WSGI middleware¶

The same reporting as the ASGI middleware, for Flask and anything else that speaks WSGI. No framework dependency, no new packages.

Mount it¶

FlaskAny WSGI app

from queryspy.wsgi import QuerySpyMiddleware

app.wsgi_app = QuerySpyMiddleware(app.wsgi_app, budget=10)

from queryspy.wsgi import QuerySpyMiddleware

application = QuerySpyMiddleware(application, budget=10)

Options are identical to the ASGI middleware — budget, threshold, capture_stacks, add_headers, logger, on_report, panel, panel_path, history. The panel works the same way; see the ASGI guide.

When the report is emitted¶

This is the one thing that genuinely differs from ASGI.

A WSGI application returns an iterable, and for a streaming response the queries keep coming after __call__ has already returned. Finalising there would report a fraction of the truth, so queryspy keeps the recording window open and wraps the returned iterable: the report is emitted when that iterable is exhausted or closed, whichever happens first.

Two consequences worth knowing:

Counter headers reflect the moment the response started. They have to — headers go out before a streaming body runs. The log line is emitted at the end and is always the final figure.

In tests, consume the body. A real server always does; test clients are often lazy:

def test_projects(client):
    response = client.get("/projects")
    response.get_data()  # <- without this, the report has not landed yet
    assert response.headers["x-queryspy-queries"] == "2"

Concurrency¶

A threaded WSGI server gives each request its own thread, and a fresh thread gets a fresh context — so the recorder's context variable isolates requests exactly as it does for ASGI tasks. Interleaved requests never record each other's queries.

Which middleware do I want?¶

Stack	Use
FastAPI, Starlette, Litestar, Quart	`queryspy.asgi`
Flask, Pyramid, Bottle, Django (WSGI)	`queryspy.wsgi`

Neither is required — the pytest integration is independent of both and is where most of the value is.