Traffic

Traffic is a modal HTTP traffic window. It opens with the traffic button on the ZP7 page, the /?page=zp7#traffic link and the Alt+5 hotkey.

Nodes and the local runtime write traffic to a file, and the window reads its tail — the same way allLogs reads logs.

Sources#

The window polls two kinds of source at once and merges the records into one list:

Node addresses come from GET /zp/nodes?probe=false. An unreachable node does not get in the way of the others: its error is shown as a line above the list, and records from the reachable sources are displayed as usual.

Each side keeps one file per machine, written under a named mutex and rotated at 50 MB with the five most recent files kept. Reading goes from the end: the last N records are requested, and rotated files are pulled in when the current one is not enough. The scan ceiling is 64 MB; once it is used up the answer is marked truncated.

List#

Filters: machine, project, method, status, a URL substring, and the number of records per source. The project goes into the request and is filtered while the file is read; the remaining filters apply to the records already loaded. The window refreshes every 3 seconds while open, and the Auto button stops that. Clear view resets the window buffer — the files themselves are untouched.

Details#

The panel shows the URL, method, status, duration, machine, project, account, session, proxy and task_id, plus collapsible Request and Response sections with headers, cookies and body. Clicking a value copies it.

Replay#

Play opens the request in an editor: method, URL, headers and body can be changed before sending. The request goes through a separate replay listener — POST /http-replay, the same one HAR uses. The response can be copied, and there is a cURL button next to it built from the current field contents.

The listener port is always the dashboard port + 1, both on the server and on the front end; there is no setting for it. It listens on localhost only.

Code#

For a selected record, ready-made calls are generated: HttpClient, ZP7, Hybrid, Python, TypeScript, cURL. API Skeleton and API Example fold every displayed request into an endpoint description — either field types or the last real values.

cURL Import parses a pasted curl command and opens it in the details panel as an ordinary record: it can be replayed and converted, and it does not appear in the list.

Record format#

{
  "timestamp": "2026-09-12 17:40:11.325",
  "startedDateTime": "2026-09-12T22:40:11.1000000Z",
  "method": "POST",
  "url": "https://example.com/api",
  "statusCode": 200,
  "durationMs": 214,
  "proxy": "",
  "request":  { "headers": [], "contentType": "", "userAgent": "", "cookies": "", "cookiesSource": "", "body": "" },
  "response": { "headers": "", "body": "" },
  "machine": "PC", "project": "", "account": "", "session": "",
  "task_id": "", "source": "z3n7"
}

The format is shared by the node and the runtime, so the window parses both sources with the same code.

API#

• GET /traffic?tail=N&project=&task_id= — the local z3nDash file;
• GET /zp/traffic?machine=&tail=N&project=&task_id= — forwarded to a node;
• GET /zp/nodes?probe=false — the node list;
• POST /http-replay — replaying a request (dashboard port + 1).

Node requirement#

The tail of the file is served by the tail parameter of GET /traffic on the z3n7 side. If a node returns the first N records instead of the last ones, it is running a z3n7.dll build without tail support.