enki-ai

JavaScript bindings for Enki's Rust agent runtime, published as a native Node.js package via napi-rs.

Install

npm install enki-ai

The package ships prebuilt native binaries for:

• Windows x64 and arm64
• macOS x64 and arm64
• Linux x64 and arm64 (GNU libc)

API

The package exposes two layers:

• EnkiAgent: thin wrapper over the native runtime
• Agent: higher-level JavaScript wrapper for tools, memories, and custom LLM providers

It also exports NativeEnkiAgent, Tool, MemoryModule, MemoryBackend, LlmProviderBackend, RunContext, and AgentRunResult.

EnkiAgent

Use EnkiAgent when you want a simple session-oriented interface backed directly by the native runtime.

const { EnkiAgent } = require('enki-ai')

async function main() {
  const agent = new EnkiAgent({
    name: 'Assistant',
    systemPromptPreamble: 'Answer clearly and keep responses short.',
    model: 'ollama::llama3.2:latest',
    maxIterations: 20,
    workspaceHome: process.cwd(),
  })

  const output = await agent.run('session-1', 'Explain what this project does.')
  console.log(output)
}

main().catch(console.error)

Constructor options:

• name?: string
• systemPromptPreamble?: string
• model?: string
• maxIterations?: number
• workspaceHome?: string

Agent

Use Agent when you want to register JavaScript tools or plug in your own LLM provider.

const { Agent } = require('enki-ai')

async function main() {
  const agent = new Agent('demo-model', {
    instructions: 'You are a dice game.',
    workspaceHome: process.cwd(),
  })

  agent.toolPlain(
    function rollDice() {
      return '4'
    },
    {
      description: 'Roll a six-sided die and return the result.',
      parametersJson: JSON.stringify({
        type: 'object',
        properties: {},
        additionalProperties: false,
      }),
    },
  )

  const result = await agent.run('My guess is 4', {
    sessionId: 'session-tools-1',
  })

  console.log(result.output)
}

main().catch(console.error)

Context-aware tools

agent.tool() injects a RunContext as the first argument so your tool can access runtime dependencies.

const { Agent } = require('enki-ai')

const agent = new Agent('demo-model')

agent.tool(
  function getPlayerName(ctx) {
    return ctx.deps.playerName
  },
  {
    description: "Get the player's name.",
    parametersJson: JSON.stringify({
      type: 'object',
      properties: {},
      additionalProperties: false,
    }),
  },
)

Then pass dependencies at run time:

const result = await agent.run('Say hello.', {
  sessionId: 'session-ctx-1',
  deps: { playerName: 'Anne' },
})

Custom LLM providers

Pass either a subclass of LlmProviderBackend or a function through the llm option.

const { Agent, LlmProviderBackend } = require('enki-ai')

class DemoProvider extends LlmProviderBackend {
  complete(model, messages, tools) {
    return {
      model,
      content: `Received ${messages.length} message(s) and ${tools.length} tool(s).`,
    }
  }
}

const agent = new Agent('demo-model', {
  llm: new DemoProvider(),
})

Development

From crates/bindings/enki-js:

yarn install
yarn build
yarn test

Useful scripts:

• yarn build: build the native addon in release mode
• yarn build:debug: build without release optimizations
• yarn test: run the AVA test suite
• yarn lint: run oxlint
• yarn format: run Prettier, cargo fmt, and taplo format

Notes

• Agent can register tools, memories, and custom LLM providers in JavaScript.
• The default native constructor uses 20 max iterations when none is provided.
• workspaceHome lets you control where the runtime creates and resolves workspace state.