mcp-vector-store mcp

v1.0.0 · MCP Tool · ai · registry.pascalai.org

Store, search and manage vectors using MakerAI vector memory (Mem_VStore / Mem_VSearch / Mem_VQuery). Supports upsert, semantic similarity search, metadata filtering and namespace isolation for multi-tenant RAG.

vectorstoreragsemanticsearchembeddingsmemory

MakerAI Pipeline

Input Parameters

ParameterTypeDescription
operationrequired string 'upsert': store/update a document with its embedding; 'search': find similar docs by text; 'query': VQL vector query; 'delete': remove by ID; 'list': list stored IDs; 'stats': storage statistics. One of: upsert, search, query, delete, list, stats.
namespaceoptional string Namespace for isolation (e.g. agent ID, project). Default: 'default'. Default: default.
idoptional string Document ID (upsert, delete).
contentoptional string Text content to embed and store (upsert).
vectoroptional array[number] Pre-computed embedding vector (upsert). If omitted, content is auto-embedded.
metadataoptional object Arbitrary metadata to store alongside the vector.
queryoptional string Text query for semantic search (search operation).
vqloptional string VQL (Vector Query Language) expression for advanced queries (query operation).
limitoptional integer Max results to return. Default: 5. Default: 5.
min_scoreoptional number Minimum similarity score threshold [0, 1]. Default: 0.0. Default: 0.0.
filteroptional object Metadata filter conditions for search results.

Output Fields

FieldTypeDescription
operation string
id string ID of upserted document.
results array[object] Search results ordered by similarity.
count integer
stats object Storage stats: total vectors, namespaces, dimensions.

Examples

Store a document chunk

// Input
{
  "operation": "upsert",
  "namespace": "project-docs",
  "id": "doc_001_chunk_0",
  "content": "MakerAI provides a unified interface for LLM providers.",
  "metadata": {
    "source": "readme.md",
    "page": 1
  }
}

// Output
{
  "operation": "upsert",
  "id": "doc_001_chunk_0"
}

Semantic search with score threshold

// Input
{
  "operation": "search",
  "namespace": "project-docs",
  "query": "how to connect to LLM providers?",
  "limit": 3,
  "min_score": 0.75
}

// Output
{
  "operation": "search",
  "count": 2,
  "results": [
    {
      "id": "doc_001_chunk_0",
      "content": "MakerAI provides...",
      "score": 0.923,
      "metadata": {
        "source": "readme.md"
      }
    }
  ]
}

Install & Discovery

Install

ppm install mcp-vector-store

Get JSON Schema

GET /v1/packages/mcp-vector-store/1.0.0/schema

Discover by keyword

GET /v1/mcp/discover?q=vector
Discovery hint: Install with ppm install mcp-vector-store or invoke remotely via POST /v1/invoke/mcp-vector-store on the MCP Service.

PascalAI Usage

uses toolslib;
var Tool := LoadTool('mcp-vector-store');
Tool.Call(JsonObj(['operation','upsert','namespace','myapp','id','doc1','content',text]));
var R := Tool.Call(JsonObj(['operation','search','namespace','myapp','query',userQuestion,'limit',5]));
for var Item in R['results'] do Writeln(Item['content']);