Skip to content

Connecting to ChatGPT

Use this guide when you want Fluent inside ChatGPT.

Fluent is one product across Meals, Style, and Health. In ChatGPT, it should feel like a connected memory layer: ChatGPT can ask Fluent what it knows, show supported rich views when available, and fall back to complete text when a view cannot render.

In ChatGPT v2, Fluent focuses on compact Meals onboarding, Grocery List v2, Style setup/calibration, and Style purchase analysis with direct or uploaded product images when needed.

Fluent is in early access. Use managed early access, run the open-source runtime, or follow the access instructions you were given.

Start Here

After Fluent is connected, ask:

"What can Fluent already help me with right now?"

That first check can show account state, ready domains, remembered context, and suggested next actions without showing technical details.

Or go straight into Meals:

"Plan dinners for next week and turn them into a grocery list."

Meals In ChatGPT

Meals is the clearest first ChatGPT workflow.

Try:

"Plan dinners for next week, lock the plan once it looks good, and show me the grocery list."

Expected flow:

  1. ChatGPT checks whether Meals is ready.
  2. Fluent generates a candidate week from saved preferences, household context, and kitchen state.
  3. You review and adjust the plan in chat.
  4. When you approve it, Fluent saves the plan.
  5. Fluent creates the grocery list.
  6. ChatGPT shows a grocery-list view when it supports it, or gives you the same list in text.

Fluent can help prepare a grocery list and reconcile what you already have. It does not complete retailer checkout, payment, or order submission inside ChatGPT.

Style In ChatGPT

For purchase decisions, give ChatGPT item details and an image when possible:

"Should I buy this jacket, given what I already own?"

Fluent can compare the candidate against your closet and style profile. The result should help you decide whether to buy, wait, or skip. It should not imply autonomous purchase completion.

If a rich purchase view cannot render, ChatGPT should still give a complete text answer from Fluent's structured analysis.

Health In ChatGPT

Health is text-first.

Use it for ordinary fitness planning:

  • today's training context
  • current goals
  • active training block
  • workout adjustments after a missed or changed session
  • weekly review

Fluent Health is not medical advice, diagnosis, treatment, emergency support, or clinical nutrition.

What To Expect From Rich Views

Rich views are helpful when they are available, but they are not the source of truth.

  • A recipe card, backed by meals_render_recipe_card, is a better way to read a saved recipe.
  • A grocery-list view, backed by meals_render_grocery_list_v2, is a better way to shop from a plan.
  • A style decision view is a better way to compare a candidate item.

Saved Fluent data changes only when you clearly approve an update. Temporary UI choices, like an expanded section or selected tab, are not the saved record.

If Something Falls Back To Text

That is normal.

You should still get:

  1. an answer based on saved Fluent data
  2. the result written clearly in chat
  3. a brief note when the visual view is not available

Managed access is in early access. The open-source runtime is available today.