Graham Dumpleton Introduces Wrapture Library

Summary

Graham Dumpleton (wrapt, mod_wsgi) introduced Wrapture, a young Python library. The library combines runtime instrumentation of uncontrolled code, including monkeypatch-style testing (an alternative to unittest.mock) and a configuration-file-based tracing mode with OpenTelemetry support. Simon Willison noted that the project is promising while explicitly emphasizing that it is only a few weeks old.

Why it matters

Observing code you don't own without disrupting it has long been a persistent issue in Python. Wrapture unifies these two workflows—creating stubs in tests and tracing in production—into a single API. This could reduce the tooling complexity surrounding the combination of unittest.mock and a separate APM agent. Configuration-driven tracing (a TOML-like block pointing to a target and a JSONL sink) offers a low-friction starting point for teams who want to add OpenTelemetry spans without writing instrumentation code.

Highlights

• A single API covers both test mocking (returns / transforms_result) and live tracing, replacing the need for a mock plus APM agent combination.
• Configuration-based tracing allows you to add spans to an existing project with a few lines of declarative configuration instead of code changes.
• Dumpleton stated that every line was generated by AI under his own direction. He clearly distinguishes this from 'vibe coding' and positions the project as his first fully agent-directed project.

Critical take

The library was released weeks ago. There is no evidence of third-party adoption, benchmarks, or edge-case stress tests. Therefore, calling it a 'promising start' may be generous. Additionally, while the AI authorship claim is transparent, it raises the question of how much independent review the library has undergone before being recommended as a direct replacement for unittest.mock in production test suites.