| --- |
| title: English Learning Tinykan |
| emoji: 🎓 |
| colorFrom: blue |
| colorTo: indigo |
| sdk: gradio |
| app_file: main.py |
| pinned: false |
| license: mit |
| --- |
| |
| # Multi-Agent KAN System for Adaptive English Learning |
|
|
| Three-agent architecture (Evaluator · Teacher · Planner) for CEFR-aligned |
| English language learning, built on a distilled Kolmogorov–Arnold Network |
| (TinyKAN-Distilled, 32,080 parameters) coupled with Gemini 2.5 Flash. |
|
|
| ## Required Space secret |
|
|
| Set this under **Settings → Variables and secrets** before the first run: |
|
|
| | Name | Value | |
| |---|---| |
| | `GEMINI_API_KEY` | your Gemini API key from https://aistudio.google.com | |
|
|
| ## Required files (upload alongside these) |
|
|
| - `tinykan_cosine.pth` — trained model weights |
|
|
