Tonic/nutonic-terramind-tim

NU:TONIC — TerraMind TiM (Gradio ZeroGPU Space)

Published from inference/terramind_tim_local in the NU:TONIC monorepo.

CI deployment

This Space is deployed by .github/workflows/huggingface-deploy.yml using:

python tools/hf_deploy/deploy_space.py --service terramind_tim --repo-id Tonic/nutonic-terramind-tim

The workflow first runs python -m pytest inference/terramind_tim_local/tests -q, then mirrors the staged Gradio SDK Space tree and syncs runtime settings from tools/hf_deploy/profiles/terramind_tim.yaml.

Hardware

This service is deployed as sdk: gradio because Hugging Face ZeroGPU only supports Gradio SDK Spaces. Use ZeroGPU so TerraTorch forwards have CUDA. CPU-only is suitable only for tiny smoke configs outside the ZeroGPU deployment.

Endpoints

• —Gradio API api_name="health" — patch and service diagnostics.
• —Gradio API api_name="tim_infer" — JSON body: { "config": { ... } } or { "config_yaml": "<yaml>" } matching the CLI YAML schema.

The package still exposes FastAPI routes when run locally or as a Docker Space, but the production ZeroGPU deployment uses the Gradio SDK app.

Environment variables