The original author of this repository, PianothShaveck, decided to stop working on it, especially to not be affiliated anymore with owner of the API this repository used.
This repository contains the source code for LLM Playground, a web-based interface for testing various Large Language Models (LLMs).
- index.html: Main HTML file structuring the webpage.
- s.js: JavaScript file handling chat interface functionality.
- s.css: CSS file styling the webpage.
- Images:
discord.png
,light.png
,favicon.png
.
LLM Playground offers users an intuitive interface to interact with a variety of LLMs. Supports streaming. It's designed with a focus on a pure JavaScript and CSS implementation, minimizing reliance on external libraries for a lightweight and efficient experience.
Key features include:
- Model Selection: Choose from popular LLMs and custom endpoints.
- Streaming Support: Interact with models that stream responses in real-time.
- File Processing: Attach and process various file types, including spreadsheets, documents, PDFs, PowerPoints, EPUBs, RTFs, and plain text files.
- Chat Management: Maintain, export, and import chat histories. Search through past conversations.
- Message Control: Edit, delete, copy, and change the role of individual messages.
- Audio Support: Record and transcribe audio messages (requires Groq API key).
- Theming: Toggle between light and dark themes.
- System Prompts: Save, load, and customize system prompts.
- Web Search: Enable, disable, or automate web searches to enhance LLM responses.
- Parameter Control: Adjust max tokens, temperature, top_p.
This project is built with pure JavaScript, with minimal external libraries used only for parsing specific file types. This approach ensures a lightweight, fast-loading application with no unnecessary dependencies.
Enjoying LLM Playground? Consider supporting its development!
For a detailed project roadmap and a list of known issues, please refer to the TODO.md file.
- Clone the repository.
- Open
index.html
in a web browser.
No additional installation is required.
This project is licensed under the MIT License.