Before you begin

Colattice is a private research workspace. Use the username and password provided by the person who runs your installation. There is no public account signup. The owner configures the AI provider, available tools, and access policy.

If Ask reports that agents are unavailable, ask your workspace owner to finish agent setup. Your saved projects, agents, tools, and workflows remain available. After signing in, Help stays at the bottom right of every workspace page. Ask the Product Support agent about controls, or read the built-in guides even when AI is unavailable. AI support uses the selected project to retain its questions and answers; no attachments are sent and it has no tools.

Ask your first question

Start small: understand a dataset before asking for a complete modeling study. A CSV spreadsheet or PDF paper is enough to begin.

  1. Open Ask and attach a CSV or PDF with the paperclip, or drag a file into the composer.
  2. Write a specific goal, such as “Check this spreadsheet for missing values and duplicate rows. Explain the problems. Do not build a model.”
  3. Turn on “Show me the plan first” when you want a checkpoint before the investigation proceeds.
  4. Send with Enter or the arrow button. Use Shift + Enter to add a new line. A first request can create its own project; you can select an existing project for related work.

Follow the work

The request shows its current phase and activity. “Needs your input” means there is a question or plan to review. “Crunching numbers” means a calculation is running. A queued request has been accepted but has not started yet.

Open the research team activity to inspect the available coordinator and specialist updates. Simple questions may be answered directly; not every request requires a team of specialists or a scientific calculation.

Read the answer, then the evidence

Read the response alongside its saved results. Follow the source references, inspect warnings, and check which files and methods were used. A completed run is not a guarantee that a scientific conclusion is correct.

Ask a follow-up in the same project to retain relevant context. Before sending, check the file chips: selected earlier files are included, and you can switch off any that should be left out.