[beta] local read-only tool: which lines of your Python OpenAI client break when the API changes?

[sambanideepak]

I built a small Python-first, read-only scanner and would value maintainers' read on it.

In one command it takes a Python repo that calls a REST/OpenAPI API, plus the OLD and NEW API specs, and reports exactly which files / functions / lines will break — with a confidence score and the evidence behind it.

Why it may fit here: the OpenAI API changes very fast, and openai-python is a high-traffic client where an upstream breaking change can silently break callers.

It runs entirely on your machine: no source upload, no account, no telemetry. It's a beta; I'm not selling anything — I just want to know if it saves investigation time when the API changes.

If you're up for a ~20-minute test on a public copy of openai-python + two OpenAI OpenAPI specs (old/new), reply here and I'll share the package + a 2-minute quickstart (I'm not pushing the code publicly yet). Feedback goes back via a short form; stop anytime.

Thanks,
Deepak

---

UPDATE (self-serve): the code is now public – you can test it without waiting for me to share files. Clone and run in ~2 minutes, no account:

https://github.com/sambanideepak/upstream-sentry

It's a Python-first, read-only, local scanner (no source upload, no telemetry). Pick two OpenAPI specs (old/new) of the API you depend on and a copy of your client, then bash run_beta_scan.sh --repo ... --old-spec ... --new-spec ... --output ./out. Feedback still welcome via the short form in the repo. Still a beta; not selling anything.

UpstreamSentry — know what breaks before the API change reaches you.