Skip to content
WeftKit

Reports & evidence

A report is the audit record of one run or one recording. It is stored locally, organised into seven tabs, and written to be defensible: it separates what was observed from what was verified, and states its own gaps.

The WeftKit run report summary tab showing outcome, steps, duration, AI cost and coverage-gap cards, a first material divergence banner, an execution contract table, and an evidence availability panel.
Summary: the verdict, the execution contract that pins how to read it, and evidence availability.

Summary

The summary opens with the verdict, the executed-step ratio, wall-clock duration, AI cost, and the number of coverage gaps. Below it sit two panels that make the result interpretable rather than merely colourful:

  • Execution contract — the pinned inputs used to interpret the result: report type, creation time, browser, environment, replay mode, suite revision, AI request count, and artifact size.
  • Evidence availability — which evidence channels are present, stated separately from integrity. Report integrity that has not been verified is labelled as a coverage gap instead of being omitted.
Availability answers “is this reference present?”, not “was every channel captured?”. The report says so in the panel itself rather than leaving you to assume.

The seven tabs

The tabs of a WeftKit run report
TabContents
SummaryVerdict, execution contract, first material divergence, evidence availability.
TimelineEvery executed step in order with its status, action type, and timing.
EvidenceThe screenshot gallery, linked to the step that produced each capture, with click, screenshot, and step counts.
NetworkRequest metadata — status, method, URL, resource type, duration. Sensitive URL values are redacted at capture.
ConsoleBrowser console output captured during the run, timestamped by level.
PrivacyWhat the run touched in privacy terms, including the count of secret references involved.
AI costThe provider transactions for this run — including the explicit zero-cost transaction when no provider was contacted.
The timeline tab of a WeftKit run report, listing each executed step in order with its status badge and action type.
Timeline: the executed order, not the authored order.
The evidence tab of a WeftKit run report showing click, screenshot and step counts above a gallery of step-linked screenshots.
Evidence: screenshots stay linked to the step that produced them.
The network tab of a WeftKit run report showing a table of requests with status, method, URL, type and duration, including a redacted search query and a 503 response.
Network: metadata only. Note the redacted query value and the 503 that caused the failure.

Redaction is applied at capture

Credential headers, sensitive URL values, secret-bearing fills, and sensitive screenshots are filtered before evidence enters a report. This is a collection boundary, not a display filter — there is no setting that turns it off, and no hidden copy of the unredacted value.

The AI cost tab of a WeftKit run report stating deterministic execution with zero AI cost and explaining that the zero-cost transaction is retained.
Even a run that never contacted a provider gets a transaction — that is what proves the deterministic-first path.

Exporting

Two exports sit in the report header. Copy JSON summary puts a machine-readable summary on the clipboard for a ticket or a CI comment. Open HTML report opens the standalone local report. Both are produced from the same stored record, so an exported artifact cannot disagree with the app.

LimitationRecording reports and run reports are different documents. A recording report describes what was demonstrated; a run report describes what was executed and asserted. The report list labels each one so they are never conflated.
Next: Providers & models