🚀

Agentic Space Factory

Build, inspect, validate Gradio Spaces from model cards.

HF Auth: unknown Sign in
Run storage
loading…

Checking private bucket access…

Checking
⌘

New Build

Model pre-scanNot scanned yet
Required

Run a fast metadata scan before launching a paid Job. It checks the model card guidance, runnable examples, Diffusers signals, files, safetensors, gated access and custom-code risk without downloading weights.

Strict inference requires a real working inference endpoint. The run fails if the generated Space only boots without producing a valid output.

Checking your run bucket before enabling builds…

â–·

Active Run

Idle
Run ID—
Target Space—
Model—
Build goal—
Target GPU—
Fallback GPU—
Expected output—
Pi assistant—
Elapsed time—
Space ↗ Settings ↗ Job ↗ Artifacts ↗
Overall progress0%

Latest events

Typed activity
No activity yet.

Inference latency

measured when smoke test passes
Observed—
ZeroGPU duration—

When available, the live smoke-test latency is used to recommend the `@spaces.GPU(duration=...)` value for ZeroGPU apps.

Run traces

Document dock · source files
Source traces and documents will appear as files are written to the bucket.
☰

Runs Explorer

⌕
Loading runs

Run Stats

No runs
Run stats will appear after the first refresh.
¢

Billing & compute

Check billing
This can cost money.

Builds launch Hugging Face Jobs. Pi uses Hugging Face Inference Providers for Qwen and other selected assistant models, and usage may consume paid inference credits. Before running builds, open your Hugging Face Billing dashboard and make sure payment, quotas and spend limits are okay.

Open billing dashboard ↗