Skip to main content
Each workflow is a standalone uv project to download. The source and assets ship with the SDK, so you do not need repository access. Each app defines its images, compute, and storage mounts in Python. Running or deploying the app applies those definitions, so mounted volumes need no separate CLI step to create them. Examples that need credentials include a Python setup step for them.

OpenAI-compatible LLM service

Serve a language model behind an authenticated OpenAI-compatible API on a GPU.

Train a YOLO object detector

Train on a GPU, keep the checkpoint in a volume, and predict from a second function.

Document processing with FastAPI

Upload a PDF or image in your browser and download the extracted text.

Sandboxed coding agent

Keep model credentials in a function while generated code runs in a network-blocked sandbox.

Parallel Parquet processing on S3

Fan partition work out through a function and write the summary back to your bucket.

​
Set up the examples

Install the CLI with uv, then choose a project:
Each download includes a README and dependencies pinned to the installed SDK. Run the guide’s commands from the downloaded directory, and use uv run for Python scripts and CLI workflows. Commit the generated uv.lock and use uv sync --locked in CI. GPU libraries install in remote images, so you do not need a local GPU.
If you use several workspaces, select one with uv run lazycloud workspace use <name>. Before running a GPU example, check your credit balance and GPU access in Settings, then Billing. The storage examples keep their files after a run so you can reuse the outputs. Follow each guide’s cleanup steps when you finish. The catalog also includes quickstart, artifacts, and all-workloads. Their READMEs cover run commands and cleanup. Use lazycloud example download all to write every project into its own directory.