File size: 1,591 Bytes
d9faf63 7be0622 d9faf63 7be0622 d9faf63 7be0622 |
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 |
---
title: Smol Private Spreadsheet
emoji: 💻
colorFrom: blue
colorTo: yellow
sdk: docker
pinned: false
short_description: A spreadsheet powered by smolLM2 using WebGPU
tags:
- custom-components-for-agents-track # 🧩 Track 2
- track2
- gradio
- smolllm
- spreadsheet
---
# Smol-Private-Spreadsheet ✨
A tiny proof-of-concept that embeds **SmolLM 2** directly inside a sheet.
Type a formula, get an answer—no extra UI required.
## Why it matters
| Feature | What it does |
|---------|--------------|
| **`=SMOLLM("prompt")`** | Asks the model and returns the reply inside the cell. |
| **Context support** | `=SMOLLM("prompt", A1)` or a range `A1:C3` to pass extra context. |
| **GPU / WebGPU ready** | Runs locally inside the Space; zero external API calls. |
| **First Step Towards Creating a WebGPU custom Gradio Component** |I'm hoping that this space can showcase the power of UI abstraction when combined with SmolLM2 and allow me to reach mentors that are willing to teach me how to create a custom gradio component that allows the use of small models using WebGPU. |
| **Next Steps** | I'm still trying to create one Gradio app that allows me to do both, run a model locally using WebGPU and use MCP tools and other models using inference to bring the best of both worlds. It's a very challenging journey for me, especially given the limitation that all my ideas are competing for the same port at the same time: 7860. |
## Quick start
1. **Launch** the Space.
2. In any empty cell, enter:
```text
=SMOLLM("Summarise A1:C3 in one sentence", A1:C3) |