Skip to main content
Tell chat about a change in your health or routine. getbased can prepare updates in the relevant records, so you do not have to find every field yourself. Nothing is saved until you choose Apply on a proposal card. You can accept some changes and discard others.

Start with what changed

For example, you could say:
I started swimming twice a week. Please add improving endurance as a low-priority health goal.
Chat can read the relevant saved records and propose the new facts. It should preserve information you did not change and ask when a detail is ambiguous. A statement such as “my digestion changed” needs more detail before it can become a useful record. Ordinary questions, hypothetical examples, and advice do not automatically become records. A suggested supplement is not a treatment you have started, and a suspected diagnosis is not a confirmed condition. You can also say that you do not want something remembered.

Review and save

  1. Send your message and wait for the response to finish. Tool activity can show which records chat is reading or preparing.
  2. Find the Proposed change cards. Each identifies the destination, the change, your supporting words, and a reason. Existing and proposed values are shown where applicable.
  3. Choose Apply for a change you want, or Discard to leave that record as it is.
  4. Check the result in its normal card or record view. You can continue editing there.
A sentence saying “I prepared a draft” is not a proposal. If there is no native card with Apply / Discard, no reviewable update has been prepared. Ask chat to prepare the supported change again, or use the normal editor. If a proposal needs correction, tell chat what to change and review the replacement. A newer overlapping proposal can supersede an older one. Editing a saved record, changing profiles, or changing the chat permission can make an older review unusable; request a fresh proposal instead of applying an outdated one. When a save result is uncertain, use the card’s Check saved change action when offered and inspect the destination before trying again.

Paste a questionnaire or health summary

You can paste a summary you already have and ask chat to prepare the relevant updates. Replace the example facts below with your own before sending:
Please prepare reviewable updates to my getbased profile from this health summary: I eat a vegetarian diet, sleep 7–8 hours, and swim twice a week. My goal is to improve endurance, low priority. Record only these stated facts. Keep other saved answers and show proposals for me to review.
Chat can prepare several proposals in one response, including changes to different context cards and individual records. Review them separately: applying one card does not approve the others. This is not an all-or-nothing import, and the model may need a follow-up for missing details or conflicts. Do not assume every sentence in a long intake has been captured. For reports containing measured lab results, use the reviewed lab-import workflow. Selecting and analyzing a file is a separate action from telling chat about it.

What chat can help update

The exact fields and operations depend on the record and current permissions. This is not unrestricted access to every editable field. Calculated results are not invented or set as measurements. Deleting a complete lab sample, deleting an imported report, administering light devices, and operating live timers remain in their native workflows. Report history, saved device specifications, and light audits can be readable without being editable through chat.

What stays in the normal UI

Use Settings and the relevant editors for provider or model switching, units and display preferences, profile switching or deletion, backups, Sync, payments, connections, Knowledge Base management, custom personas, file selection, and hardware or timer controls. Chat can explain where to go without changing these controls. Agent Access for an assistant outside getbased is a separate, read-only context-sharing feature.

Providers and cost

The same review system serves direct API chat and supported local CLI agents. The selected model and adapter still need compatible tool or structured-output support. A provider appearing in Settings does not guarantee every model supports this workflow. Personal-gateway targets do not receive the local getbased tool bridge. Reading records, preparing proposals, and responding after tool calls can take several model requests and more tokens than an ordinary reply. Cost depends on the provider, model, enabled context, and number of changes; check the response’s usage information.

Turn the feature off

Open Settings → AI and turn Health updates through chat off. It is on by default. Ordinary conversation and enabled baseline context remain available; manual editors continue working. This setting belongs to the current browser. Turning it off revokes pending tool authorization. Turning it back on does not reactivate older pending reviews: request fresh proposals. Chat cannot turn this permission on for you. Context-source switches under Insight → Manage → Context → Data sources are separate. They still control what stored context and exact reads can disclose. Turning this feature on does not override a source you turned off.