HAR-X

A rewrite of the old capy/har.analyze.py pair into two small, coherent tools joined by one interface: the HAR file.

harx/
├── capture/     Go tool: pops up a real Chrome window, records every
│                network request during your session, writes a spec-
│                compliant .har file.
├── analyzer/    Python + Streamlit app: loads .har files (from capture/
│                or any browser export) and gives you an interactive
│                waterfall, filters, and summary stats.
└── captures/    Shared output folder. capture/ writes here by default;
                 analyzer/ scans it by default. (Two real-world sample
                 .har files are included so you can try the analyzer
                 immediately.)

Why the old version didn't work

The old capy tool did a single raw net/http.Get() and faked a one-entry HAR from it, plus an unrelated raw pcap dump. It never ran a browser, so it could never see JS-driven requests, CSS/images/XHR/fetch, or real per-phase timings (DNS/connect/TLS/send/wait/receive) — everything that makes a HAR a HAR. capture/ replaces that outright by driving a real Chrome instance over the DevTools Protocol (CDP) and recording its actual Network.* event stream, the same way DevTools' own "Network" tab and Lighthouse do it.

1. Capture

cd capture
go build -o harx-capture ./cmd/harx-capture
./harx-capture -out ../captures                 # opens a blank window
./harx-capture -url https://example.com -out ../captures -title example

A real, visible Chrome window opens. Browse normally — log in, click around, trigger whatever XHR/fetch/SPA traffic you care about. Close the window (or Ctrl+C the terminal) to stop recording; a .har file lands in captures/.

Flags:

• -url — optional starting page (default: opens a blank tab)
• -out — output directory (default ./captures)
• -profile — reuse a Chrome user-data-dir, so you don't have to log in every session
• -title — label included in the output filename

2. Analyze

cd analyzer
pip install -r requirements.txt
streamlit run app.py

Opens in your browser. Pick a capture from the sidebar (auto-discovered from ../captures/) or upload any .har file — this works on browser DevTools exports too, not just harx-capture output. You get:

• an interactive waterfall (stacked by timing phase: blocked/DNS/connect/ SSL/send/wait/receive)
• filters: status range, method, resource type, domain, content-type
• status code / content-type / size distributions
• a sortable request table
• per-request detail: headers, POST body, response body preview, full timing breakdown

Status

• Capture rewritten on chromedp/CDP — produces real, multi-entry, spec-compliant HAR with per-phase timings
• Analyzer rewritten in Streamlit — interactive replacement for har.analyze.py