har

Topics in the source

• HAR analysis
• When to load
• When there is no capture yet
• The two-command answer
• The scripts
• When a search finds nothing
• References

CANARY: har-loaded-2026-08-30-7f3a91c4-2b6d4e8a1c9f0b73 A browser capture holds every request, header and response body of a session, in 5–500 MB of JSON. The few hundred bytes that matter — which endpoint returns the data, which parameter pages it, which header authenticates it — are reached through these scripts rather than by reading the file. A .har file has to be searched, summarised, filtered, reduced or made safe to commit. The question is which request returned a string seen on a page. A scraper is being built from a captured session. There is no capture yet and nobody at a browser to take one. A JSON or HTML file that did not come out of a capture needs none of this. It records into a fresh browser context, so no operator cookie or login ends up in the file, and writes the HAR even when navigation times out. --headed to watch, --wait S for late XHR. Read references/capturing.md before changing any of that. Paste a string that was visible on the page; get back the request that returned it, then its shape: --schema prints keys, types and array lengths, often 100× smaller than the body. On an unfamiliar capture, run analyzehar.py --input capture.har first: its XHR/JSON count

This is an excerpt. Refer to the complete source and its current requirements.