Skip to content

Conversation

@varun29ankuS
Copy link

Summary

Adds shodh-memory skill - a persistent memory system for AI agents that maintains context across conversations.

What this skill teaches Claude:

  • When to call proactive_context - Every user message to surface relevant memories
  • How to structure memories - Rich content with reasoning, not just facts
  • Memory types - Decision, Learning, Error, Discovery, Pattern, Context, Task
  • Retrieval strategies - Semantic, associative, and hybrid search
  • Best practices - Consistent tagging, periodic review, trust the decay system

Features of the underlying MCP server:

  • Neuroscience-inspired memory consolidation (working → session → long-term)
  • Hebbian learning for association strengthening
  • Hybrid decay model (exponential + power-law) based on Wixted & Ebbesen (1991)
  • Multi-modal retrieval (semantic similarity + graph traversal)
  • Memory replay and interference detection

Links

Checklist

  • SKILL.md with proper YAML frontmatter
  • README.md with installation instructions
  • Tested with Claude Code

Shodh Memory provides persistent memory across conversations for AI agents.
This skill teaches Claude:
- When to call proactive_context (every user message)
- How to structure memories for optimal retrieval
- Memory types and their importance weighting
- Retrieval strategies (semantic, associative, hybrid)
- Best practices for tagging and organization

Features:
- Neuroscience-inspired memory consolidation
- Hebbian learning for association strengthening
- Hybrid decay (exponential + power-law)
- Multi-modal retrieval (semantic + graph traversal)

GitHub: https://github.com/varun29ankuS/shodh-memory
MCP Server: Available on npm
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant