File size: 2,114 Bytes
38979e8
db6c943
 
 
 
38979e8
db6c943
38979e8
 
 
 
db6c943
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
---
title: SupraChat
emoji: 🤖
colorFrom: blue
colorTo: gray
sdk: gradio
sdk_version: 4.0.0
app_file: app.py
pinned: false
---

# SupraChat – ChatGPT‑like Interface for SupraLabs Models

A clean, modern chat interface built with Gradio, supporting all text‑based models from [SupraLabs](https://huggingface.co/SupraLabs).  
Conversation history is stored temporarily in RAM and is cleared when you leave the page.

## Supported Models

- **Supra-50M-Instruct** – 50M instruction‑tuned model for general dialogue.
- **Supra-50M-Reasoning** – 50M reasoning model that outputs a thought process and a solution.
- **Supra-1.5-50M-Instruct-exp** – Experimental 50M model with 5K context length.
- **Supra-50M-Base** – Pure base model (no instruction tuning).
- **StorySupra-10M** – 10M model specialised for story generation.
- **Supra-Mini-v5-8M** – Ultra‑small 8M model for fast experimentation.

## Features

- 💬 ChatGPT‑style conversation interface
- 🔄 Switch between any supported model on the fly
- 📝 Automatic conversation title generation using `Supra-Title-Flan-85M`
- 💾 History stored in memory (cleared on page close)
- 🎨 Clean, responsive UI with adjustable temperature and token length

## Usage

1. Select a model from the dropdown.
2. Adjust temperature (creativity) and maximum output length.
3. Type a message and press Enter or click Send.
4. The title is auto‑generated from the first message.
5. To start a new conversation, click “Clear Chat”.

## Technical Details

- The app loads models on demand and caches them for subsequent requests.
- For reasoning models, the prompt is automatically prefixed with `<|begin_of_thought|>` to trigger the thought process.
- All dependencies are listed in `requirements.txt`.

## Deployment on Hugging Face Spaces

1. Create a new Space with Gradio SDK.
2. Upload `app.py`, `requirements.txt`, and this `README.md`.
3. The Space will install dependencies and launch the app.

## License

This project is for demonstration purposes. The underlying models are released by SupraLabs under their respective licenses.