Skip to content

Navigation

Every page after sign-in shares one frame: the rail on the left, one title line across the top, and the assistant at the bottom right. The rail is the only menu.

The rail

The rail has three labelled sections, and Settings at its foot:

SectionPages
FetchAsk · Briefs
SimulateOptions · Why
TwinsActionable Twins
(foot)Settings

Below Settings are the signed-in account, with Log out, and the instance FACE is acting for. Select it to choose another instance.

Pages that are not in the rail light the entry they belong to:

You are onLit entry
A past run (/fetch/run/<id>), or a live view opened from a connection’s card (Surveillance, Live sensors, Field reader, Devices)Ask
A brief’s editorBriefs
A scenario’s details pageOptions
The Map view, site plans, twin entities or one entity’s page, Follow-ups, a decision artifact, a remediation screenActionable Twins
Admin, and the connection wizardSettings

An address the app does not know lights Ask.

Each rail entry is a button with a visible focus ring. Screen readers announce its label and whether it is the page you are on.

The title line

The title line reads FACE and the page’s name. On Ask, Options, Why and Actionable Twins it also names the selected run, with Run actions (Change run, Clear). On every other page it carries the audit: how the last run ended, as Passing, Needs attention, Failing, Running or Not run yet, with the reason and Open audit. When runs have proposed connections, the audit counts them.

Not run yet appears only when the server has recorded no run at all.

The activity card

While a run or another long operation works, a card narrates it over whatever page is open. Long operations include testing a connection, indexing documents and reading a document source.

  • The headline says what is running, for example Running your fetch…, then how it ended.
  • The steps are the phases as the server reports them, each running, done, failed, out of time, skipped or partial.
  • Show reasoning opens the run’s full reasoning, with every step.
  • While a run works, the card may show messages from Runink’s partners, each under the partner’s name (From {name}).

After the answer

The answer coming back does not finish a run. The analysis and the twins’ decisions run after it. From the answer on, the card tells the run’s story:

  • The headline follows how the run ends: Fetch answered — still analysing while a phase runs; Fetch finished when every phase finished; Fetch partly finished, or Fetch partly finished — interrupted by a server restart; Fetch failed; or Fetch answered — its phases were not recorded.
  • A partly finished run says Did not finish: and names each phase.
  • The steps are four items, one per page: Ask (what was read), Why (what it means), Options (what could happen) and Actionable Twins (what we do). Each shows its state and one line from the run’s own record. While a phase works it says so, for example Judging the scenarios…, or Waiting to start. Nothing recorded appears only for a phase that ended and left nothing.
  • Select an item to open its page for this run.

A finished card closes itself after a short while, but not while the pointer is over it or focus is inside it. A running card never closes on its own. If several operations run at once, the card shows the most recent.

The automated analysis notice

If automated analysis is not ready, a strip under the title line says so:

  • Automated analysis is starting up… while it loads;
  • Automated analysis is not ready — features that need it are unavailable. when it cannot answer.

The strip adds Everything else works now. and offers Check again. It goes away once automated analysis is ready.

The FACE AI assistant

The round button at the bottom right of every page opens the assistant. Its tooltip is FACE AI, and screen readers announce Open FACE AI assistant. It floats over the page.

The assistant knows which page you are on. It suggests the question of the selected run, your persona’s question, then two prompts for the page: on Ask and Integrations, Create new connection and Find extra docs; on Options, Trigger Shadow Simulation and Optimize sliders; on Why, Check Rules and Find logic gaps; elsewhere, Help me with this screen and What can I do here?.

ControlLabel
Text boxAsk FACE…
SendSend message
MicrophoneSpeak your question. It is sent when you stop.
CameraTake a photo
AttachmentAttach a file

The answer lists every step FACE ran as done, failed or blocked, with the reason. How this was worked out shows the reasoning. FACE AI could not answer means the request failed, and the reason follows. No answer came back means the assistant finished but returned nothing.

Behind the frame: FetchService.RunFetchStream, ListFetchRuns and ModelService.GetModelStatus. See the API reference.