OpenModelKit

List open models, filter them, and chat — with a model name and a prompt.

Zero runtime dependencies. Works with Ollama, NVIDIA and many more. Includes a small JS SDK, CLI, and first-class LangChain / LangGraph support via ChatOpenModelKit.

npm install openmodelkit

Requires Node.js 18+.

Quick start

1. List free models

import { listModels } from 'openmodelkit';

const { models } = await listModels({
  provider: 'ollama',
  free: true,
});

for (const model of models || []) {
  console.log(model.id || model.name);
}

2. Chat with a model

import { chat } from 'openmodelkit';

const reply = await chat({
  provider: 'ollama',
  model: 'gemma4',
  prompt: 'Explain gravity in simple words',
});

console.log(reply);

3. Full flow (list → chat)

import { listModels, chat } from 'openmodelkit';

const { models } = await listModels({ provider: 'ollama', free: true });
const model = models[0].id || models[0].name;

const reply = await chat({
  provider: 'ollama',
  model,
  prompt: 'Say hello in one sentence',
});

console.log(`Model: ${model}`);
console.log(reply);

Providers

| Provider | Usage | | --- | --- | | Ollama | provider: 'ollama' | | NVIDIA | provider: 'nvidia' |

Model filters

Ollama

| Filter | Values | Description | | --- | --- | --- | | free | true | Free models only | | premium | true | Premium models only | | category | all · free · cloud · embedding · vision · tools · thinking | Filter by model type | | order | popular · newest | Sort order | | page | number | Page number | | limit | number | Results per page |

import { listModels } from 'openmodelkit';

await listModels({ provider: 'ollama', free: true });
await listModels({ provider: 'ollama', premium: true });
await listModels({ provider: 'ollama', category: 'vision' });
await listModels({ provider: 'ollama', category: 'embedding' });
await listModels({ provider: 'ollama', category: 'tools' });
await listModels({ provider: 'ollama', category: 'thinking' });
await listModels({ provider: 'ollama', category: 'cloud' });
await listModels({
  provider: 'ollama',
  category: 'all',
  order: 'newest',
  page: 1,
  limit: 20,
});

NVIDIA

| Filter | Values | Description | | --- | --- | --- | | free | true | Free endpoint models only | | premium | true | Partner / premium models only | | page | number | Page number | | limit | number | Results per page |

import { listModels } from 'openmodelkit';

await listModels({ provider: 'nvidia', free: true });
await listModels({ provider: 'nvidia', premium: true });
await listModels({ provider: 'nvidia', page: 1, limit: 50 });

Authentication

Pass your provider API key when chatting (recommended for production):

await chat({
  provider: 'ollama',
  model: 'gemma4',
  prompt: 'hi',
  apiKey: process.env.OLLAMA_API_KEY,
});

await chat({
  provider: 'nvidia',
  model: 'meta/llama-3.1-8b-instruct',
  prompt: 'hi',
  apiKey: process.env.NVIDIA_API_KEY,
});

Or set environment variables:

export OLLAMA_API_KEY=your_ollama_key
export NVIDIA_API_KEY=your_nvidia_key

Never commit API keys to source control.

LangChain and LangGraph

OpenModelKit ships ChatOpenModelKit, a LangChain BaseChatModel. Use it with LangChain tools, bindTools(), and LangGraph graphs (agent + tool loops).

@langchain/core is an optional peer dependency. Install it when you use this integration. For graphs, also install @langchain/langgraph.

npm install openmodelkit @langchain/core @langchain/langgraph

LangChain chat

import { ChatOpenModelKit } from 'openmodelkit';
import { HumanMessage } from '@langchain/core/messages';

const llm = new ChatOpenModelKit({
  provider: 'ollama',
  model: 'gemma4',
  apiKey: process.env.OLLAMA_API_KEY,
});

const reply = await llm.invoke([
  new HumanMessage('Explain gravity in simple words'),
]);

console.log(reply.content);

Tools with LangChain

import { ChatOpenModelKit } from 'openmodelkit';
import { tool } from '@langchain/core/tools';
import { z } from 'zod';

const weather = tool(
  async ({ city }) => `Weather in ${city}: sunny, 24°C`,
  {
    name: 'get_weather',
    description: 'Get the current weather for a city',
    schema: z.object({ city: z.string() }),
  }
);

const llm = new ChatOpenModelKit({
  provider: 'ollama',
  model: 'gemma4',
}).bindTools([weather]);

const reply = await llm.invoke('What is the weather in Paris?');
console.log(reply.tool_calls);

LangGraph agent

ChatOpenModelKit works as the model node in a LangGraph StateGraph. Bind tools, then use ToolNode and toolsCondition for the agent loop.

import { ChatOpenModelKit } from 'openmodelkit';
import { StateGraph, MessagesAnnotation, START } from '@langchain/langgraph';
import { ToolNode, toolsCondition } from '@langchain/langgraph/prebuilt';
import { HumanMessage, SystemMessage } from '@langchain/core/messages';
import { tool } from '@langchain/core/tools';
import { z } from 'zod';

const weather = tool(
  async ({ city }) => `Weather in ${city}: sunny, 24°C`,
  {
    name: 'get_weather',
    description: 'Get the current weather for a city',
    schema: z.object({ city: z.string() }),
  }
);

const tools = [weather];
const llm = new ChatOpenModelKit({
  provider: 'ollama',
  model: 'gemma4',
}).bindTools(tools);

const graph = new StateGraph(MessagesAnnotation)
  .addNode('agent', async (state) => {
    const out = await llm.invoke([
      new SystemMessage('You are a helpful assistant. Use tools when needed.'),
      ...state.messages,
    ]);
    return { messages: [out] };
  })
  .addNode('tools', new ToolNode(tools))
  .addEdge(START, 'agent')
  .addConditionalEdges('agent', toolsCondition)
  .addEdge('tools', 'agent')
  .compile();

const result = await graph.invoke({
  messages: [new HumanMessage('What is the weather in Paris?')],
});

console.log(result.messages.at(-1).content);

NVIDIA works the same way — set provider: 'nvidia' and model to an NVIDIA model id.

CLI

# List models
npx openmodelkit models --provider ollama --free
npx openmodelkit models --provider ollama --premium
npx openmodelkit models --provider nvidia --free

# Chat
npx openmodelkit chat --provider ollama -m gemma4 "Say hello"
npx openmodelkit chat --provider nvidia -m meta/llama-3.1-8b-instruct "Say hello"

# Help
npx openmodelkit help

API overview

| Export | Purpose | | --- | --- | | listModels(options) | List models with provider filters | | chat(options) | Send a prompt and get a response | | listProviders(options) | List available providers | | OpenModelKit | Reusable client for repeated calls | | ChatOpenModelKit | LangChain BaseChatModel for LangChain and LangGraph | | OpenModelKitError | Structured error (message, code, status) |

TypeScript types are included.

Requirements

• Node.js 18+ (native fetch)
• A valid provider API key when the upstream service requires one
• Optional: @langchain/core (and @langchain/langgraph) for LangChain / LangGraph

Author

Built and maintained by Aakash Chhoker.

• GitHub: github.com/aakashchhoker
• Package: npmjs.com/package/openmodelkit
• Repository: github.com/aakashchhoker/openmodelkit

License

MIT