Fetch

Fetch Center is where you ask your data a question. You can type it, dictate it or attach files to it. It runs against the sources you select, and the answer comes back as Twin Insights. Open it with Fetch in the bottom navigation (/fetch).

The status line

The banner at the top says where you stand, and it offers the next step:

You see	Button
A fetch is running…
Checking your configured sources…
Could not load your data sources., with the transport error	Retry
No data source is configured yet.	Connect a data source
Sources are configured, but no fetch has run on this device yet.	Ask your first question
Your last fetch failed.	Open the run
Your last fetch is queued offline and has not run yet.	Open the run
Your last answer came back low-confidence.	Ask again
Your last answer was reused from cache, not fetched live.	Ask again
Your last fetch succeeded.	Ask a new question

Could not load your data sources. means the cockpit could not look. No data source is configured yet. means it looked and found no sources.

Run a fetch

### Choose your sources Under **ACTIVE CONNECTIONS**, tap a source card to include or exclude it. The caption reads **Included in fetch** or **Not included**. It describes your selection. It does not say whether the source is reachable. When the connection list loads live, every configured source starts out included. ### Ask your question Type in **Ask a question or use voice command…**, or tap a card under **FETCH STARTERS** to fill the box with a prepared question. To add input, use the buttons beside the box: - **Voice Input** dictates. Dictation stops after a pause and the fetch starts by itself. - **Camera Input** attaches a photo. - **Attach File** attaches one or more files. Attachments appear as chips that you can remove. ### Choose how to be told **Call Me When Finished** and **WhatsApp Me** ask FACE to notify you when the run ends. ### Run it Press Enter or select **Run Fetch**. A progress dialog, **Analyzing your data…**, lists each step as the server streams it. A finished step gets a check mark. The dialog shows **Ready** and closes when the run ends. ### Read the answer Under **Twin Insights**, the spoken summary comes first, followed by the cards. The summary is also read aloud.

A photo taken with Camera Input is also offered to the surveillance pipeline. The snackbar tells you which of three things happened: the photo was offered, the offer failed, or it was attached to the prompt only. The photo is listed under Captured media on Reconcile.

If the cockpit cannot reach the server, the run is queued on your device (Offline mode: Fetch task queued for background synchronization.). It runs when the connection returns.

Twin Insights

Card	Shows
Decision	The recommendation, a {percent}% confidence badge, and Why:, Plan: and Risks: lines.
Recommended Action	A priority badge and Impact:.
Forecast	Forecast:, and an ANOMALY badge where one was detected.
Signal	Its source and Relevance:.

Each proposed action is judged before it reaches you. A decision or action card that was judged carries a footer: Judgement: followed by concurs, dissents, unable to judge — not agreement or out of scope — not judged, then the judge’s reason. assessed by the model or checked without a model says how the verdict was reached. An action the judge dissented from is removed from the answer and listed under Actions rejected, with a count of how many actions it concurred with, dissented from, or could not judge.

Compute runners

COMPUTE RUNNERS lists the runners registered for the instance. Runs use the managed runner, which is marked Active. Other runners show Standby.

Execution history

EXECUTION HISTORY lists your runs, newest first. The automatic session fetch is listed here too. Each run card shows the command, the result or error, the duration, Runner: and Connections:, and a provenance badge. The badge reads Live, Cached, Offline — queued, Low confidence or Failed.

Reuse Prompt runs the same command again.
Delete Execution removes that one run, after you confirm Delete this run?.
Clear Log clears the whole history, after you confirm Clear Execution History.

Run history is stored on this device.

Below the runs are traces: the lineage datasets FACE has recorded, with the schema and how fresh each one is. Tap a trace to open its details: rows, size, freshness, columns, downstream datasets and the description.

Open a past run

Tap a run card to open Fetch details (/fetch/run/<id>). It shows:

the status: Completed, Failed or Queued offline;
Query, Runner, Connections, Started and Duration;
Error, when there was one;
Result;
What happened to the sources: notices about individual sources, word for word;
AI analysis;
Generated SQL.

Reuse Prompt in the title bar returns to Fetch Center and runs the command again. This run is no longer in your history. means the run was deleted or the history was cleared.

Schedules

Fetch Schedules, in the link row above the history, opens Active Schedules (/active_schedules). The subtitle counts the tasks, for example 2 OF 3 TASKS RUNNING.

The filter menu (Filter Tasks) offers All, Active only and Paused only.
Each card shows ACTIVE or PAUSED, FREQUENCY, NEXT RUN and LAST RUN.
The switch (Toggle schedule status) pauses or resumes the schedule.
Edit Schedule opens the schedule form.
Delete Schedule asks Delete schedule? before it deletes.
CREATE NEW SCHEDULE opens a blank form.

No schedules match this filter. means the list loaded and nothing matched. A failed load shows the error instead.

Create a schedule

### Open the form Select **CREATE NEW SCHEDULE**. The form is titled **SCHEDULE FETCH** — **Automate Data Discovery**. ### Pick the task Under **TASK CONFIGURATION**, choose a **FETCH TOPIC** from the fetch starters and a **RUNNER TYPE**. A scheduled run always executes on the managed runner. ### Pick the frequency Under **RECURRENCE LOGIC**, choose **DAILY**, **WEEKLY** or **MONTHLY**. **NEXT RUN** shows when it would next run. ### Save Select **SAVE SCHEDULE**. **Schedule Saved Successfully** closes the form. **Could not save the schedule. It was not created.** means nothing was stored.

Scheduled runs send no alerts. The form says so under NOTIFICATIONS. To be told when a run ends, turn on Call Me When Finished or WhatsApp Me for a run you start yourself.

Decision artifacts

A decision artifact is the full record of one proposed action, with the drafts it would produce: an email, a calendar invite, an ERP change. Open one from a Twins card:

View full decision artifact opens a preview, marked PENDING. Approve All approves the action.
Approving a card that has artifacts opens the outcome, marked QUEUED. The footer reads Action approved — queued for execution.

The page shows DECISION ARTIFACT #{id}, the title, the decision trail and the GENERATED DRAFTS. A link that carries no decision shows No decision artifact.

RPCs behind this page

FetchService.RunFetchStream (server stream), ListFetchStarters, ListTraces, GetTraceDetails.
ConfigService.ListConnections, ListRunners.
SchedulesService.ListSchedules, CreateSchedule, UpdateSchedule, SetScheduleActive, DeleteSchedule.
ModelService.SpeakFetchSummary, SynthesizeSpeech.
TwinsService.ExecuteAction, for Approve All.

See the API reference.

Navigation Reconcile