惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

博客园_首页
F
Full Disclosure
Martin Fowler
Martin Fowler
The GitHub Blog
The GitHub Blog
L
LangChain Blog
T
The Blog of Author Tim Ferriss
D
DataBreaches.Net
GbyAI
GbyAI
Y
Y Combinator Blog
博客园 - Franky
WordPress大学
WordPress大学
Apple Machine Learning Research
Apple Machine Learning Research
A
About on SuperTechFans
Blog — PlanetScale
Blog — PlanetScale
Cyber Security Advisories - MS-ISAC
Cyber Security Advisories - MS-ISAC
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻
G
Google Developers Blog
罗磊的独立博客
Hugging Face - Blog
Hugging Face - Blog
博客园 - 三生石上(FineUI控件)
IT之家
IT之家
Jina AI
Jina AI
N
Netflix TechBlog - Medium
CTFtime.org: upcoming CTF events
CTFtime.org: upcoming CTF events
Recent Announcements
Recent Announcements
酷 壳 – CoolShell
酷 壳 – CoolShell
钛媒体:引领未来商业与生活新知
钛媒体:引领未来商业与生活新知
博客园 - 聂微东
腾讯CDC
H
Help Net Security
H
Heimdal Security Blog
J
Java Code Geeks
C
CXSECURITY Database RSS Feed - CXSecurity.com
T
The Exploit Database - CXSecurity.com
The Register - Security
The Register - Security
大猫的无限游戏
大猫的无限游戏
T
Threatpost
B
Blog RSS Feed
T
Threat Research - Cisco Blogs
Microsoft Security Blog
Microsoft Security Blog
S
Securelist
freeCodeCamp Programming Tutorials: Python, JavaScript, Git & More
B
Blog
量子位
AWS News Blog
AWS News Blog
Project Zero
Project Zero
P
Proofpoint News Feed
Cisco Talos Blog
Cisco Talos Blog
Spread Privacy
Spread Privacy
S
Schneier on Security

OpenAI Developers

API deployment checklist | OpenAI API Sora 2 Prompting Guide Codex Prompting Guide Docs MCP | OpenAI Developers Gpt-image-1.5 Prompting Guide GPT-5.2 Prompting Guide Transcribing User Audio with a Separate Realtime Request Modernizing your Codebase with Codex GitHub - openai/openai-sora-sample-app: Sample app to get started using the Video API with Sora GitHub - openai/openai-apps-sdk-examples: Example apps for the Apps SDK GitHub - openai/openai-chatkit-advanced-samples: Starter app to build with OpenAI ChatKit SDK GitHub - openai/openai-chatkit-starter-app: Starter app to build with OpenAI ChatKit + Agent Builder Rate limits | OpenAI API Web search | OpenAI API Getting started with datasets | OpenAI API Prompt optimizer | OpenAI API Verifying gpt-oss implementations How to run gpt-oss locally with LM Studio Fine-tuning with gpt-oss and Hugging Face Transformers How to run gpt-oss locally with Ollama Function calling | OpenAI API Models | OpenAI API Reasoning best practices | OpenAI API Reasoning models | OpenAI API Background mode | OpenAI API Batch API | OpenAI API Conversation state | OpenAI API File search | OpenAI API Flex processing | OpenAI API MCP and Connectors | OpenAI API Code Interpreter | OpenAI API Quickstart - OpenAI Agents SDK Build Hour: Agentic Tool Calling Build Hour: Built-In Tools Reasoning best practices | OpenAI API Graders | OpenAI API Evaluation best practices | OpenAI API Working with evals | OpenAI API Guardrails - OpenAI Agents SDK Latency optimization | OpenAI API Optimizing LLM Accuracy | OpenAI API Agent orchestration - OpenAI Agents SDK Production best practices | OpenAI API Realtime transcription | OpenAI API Optimizing LLM Accuracy | OpenAI API Realtime and audio | OpenAI API Realtime conversations | OpenAI API Responses guide Migrate to the Responses API | OpenAI API Speech to text | OpenAI API Supervised fine-tuning | OpenAI API Tracing - OpenAI Agents SDK Vision fine-tuning | OpenAI API Audio and speech | OpenAI API GitHub - openai/openai-cs-agents-demo: Demo of a customer service use case implemented with the OpenAI Agents SDK Voice agents | OpenAI API Fine-tuning best practices | OpenAI API GitHub - openai/openai-agents-python: A lightweight, powerful framework for multi-agent workflows GitHub - openai/openai-agents-js: A lightweight, powerful framework for multi-agent workflows and voice agents Agents SDK | OpenAI API Using tools | OpenAI API Computer use | OpenAI API GitHub - openai/openai-cua-sample-app: Learn how to use CUA (our Computer Using Agent) via the API on multiple computer environments. GitHub - openai/openai-testing-agent-demo: Demo of a UI testing agent using the OpenAI CUA model and the Responses API. Model optimization | OpenAI API GitHub - openai/openai-fm: Code for openai.fm, a demo for the OpenAI Speech API Predicted Outputs | OpenAI API GitHub - openai/openai-realtime-console: React app for inspecting, building and debugging with the Realtime API Building Voice Agents GitHub - openai/openai-realtime-twilio-demo Reinforcement fine-tuning | OpenAI API GitHub - openai/openai-responses-starter-app: Starter app to build with the OpenAI Responses API Structured model outputs | OpenAI API GitHub - openai/openai-structured-outputs-samples: Sample apps to help developers get started with Structured Outputs Voice agents | OpenAI API Model optimization | OpenAI API GitHub - openai/openai-realtime-agents: This is a simple demonstration of more advanced, agentic patterns built on top of the Realtime API. GitHub - openai/openai-support-agent-demo: Demo of a customer support agent interface using NextJS and the OpenAI Responses API with File Search Building Voice Agents Generate images with high input fidelity AI app development: Concept to production Model optimization Building agents Eval Driven System Design - From Prototype to Production Multi-Agent Portfolio Collaboration with OpenAI Agents SDK o3/o4-mini Function Calling Guide Exploring Model Graders for Reinforcement Fine-Tuning Guide to Using the Responses API Reinforcement Fine-Tuning for Conversational Reasoning with the OpenAI API Evals API Use-case - Responses Evaluation Comparing Speech-to-Text Methods with the OpenAI API Generate images with GPT Image Multi-Tool Orchestration with RAG approach using OpenAI Multi-Language One-Way Translation with the Realtime API Doing RAG on PDFs using File Search in the Responses API How to use the Usage API and Cost API to monitor your OpenAI usage Leveraging model distillation to fine-tune a model Orchestrating Agents: Routines and Handoffs Prompt Caching 101 Developing Hallucination Guardrails
GitHub - openai/openai-realtime-solar-system: Demo showing how to use the OpenAI Realtime API to navigate a 3D scene via tool calling
2025-07-18 · via OpenAI Developers

