Is it possible to get more details from repsonse time, like cdn lookup time, request connect time, response bytes read time etc...

[mer163]

Is it possible to get more details from repsonse time, like cdn lookup time, request connect time, response bytes read time etc...

[ketpatil77]

For transport phases, use the request's trace extension. A response hook alone does not separate connection setup from response-body reads. This helper records completed phases with a monotonic clock:

from time import perf_counter

def timed_get(client, url):
    starts, durations = {}, {}
    def trace(name, info):
        phase, _, state = name.rpartition(".")
        if state == "started":
            starts[phase] = perf_counter()
        elif state in ("complete", "failed") and phase in starts:
            durations.setdefault(phase, []).append(perf_counter() - starts.pop(phase))
    response = client.get(url, extensions={"trace": trace})
    return response, durations

Use an async def trace(name, info) callback with AsyncClient instead. The default non-streaming get consumes the body before returning; with streaming, keep the callback alive through body iteration.

These are implementation-level event names, such as connection.connect_tcp and http11.receive_response_body, rather than a stable set of browser waterfall fields. Pooled requests may have no connect event. DNS resolution is included in connection establishment; this does not expose a separate DNS duration, CDN-internal lookup, or server processing time. Concurrent/repeated events would need request-local correlation beyond this simple sequential GET helper.

Validated against a loopback HTTP server, observing TCP connect and response-body events. Trace extension.

Validation version: HTTPX 0.28.1, Python 3.11.9.