OpenAI unveiled a major update during Dev Day yesterday (Oct 6), introducing a new suite of tools to make building and deploying AI agents much easier.
โจ Whatโs New:
๐ง The launch includes AgentKit, which gives developers and business users the ability to build, deploy, and optimize agentic AI systems, and ChatKit, a framework for creating rich chat experiences without reinventing the UI layer.
๐ฌ ChatKit lets you embed a production-ready chat interface into your app or website with support for file uploads, tool invocation, and chain-of-thought visualization โ all within minutes.
Together, AgentKit and ChatKit bridge the gap between agent logic and user interaction, making it simpler to bring real AI agents into production products.
๐ก ๐๐ฐ๐จ ๐๐๐ฒ๐ฌ ๐ญ๐จ ๐๐ฌ๐ ๐๐ก๐๐ญ๐๐ข๐ญ::
According to OpenAIโs documentation, ChatKit can be integrated in two ways:
1๏ธโฃ ๐๐๐๐จ๐ฆ๐ฆ๐๐ง๐๐๐ ๐๐ง๐ญ๐๐ ๐ซ๐๐ญ๐ข๐จ๐ง:
Let OpenAI host and scale everything โ you embed the ChatKit widget in your frontend and connect it to an OpenAI Agent Builder backend.
2๏ธโฃ ๐๐๐ฏ๐๐ง๐๐๐ ๐๐ง๐ญ๐๐ ๐ซ๐๐ญ๐ข๐จ๐ง:
Host ChatKit on your own infrastructure, use the ChatKit SDK, and connect it to any backend or model endpoint.
๐งฑ What I Built
I implemented the Advanced Integration version โ a self-hosted Chat Framework combining:
โ ๐๐ก๐๐ญ๐๐ข๐ญ ๐๐ (๐
๐ซ๐จ๐ง๐ญ๐๐ง๐): A modern React + Next.js interface built with ChatKit components. Supports message history, placeholders, and full customization.
โ ๐
๐๐ฌ๐ญ๐๐๐ (๐๐๐๐ค๐๐ง๐): A lightweight layer exposing /api/chat and /health endpoints. Handles message serialization, temperature control, and integrates with any OpenAI-compatible API.
โ ๐๐จ๐ง๐๐ข๐ ๐ฎ๐ซ๐๐๐ฅ๐ ๐๐จ๐๐๐ฅ ๐๐ง๐๐ฉ๐จ๐ข๐ง๐ญ: Flexible backend integration for connecting to any model API or agent orchestration layer.
This setup delivers the same smooth ChatKit experience โ but entirely under developer control. Itโs modular, lightweight, and can easily connect to private APIs, enterprise systems, or custom agent tools.
โ๏ธ How It Works
โ
The ChatKit UI sends messages to my FastAPI backend.
โ
The backend processes, formats, and forwards them to the model API.
โ
The response is returned to the UI and rendered instantly in the chat view.
โ
All data flow follows the OpenAI Chat Completions schema, so itโs plug-and-play with any model or agent backend.
For those whoโd like to explore the setup, Iโve published the full implementation here:
https://lnkd.in/eCvN8mjh
Top comments (0)