---

name: "python-client" description: "Routes UnrealCV Python client tasks for connecting to servers, sending commands, decoding responses, and launching local binaries." disable-model-invocation: true metadata: disco-role: operating license: MIT

Python Client

Use this sub-skill for anything that talks to a running UnrealCV server through the Python package. The runtime helpers are bundled here, but this route still operates against a live server or local dummy server rather than a packaged source snapshot.

Read first

• references/api-reference.md for the verified public API surface, command families, and method groups.
• references/workflows.md for quick-start and live-server recipes.
• references/troubleshooting.md for connection, decoding, version, and launcher failures.
• scripts/local_client_smoke.py for a safe dummy-server smoke check.
• scripts/connect_and_request.py if you need a live-server request helper.
• references/api-reference.md for the public API surface and bundled install facts.

What this sub-skill covers

• Client, SocketMessage, and ApiVersionManager
• UnrealCv_API camera, object, scene, image, capture, recording, pak, and panoramic helpers
• MsgDecoder and utility functions such as read_png, read_npy, and parse_resolution
• Data models from unrealcv.models
• Launcher helpers (RunUnreal, RunDocker, UE4Binary) when the client workflow also starts a binary
• Optional command-capability checks and UnrealZoo-only helpers when the connected server supports them

Typical triggers

• "How do I connect to UnrealCV from Python?"
• "How do I request camera location, depth, or object data?"
• "How do I decode lit/depth/normal responses?"
• "How do I start a local binary or Dockerized Unreal environment and then connect?"
• "How do I handle command-capability warnings or version mismatches?"

What belongs elsewhere

• Plugin build, install, or packaging requests belong in ../plugin-build/.
• Refreshing the generated command schema or public API snapshot belongs in ../maintenance/.
• Do not send future agents back to the original repo examples or tests for runtime use; use the bundled references and scripts instead.

Usage pattern

1. Confirm the package is installed with pip install unrealcv or by using the bundled package snapshot under ../../references/unrealcv-source/client/python when you want a self-contained inspection target.
2. Connect with Client for low-level transport or UnrealCv_API for high-level camera/object/image helpers.
3. Use ApiVersionManager when the task needs command discovery or version gating.
4. Use the bundled smoke helper before asking anyone to debug a live UnrealCV server.

Failure modes to keep in mind

• No running server or wrong endpoint.
• Server too old to report supported commands.
• Image/depth payloads that do not match the requested mode.
• UDS only available on Linux and only after the binary creates its socket.
• Live launch helpers failing because the binary path or unrealcv.ini state is wrong.