mcpField
V
vector-memory

vector-memory

github.com/winsorllcAgent workflows0151 installsUpdated 35w ago
Repository

About vector-memory

A vector-based semantic memory skill for agents that creates embeddings through the OpenAI API and stores text and metadata as JSON. It supports meaning-based retrieval and memory management without an external vector database.

Key features

  1. 01Generate text embeddings automatically
  2. 02Search memories by semantic similarity
  3. 03Find memories similar to an ID
  4. 04Store metadata such as timestamps and tags
  5. 05List, delete, or clear memories

Use cases

  1. 01Retrieve past conversations by meaning for an agent
  2. 02Find documents similar to a selected memory
  3. 03Group or cluster related memories

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/winsorllc/upgraded-carnival --skill vector-memory

After installing, type/vector-memoryin Claude Code to confirm it loaded.

Bundled files

References and scripts installed along with SKILL.md.

test-results.json
test.js
vector-memory.js

Details

From the project README

Vector Memory Skill

This skill provides vector-based semantic memory storage using embeddings for intelligent recall by meaning.

When to Use

  • You need semantic search (find memories by meaning, not keywords)
  • You want to retrieve similar documents or conversations
  • You're building an agent that needs context-aware memory
  • You need to cluster or group related memories

Capabilities

  • vstore: Store text with automatic embedding generation
  • vsearch: Search memories by semantic similarity
  • vdelete: Remove a memory by ID
  • vlist: List all stored memories
  • vsimilar: Find memories similar to a given ID
  • vclear: Clear all memories

How It Works

  1. Text is converted to embeddings using OpenAI's API
  2. Embeddings are stored in JSON with metadata
  3. Search uses cosine similarity to find semantically related memories
  4. No external vector database required - pure JSON storage

Environment Variables

Required:

  • OPENAI_API_KEY - For generating embeddings

Optional:

  • VECTOR_MEMORY_DIM - Embedding dimensions (default: 1536 for text-embedding-ada-002)

Usage Examples

// Store a memory with semantic embedding
vstore('Meeting notes: Discussed Q1 roadmap and budget allocation')
// Returns: "Stored memory with ID: mem_abc123"

// Search by meaning (not keywords)
vsearch('What did we talk about regarding money?')
// Returns: Memories about budget, funding, financial discussions

// Find similar memories
vsimilar('mem_abc123')
// Returns: Semantically similar memories

// List all memories
vlist()
// Returns: List of stored memories with metadata

// Clear all
vclear()
// Returns: "Cleared all vector memories"

Features

  • Semantic search:Find by meaning, not keywords
  • Similarity scoring: Results ranked by relevance score
  • Automatic embeddings: No manual vector generation needed
  • Metadata support: Store timestamps and tags with memories
  • Pure JSON: No external database dependencies

Comments

Markdown supported