LLM PlayGround
A playground for AI models from providers and routers such as OpenRouter and Requesty. Pick a model, send it something, and see exactly what comes back. It also has a few small tools and games built on those models.
- Tools: draw the last scene of a DreamGen story, turn it into a short video clip, or make videos from your own prompts.
- Games: AI Group Chat, where several AI characters, each on its own model, talk with you and with each other.
- API tester: one tab per kind of request (chat, images, video, speech, transcription, embeddings and more), showing the answer, the raw response and a ready-to-run
curl command.
Try it now at https://hkarn.github.io/LLM-PlayGround/. Nothing to install.
1. Get an API key
The page sends your requests to an AI provider with your own key, so you need one. The Welcome page, where the page opens (click the logo at the top left to get back to it), says how to start and, further down, compares the providers and links to each sign-up page. Video works with OpenRouter, and with LLM Gateway when you run the page on your own server.
The easiest is OpenRouter: sign in (Google works), add a little credit, create a key at openrouter.ai/keys, and every model is available, including some free ones.
Paste the key into the connection panel at the top of the page (click the chip that says Add API key) and press Load models.
2. Run it on your own computer (optional)
The online page works on its own. Running it yourself adds a small local server that can hold your keys, so you don't have to paste them, and that can reach providers a browser page can't call directly (such as LLM Gateway) or a model running on your own machine (such as Ollama).
With Docker (easiest)
You need Docker Desktop (or Docker on Linux).
-
Download this project: Code → Download ZIP on GitHub and unzip it, or git clone https://github.com/hkarn/LLM-PlayGround.git.
-
Open a terminal in the project folder and start it:
docker compose up -d
-
Open http://localhost:3000.
To stop it later, run docker compose down.
With Node.js
You need Node.js 18 or later. There is nothing else to install.
-
Download the project as above.
-
In the project folder, run:
node server.js
-
Open http://127.0.0.1:3000. Press Ctrl+C in the terminal to stop it.
Keep your keys on your computer (optional)
So you don't have to paste a key each time:
- Copy
.env.example to a new file called .env in the same folder.
- Fill in the keys you have, for example
OPENROUTER_API_KEY=sk-or-....
- Start the playground again. Both Docker and
node server.js read .env.
A key typed in the page always wins over the one in .env. Only do this on your own computer: anyone who can open the page could use these keys.
What's in it
| Where | What |
|---|
| Welcome (the logo) | What the site is, how to start, links to every tool, and for each provider how hard it is to set up, the steps and its sign-up link (click the logo to get back here). The sun, moon or half circle at the top right switches between light, dark and your system's setting |
| Tools | Draw the DreamGen Scene · Animate the DreamGen Scene · Video Generator. Click Tools (or Games) in the top bar to go to the one you used last; while you are in it, that click opens a menu of them; inside one, its title is a drop-down to jump to another, and ← goes back to the list |
| Games | AI Group Chat |
| API tester ▸ | Chat, Decision, Embeddings, Rerank, Image, Image edit, Video, Speech, Transcriptions |
| Filter models | A model filter that narrows every model list on the page, and the full model list |
Every tab and tool has its own link, such as #tools/dreamgen-scene or #chat, so you can bookmark it or share it.
Privacy
Nothing is sent to us or to any third party except your own requests, with your own key, to the provider you choose. Your settings, and your key if you tick Remember, stay in your browser. The local server only passes your requests on and logs no content.
More
User guides, step by step:
Reference (all of it in docs/):
- Providers: every supported provider, its URLs, keys and quirks.
- Running and hosting: the two ways it runs, all settings, Docker, hosting it publicly, GitHub Pages, and how the server works.
- Code: how the code is laid out, for anyone changing it.
Planned: more providers as presets (such as OpenAI, Anthropic, Google and a local Ollama), and more games and tools.
It grew out of requesty-tester. Ideas and problems are welcome as GitHub issues.
