Capture

Upload a workbook or PDF report

Microphone & voice capture

The Capture screen asks your browser for microphone access the first time you tap the orb to start a voice session. If your browser blocks or has previously denied that request, the error banner at the top of Capture will show the browser's own error message instead of starting — open your browser's site settings for this page, allow the microphone, then tap the orb again.

If the orb never leaves "Tap to start, then speak a reading," check that: a stage is selected first (the orb is inactive until you pick one), your microphone isn't muted or in use by another app or browser tab, and your network allows an outbound WebRTC connection.

From a spoken reading to Pending Review

While a Capture session is connected, a reading you speak (for example, a temperature for a field in your selected stage) is sent to the server and stored as pending — it is not yet part of the permanent record. It appears immediately on the Capture grid, and also on the Pending Review screen, which lists every pending reading from every worker and stage, not just your own session.

Editing and confirming a pending reading

On Pending Review, each pending reading has Edit and Confirm buttons. Edit opens an inline field where you can type a corrected value and Save; this re-validates the value against that field's standard but does not yet save it permanently. Confirm commits the reading as-is. Every reading is reviewed and confirmed one at a time — there is no bulk or "all at once" confirm action.

When a reading becomes a saved record

A reading becomes a permanent, saved record only once it is confirmed. From that point it appears on All Readings (every confirmed reading across every stage). If a later voice reading or a manual edit corrects an already-confirmed value, the saved record is updated in place and marked "corrected" — the correction itself, including the old and new values, is permanently recorded on the Audit Trail and can never be edited or deleted.

Navigating the app

• Capture — pick your assigned stage and run a live voice session; readings you speak appear here as you go.
• Pending Review — every reading recorded but not yet confirmed, from any worker or stage; edit or confirm each one individually.
• All Readings — every confirmed reading across every stage, read-only.
• Audit Trail — a complete, read-only, chronological history of every recorded, edited, confirmed, and corrected event. Nothing on this screen can be edited or deleted.
• Standards Management — a governance log where a candidate acceptance threshold can be proposed and approved/rejected for a field that is not yet confirmed. Approving a proposal here never enforces it — see that screen's own explanation.
• Excel Import — upload an .xlsx workbook and review how its column headers map to confirmed QC fields, flagging anything ambiguous, unrecognized, duplicated, or not yet supported. Once approved and linked to a matching stage on Capture, its existing values can be imported as pending readings.

Current limitations

• Importing a workbook's values is a separate, explicit step. Excel Import lets you upload a workbook and review/approve a column-to-field mapping — approving it only saves that review, it does not create any reading by itself. On the Capture screen, after linking an approved, stage-matching workbook, use Import readings to turn its existing cell values into pending readings; like any other reading, each one still needs to be reviewed and confirmed on Pending Review before it becomes part of the permanent record. Re-linking or re-importing an unchanged workbook never creates duplicates.
• Both a standalone download of your confirmed QC entries (All Readings → Download Excel) and an export of the original linked workbook with confirmed values filled in (Capture's Workbook panel) are available.
• All five stages — Depan, Tengah I, Tengah II, Packing I, and Vacuum — support real voice capture, editing, confirmation, and audit logging. What varies is per field, not per stage: some fields have a confirmed acceptance standard that is actively checked, some have a documented value still pending confirmation, and some have no numeric threshold defined at all — none of these is silently marked pass or fail. See Field Reference for the exact status of every field.
• Standards Management has no real user authentication. Its write actions (propose/approve/reject) are gated by a single shared operational passphrase configured on the server, not a personal login — there is no concept of a signed-in user or role anywhere in this application. Approving a proposal there is recorded in that log only; it never changes what Field Reference shows as confirmed or what the backend enforces, which still requires an engineer to update the backend schema directly.
• Text and count-style fields are recorded as numbers only in this build. A few fields on the paper forms are inherently non-numeric (for example a machine ID, a yes/no note, or a size category) — these are not yet captured by voice or the grid; only the numeric fields for each stage are implemented so far.
• The Audit Trail currently shows the latest events (up to 500 at a time) without a filter or search control in the interface, even though the server supports filtering by stage and event type.