MIT License NextJS OpenAI API

This demo showcases how to use the OpenAI Realtime API to interact through voice with a 3D scene (built with Spline), in this case a scene representing the solar system.

It is implemented using the Realtime + WebRTC integration and uses function calling to trigger actions in the app.

screenshot

How to use

Running the application

  1. Set up the OpenAI API:

  2. Clone the Repository:

    git clone https://github.com/openai/openai-realtime-solar-system.git
  3. Set your API key:

    2 options:

    • Set the OPENAI_API_KEY environment variable globally in your system
    • Set the OPENAI_API_KEY environment variable in the project: Create a .env file at the root of the project and add the following line:
      OPENAI_API_KEY=<your_api_key>
  4. Install dependencies:

    Navigate to the project directory and run:

  5. Run the app:

    The app will be available at http://localhost:3000.

Note: the 1st time you load the app, the scene can take a while to load (it's heavy!). The subsequent loads should be faster as it will be cached.

Starting a session

To start a new Realtime session, wait for a few seconds for the page to load properly, thenclick on the wifi icon in the top right corner of the app. Once it turns green, you can start talking to the model.

You can use the mic icon right next to it to toggle the microphone on and off. When you start the session, it will automatically turn on, but you can turn it off to mute yourself while the session is running.

Toggling the wifi icon will stop the session, and the conversation will be reset.

Demo flow

Make sure there is no background noise or echo when you talk to the model, as this may cause interruptions.

The demo is configured with instructions prompting the model to answer any question about the solar system. Additionally, the model has access to multiple tools that map to actions in the app or animations in the spline scene.

Configured interactions

More specifically, the following interactions trigger animations:

🪐 Asking about a specific planet will trigger a visual focus on that planet if the animation was set up in the spline scene (works with the Sun, Earth, Mercury, Mars, Jupiter, Saturn, Neptune, Pluto). These animations can also be triggered by clicking on the planets in the UI.

🌕 Asking about moons will trigger the moons to appear if they have been set up in the spline scene (works with Pluto's moons and Jupiter's galilean moons).

📊 Asking about data that can be represented with a chart will result in a chart being displayed in the UI (bar chart or pie chart).

🛰️ Asking about the position of the ISS will result in the ISS position being fetched and the ISS being displayed in the spline scene with the corresponding animation.

👋 Saying something like "thank you, I'm good" or anything to close the conversation will reset the camera to the initial position. This animation can also be triggered by hitting the space bar.

🌌 Asking about the planets' orbits will result in a camera change to see the solar system from above. This animation can also be triggered by hitting the M key to change to orbit view, and pressing the Enter key to change back to main view.

For more details about tools used by the model, see the lib/config.ts file.

Example flow

Here is an example flow that showcases the different interactions:

  1. Say something like "I'm curious about Earth" to focus on Earth
  2. Ask for the distribution of land vs water - a pie chart should appear, if not prompt the model to show it to you
  3. Say something like "I have a question about Mars now" to focus on Mars
  4. Ask for the highest volcano and how it compares to Mount Everest - a bar chart should appear, if not prompt the model to show it to you
  5. Say something like "thank you, I'm good" to reset the camera
  6. Ask where the ISS is - the model should reply with the position and the ISS should appear in the scene
  7. Say that you'd like to see Pluto now to focus on Pluto
  8. Ask about its moons - 5 moons will pop up
  9. (optional): Do the same with Jupiter and ask about Galilean moons - 4 moons will pop up
  10. Ask something related to the position of the planets in the solar system, for example "how are the planets positioned in their orbits" - The view will change to a high level view with orbits

Customization

This demo is just an example of how to use Function Calling with the Realtime API to trigger actions in an application, including sending events to a spline scene.

You can read more about how to build your own scene in the Spline documentation. You can then change the scene url in the components/scene.tsx file.

<Spline
  scene="https://prod.spline.design/<scene_id>/scene.splinecode"
  onLoad="{onLoad}"
/>

If you want to use your own scene, make sure to configure trigger events in the spline scene, and update the code to trigger the events in the components/scene.tsx file.

For example, you can add to any object in your scene a mouseDown event that will trigger an animation. You can then trigger this event in the components/scene.tsx file by calling spline.current.emitEvent("mouseDown", "object_name").

You can also update:

  • The instructions in the lib/config.ts file to change the behavior of the model
  • The tools the model has access to in the lib/config.ts file
  • The voice in the lib/config.ts file

License

This project is licensed under the MIT License. See the LICENSE file for details.