README.md

---
title: ASE GLM
emoji: ⬛
colorFrom: yellow
colorTo: gray
sdk: docker
app_port: 7860
pinned: false
hf_oauth: true
# `inference-api` is the whole point: it is what lets the Space call the router
# with the signed-in viewer's token, so their account is billed and not ours.
hf_oauth_scopes:
  - inference-api
hf_oauth_expiration_minutes: 480
short_description: A minimal code harness for GLM-5.3, on a character grid.
---

# ASE · GLM

A minimal Hugging Face code harness. The whole interface is one character
grid rendered to canvas ([gutbrains](https://github.com/Bit-Trading-Company/gutbrains)),
and the backend is FastAPI proxying the Hugging Face inference router.

Models: **GLM-5.3**, GLM-5.3 Flash, DeepSeek V4 Pro, DeepSeek V4 Flash.

## Try it without an account

Press **see what it does**. It draws charts, a heat map, an animation and a
playable game, entirely in your browser — no model is called and nothing is
spent. It is the same tool the model uses when you ask it for a graph.

## Billing

Inference is billed to **you**, the viewer — via your Hugging Face sign-in, or
a key you supply. This Space holds no inference token of its own, so a visitor
cannot spend the owner's credits: without a sign-in there is simply nothing to
spend. Signing in grants this Space the `inference-api` scope, and the token is
kept in a signed cookie on the server and used only to call the router on your
behalf.

Requests are rate-limited per viewer to keep one free CPU responsive for
everyone. That limit guards the Space, not a wallet.