Insights

Insights has two tabs: an Overview of aggregate cards, and a Query tab where you ask questions in plain language.

Overview

Aggregate cards over the active team’s issues — counts by status and priority, what is overdue, what is unassigned. No configuration; it is there to be glanced at. The cards and their drill-downs read only the active team’s issues on your device and use no model.

Query

Ask a question, get an answer rendered as a chart, a table or a single number, with a written explanation.

Things it handles well:

• “How many open bugs are unassigned?”
• “Which issues are past their due date?”
• “What did we complete last month, by project?”
• “Who has the most open issues?”

It is a conversation, not one-shot. Follow-ups carry the previous turns, so “and how does that compare to the month before?” works. Threads are kept on this device so you can return to one.

How a question runs

The model writes a database query for your question. That query never runs against your workspace’s local database. For each question, Orneos copies the active team’s tasks, projects, comments, task relations and labels — and only the fields the model is told about — into a temporary, separate query space in your browser. The query runs there, and the space is discarded when the answer is ready. A query that runs longer than 5 seconds is stopped.

The copy is taken in pages while your device keeps syncing, so it is not a single frozen snapshot. A change that arrives during the copy may or may not be in the answer, and tasks and comments can reflect slightly different moments. It never includes another team’s data, and if you switch team or account mid-copy the question is stopped. Ask again for a fresh copy.

Limits you may see:

• Team size. Several limits apply at the same time: up to 25,000 tasks and 50,000 comments per team, and about 16 MB for the whole copy. The 16 MB counts each row’s field names and other structure as well as its text, so a team usually reaches it before reaching every row limit at once — 25,000 tasks together with 50,000 comments already exceeds it, even with very short text. The row limits are maximums, not a size you can always reach. A team over any limit gets a message saying so before the question is sent to the model; the Overview and filters still work.
• Long text. Descriptions and comments longer than 8,192 characters are shortened for Ask, and the answer says when that happened.
• Large results. A result shows at most 10,000 rows and 100 columns, and very long values are shortened. The answer notes when anything was cut.
• Comments. Only comments already loaded on your device are included.
• Earlier answers. Answers from before this change are still shown in saved threads, but they are not sent as context for follow-ups or analysed again. Use Refresh to re-run one.

What stays in your browser

To show recent answers quickly, this browser keeps up to 20 recent answers, including the rows they showed, for up to 7 days. It shows them only to the same person in the same team. Each browser tab also keeps your recent questions and the last conversation you had open. Saved conversations are kept with this device’s local copy of the workspace.

• Signing out removes all of it from this browser, in every open tab.
• Losing access to the company removes it when a tab that is using that company notices the change. If no open tab is using that company, its recent answers stay in this browser until they expire (at most 7 days); they are not shown unless you regain access to that team.
• Switching team keeps it; another team’s answers are never shown.
• Switching organization keeps the recent answers and this tab’s questions in this browser; they are never shown in another company. Saved conversations for the previous organization are removed with its local data.

Treat a shared computer accordingly: anyone using your signed-in session can see these answers.

Semantic grouping

Questions about themes or topics can use semantic analysis rather than a structured database query. Results can group related issues and link back to the underlying work. Check those groupings before treating them as a factual taxonomy.

What an answer is

This is the same rule as everywhere else in the product: a model proposes, a person decides.

Requirements

The Query tab uses the server-configured provider, not your personal drafting key, and needs a connection. The non-model Overview is separate.

When Anthropic is enabled, the interface asks for consent to send context and use your Orneos allowance. What is sent: your question; the team’s status, priority and type values, member names, project names and labels so the model can write the query; for follow-ups, the earlier questions, their queries and a small, size-limited sample of their results; and, for the written explanation, at most 20 size-limited rows of the result. Overview-style questions (“describe this workspace”) send a sample of task titles and short description excerpts, and theme grouping sends task titles. Query and summary can consume separate requests; interruption may still cost credit. Check the displayed allowance and shared limits. Availability depends on deployment configuration and granted access. See AI funding and usage.

Scope

Insights sees the active team’s data only: Overview reads its issues, and Query reads its tasks, projects, comments, task relations and labels. It does not span teams, and it does not read Vault content, saved views or any other part of your workspace.

Related

• Filters and saved views — the exact alternative, and how to check an answer.
• AI in Orneos
• Drafting keys