@monodox/openmantis

A small Node.js SDK that acts as an optimization layer between your app and local model runtimes like Ollama and Foundry Local.

Install

npm install @monodox/openmantis

Quick Start

const { createOpenMantis } = require('@monodox/openmantis')

const client = createOpenMantis()
const response = await client.chat({
  messages: [{ role: 'user', content: 'hello' }],
})

console.log(response.content)

CLI

npx @monodox/openmantis run "hello world"

Features

• Runtime adapters for Ollama and Foundry Local
• Disk-backed caching to skip repeated calls
• Token estimation and prompt budgeting
• Context directory indexing
• Lightweight in-memory key-value store

Configuration

| Variable | Description | Default | |----------|-------------|---------| | OPENMANTIS_CACHE_DIR | Cache directory | .openmantis/cache | | OPENMANTIS_RUNTIME_OLLAMA_URL | Ollama endpoint | http://localhost:11434 | | OPENMANTIS_RUNTIME_FOUNDRY_URL | Foundry endpoint | http://localhost:3000 | | OPENMANTIS_MAX_CONTEXT_TOKENS | Max context tokens | 4096 | | OPENMANTIS_MAX_OUTPUT_TOKENS | Max output tokens | 1024 |

API

const {
  createOpenMantis,
  createCache,
  createMemoryStore,
  createRouter,
  createOllamaRuntime,
  createFoundryRuntime,
  indexDirectory,
} = require('@monodox/openmantis')

License

MIT