River Logiceode docs

How RIA acts

What RIA can see, what it writes on a model's record, what it only proposes, and why a proposal changes nothing until you click.

RIA is eode's assistant. On a model's page, the conversation with RIA is the page, with the model's record beside it; on other screens, RIA is the console, the diamond in the corner. It reads your model and its results, answers in your own terms, and works on the model with you. What you agree with RIA goes on the record, where every change shows. A change to a built model, a what-if or a workspace layout comes as a card, and nothing on a card happens until you press its button.

RIA's loop: your message plus what's on your screen go to RIA; RIA answers, writes what you agreed on the record, or proposes a card; only your click applies a card to the modelsent withevery messageanswerswrites whatyou agreedbuildyour clickYoua message to RIAWhat you're looking atscreen · model · recordRIAreads · explains · proposesThe recordwhat you agreed, turn by turnA proposal cardApply · Approve · Yes, open itYour modelbuilt and solved

What RIA knows when you ask

Every message carries what's on your screen: which screen, which model, the scenario you've selected, the workspace, your persona, and anything you've pinned. That's why "why did this drop?" works without naming anything. On a model's page, RIA also reads the model's record, so it knows what you've agreed and what still waits on you. Each model keeps its own conversations, so RIA picks up where you left off.

While RIA works

RIA shows what it's doing as plain steps (see the console), and answers in your business's words: it doesn't show its reasoning, internal ids or the names of the tools it uses. If an answer fails partway, RIA says I couldn't finish that. Try again.

What RIA can do

RIA canHow it shows up
Explain a result, a rule, a numberAn answer in the conversation, with links to the scenarios and models it mentions
Take you to a screenIt opens the screen for you
Work out a model with youOne question at a time in the conversation. What you agree goes on the record, and Since your last message lists what each turn changed (Read the top card)
Propose an objective, or ask for figuresA Your objective or Your figures card, answered on the card (Answer a card before you type)
Build the modelOnce nothing waits on you at the Check step, or when you click Build or tell it to (Build and baseline)
Write a what-ifA what-if card: Yes, open it opens the what-if in a conversation of its own, and it solves once you tell RIA it looks right (Run a what-if)
Change a built modelA model edit card (Apply), or a change order (Preview, then Apply)
Lay out a workspaceA workspace layout card: Approve saves it to the board
Read an attached fileText files and design documents you attach, and workbooks or CSV files you upload, which also go to your Datasets

Cards come one at a time; the rest wait behind a count. See RIA's proposals.

What RIA can't do

  • Apply a card without your click on it.
  • Build a model again after it's built, unless you say so. A question about the model changes nothing.
  • Do anything you couldn't do yourself. On a model's page, only people who can change the model talk to RIA there; everyone else reads the conversation. Only owners, contributors and Org Admins can apply a card. See Who can decide what.
  • Act on a model you haven't opened, or one whose structure isn't editable yet. It says so in its reply.

Undo

Every turn and what it changed stays in the record's History (Read the history). What RIA applies from a card is a new version of the model, and an owner or Org Admin can go back to an earlier version (Versions and history).

On this page