https://github.com/explor4268/this-webserver-does-not-exist
A web server where its responses are entirely generated by an LLM based on the request path.
https://github.com/explor4268/this-webserver-does-not-exist
Last synced: 8 months ago
JSON representation
A web server where its responses are entirely generated by an LLM based on the request path.
- Host: GitHub
- URL: https://github.com/explor4268/this-webserver-does-not-exist
- Owner: explor4268
- License: mit
- Created: 2025-09-28T13:19:06.000Z (9 months ago)
- Default Branch: main
- Last Pushed: 2025-10-05T09:21:15.000Z (8 months ago)
- Last Synced: 2025-10-05T11:26:36.336Z (8 months ago)
- Language: JavaScript
- Size: 11.7 KB
- Stars: 2
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# This web server does not exist
A web server where its responses are entirely generated by an LLM based on the request path.
> [!CAUTION]
> **This is only intended to be a fun experiment. Do not actually use this concept in production environment. It's very fragile, vulnerable, and inefficient due to the nature of Generative AI.**
## Demo Video
https://github.com/user-attachments/assets/8c939363-9ce2-4341-9020-58a2f2bc21ab
## Usage
No live demos provided at this time. You may test this out by following the instructions below to run it yourself.
1. Clone or download this repo
2. Ensure that you have already installed NodeJS **version 22 or newer** beforehand. **Older versions are not tested and not guaranteed to work.** The logic itself runs in a single `index.js` file that uses NodeJS's built-in helper libraries and some recent JavaScript features. Using this approach, this experiment doesn't require any external npm dependencies.
3. Ensure that `ollama` is set up. The script uses the `qwen3:1.7b` model by default (for quick performance when testing) and interacts with it via the Ollama API (for now). Both the model and the endpoint can be changed by editing `index.js` (inside the `getAPIOpt` function for the model and after a call of `getAPIOpt()` inside `createServer` for the latter). Other options such as the listening port can also be changed inside that file.
4. Start the Ollama server by running `ollama serve` and pull some models if you haven't done one or both of it already.
5. Run the server by executing `node .` (or `node index.js` to be specific, both does the same thing)
6. The script should give you a port an a URL. Test it on your web browser. Example query: . The path can be any sensible path that the model can understand, as long as the file type is supported (and not a binary file).
## Related links
Other cool projects and experiments as my inspiration to create this experiment:
- [NeuralOS](https://news.ycombinator.com/item?id=44564531)
- [Ch.at – A lightweight LLM chat service accessible through HTTP, SSH, DNS and API](https://news.ycombinator.com/item?id=44849129)
- [Simulating a Neural Operating System with Gemini 2.5 Flash Lite - Google Developers Blog](https://developers.googleblog.com/en/simulating-a-neural-operating-system-with-gemini-2-5-flash-lite/)
- [Show HN: This Website Does Not Exist](https://news.ycombinator.com/item?id=44128441)
- [Building a Virtual Machine Inside ChatGPT](https://news.ycombinator.com/item?id=33847479)
- [Show HN: A store that generates products from anything you type in search](https://news.ycombinator.com/item?id=45231378)
- [Show HN: Vibe Linking](https://news.ycombinator.com/item?id=45362813)