GH0STTHE DEVELOPER KIT Get the kit
GH0ST / QWEN / DEVELOPER EDITIONCONNECTING TO MODEL

AN OPEN DOOR TO YOUR NEXT IDEA

Build a
presence.

A little intelligence.
A world of your own making.

Bring a voice, context and your own tools to the things you build. Qwen, connected through GH0ST.

Meet the model Explore the kit

JAVASCRIPT · TYPESCRIPT · PYTHON · HTTP

01 / THE INTELLIGENCE FIELDQWEN3 — 1.7B
One model.
Your possibilities.
ILLUSTRATIVE MODEL FIELD
1.7B

QWEN3 PARAMETERS

4,096

CONTEXT TOKENS

512

MAXIMUM OUTPUT TOKENS

Your tools.

YOUR RULES. YOUR APPLICATION.

02 / FROM LANGUAGE TO SOMETHING MORE

A thought is only
the beginning.

Build an assistant. Give a character a voice. Let an agent work with the tools you trust.

01

A voice.

Generate language that belongs in your application. Supply the personality, instructions and purpose.

TEXT & CONVERSATION ↗
02

A little context.

Bring the conversation with you. Your project decides what to remember and what the model sees.

CONTEXT YOU CONTROL ↗
03

A way to act.

Register your own tools. Validate each request, bound the work, and approve actions when it matters.

BOUNDED TOOL AGENTS ↗

03 / THE PLAYGROUND

Make the
first connection.

This is the real model.
Your words in. Something new out.

YOUR INTENT01 / INPUT
QWEN3 1.7B · LOCAL INFERENCE
THE RESPONSE02 / AWAITING A SIGNAL

A little silence,
before the signal.

REAL MODEL · NO PRERECORDED RESPONSE
LOCAL ALPHA / NO TOKEN CHARGES

Text arrives when generation completes. The illustration is not a visualization of internal model activity.

04 / MAKE IT YOURS

A few lines.
A new possibility.

Start the model API once. Connect from your project. Build the experience around it.

  1. 01
    Start the model.Run the pinned Qwen runtime locally.
  2. 02
    Connect your project.Use the client or call the HTTP endpoint.
  3. 03
    Give it a purpose.Add your context, interface and approved tools.
Read the developer guide ↗
YOUR FIRST CONNECTION
import { createQwenClient } from '@gh0st/agent-kit/client';

const qwen = createQwenClient({
  baseURL: 'http://127.0.0.1:8790/v1'
});

const response = await qwen.generate(
  'Give this idea a voice.'
);

console.log(response.content);
REAL INFERENCE. YOUR APPLICATION.0.2 / ALPHA

05 / AGENCY, WITH INTENTION

Let the model ask.
Let your code decide.

Qwen proposes a tool call. The kit checks its arguments and your approval rules. Your function runs, and the observed result returns to the model.

01Your requestA purpose and a set of rules.
02A model decisionA bounded JSON action.
03Your approved toolValidated before execution.
04An informed answerGrounded in the returned result.

The kit includes a project-catalog example. Generated text is never evaluated as code.

SOMETHING IS WAITING TO BE BUILT.

Give it
a presence.

Get the developer kit

Source, clients and examples. The next part is yours.