README.md

metadata

title: Sufficient Resolution Gate
emoji: 🖼️
colorFrom: indigo
colorTo: purple
sdk: gradio
sdk_version: 6.2.0
app_file: app.py
pinned: false

Sufficient Resolution Gate → Downstream VLM

This Space runs a small "resolution gate" model (LoRA adapter) that predicts a sufficient image resolution (max side). It then resizes the uploaded image once and sends it to a selected downstream VLM endpoint.

Secrets / Env vars

Set these in Settings → Variables and secrets:

• RITS_API_KEY (required if you use the RITS downstream models)
• HF_TOKEN (optional; only needed if your adapter repo is private)
• GATE_ADAPTER_REPO (optional; default: Kimhi/granite-docling-res-gate-lora)
• GATE_INPUT_MAX_SIDE (optional; default: 256)

If you don't have a RITS_API_KEY, pick "Skip downstream VLM (gate only)" in the dropdown.