server.json

{
  "$schema": "https://static.modelcontextprotocol.io/schemas/2025-12-11/server.schema.json",
  "name": "io.github.cyberlife-coder/velesdb-memory",
  "title": "VelesDB Memory",
  "description": "Agentic memory for AI agents in one tiny offline binary. remember/recall/relate/forget/why over a tri-engine that fuses vector search, a knowledge graph, and a columnar store — why() walks the graph to reconnect a decision with its context across sessions, surfacing linked facts that share no words with the query.",
  "version": "0.7.0",
  "repository": {
    "url": "https://github.com/cyberlife-coder/VelesDB",
    "source": "github"
  },
  "packages": [
    {
      "registryType": "cargo",
      "registryBaseUrl": "https://crates.io",
      "identifier": "velesdb-memory",
      "version": "0.7.0",
      "transport": {
        "type": "stdio"
      },
      "environmentVariables": [
        {
          "name": "VELESDB_MEMORY_PATH",
          "description": "Directory where the local memory store lives (the memory never leaves this machine).",
          "isRequired": false,
          "default": "~/.velesdb-memory"
        },
        {
          "name": "VELESDB_MEMORY_EMBEDDER",
          "description": "Embedding backend: 'hash' (default, offline, zero-dep) or 'ollama' (real semantic recall via a local model; binary must be built --features ollama).",
          "isRequired": false,
          "default": "hash"
        }
      ]
    }
  ]
}