

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add spillwavesolutions/agent-memory topic-graph

For use in Claude.ai and ChatGPT

Download Skill