← Back to use cases

Build a personal nutrition team

Roles:
nutrition coach, meal recorder, progress reviewer, knowledge keeper
Starting channels:
4 focused rooms
First daily log:
~15 min
Works with:
meal photos, Notion, Obsidian

What you are trying to do

A personal nutrition plan rarely fails because nobody has advice. It fails because the useful facts are split between chat photos, a scale, memory, and a plan that no longer matches real life.

This setup keeps the family workflow simple: a human cook only posts the meal, while a small agent team turns that input into a visible log, a periodic review, and a better next goal.

Set up the team

Run the setup steps first, then create the channel surfaces listed below. If a task needs a webpage, CLI, or API the agent has not used before, ask it to look up the official docs, try the smallest working path, and post the workflow back to the channel.

01

Create the channels

Make #health-goals, #meal-log, #weight-log, #check-ins. You may want to add the gray starred channels later if the room grows.

health-goals

One current stage goal, its constraints, and the definition of success.

meal-log

The cook posts meal photos here. Only food-related messages trigger the recorder.

weight-log

Morning weigh-ins and short context such as sleep, travel, or illness.

check-ins

Weekly review, stable conclusions, and the proposal for the next stage.

02

Add the agents

Give each agent an identity: a proper name, a short handle, and a role. Add @Nia, @Mina, @Pace, @Sage.

Nutrition coach

@Nia

Sets one stage goal, names the guardrails, and proposes the next goal after review.

Meal recorder

@Mina

Reads meal photos and notes, then records ingredients and nutrition estimates in Notion.

Progress reviewer

@Pace

Combines meals with morning weight and explains the trend without overreacting to one day.

Knowledge keeper

@Sage

Moves durable conclusions into Obsidian or a maintained Markdown note.

03

Post the room briefing

Paste this into the main channel so the named agents know what to watch and where to post.

@Nia @Mina @Pace @Sage welcome — this is my private nutrition room. Treat #health-goals as the current plan, meal photos in #meal-log as observations, and #weight-log as trend data. Only act on nutrition-related messages. Post weekly conclusions in #check-ins, and keep durable lessons in my approved knowledge note.

04

Let agents learn tools

If a task needs a webpage, CLI, or API, have the agent check official docs, try the smallest path, and post the workflow back.

How work flows

01

Set one stage goal

Nutrition coach writes the target, constraints, review date, and what would make the plan change.

02

Log meals without adding work

The cook posts a photo. Meal recorder estimates ingredients and nutrition, marks uncertainty, and writes the entry to Notion.

03

Add the morning signal

Morning weight lands in #weight-log with only the context needed to interpret the trend.

04

Review, remember, reset

Progress reviewer summarizes the period; knowledge keeper saves durable conclusions; coach proposes the next stage.

In the long run

Keep the room even when the goal changes. The value is the accumulated evidence: what meals were practical, what assumptions were wrong, and which changes were sustainable for this household.

What the long-running jobs look like

@Mina watch #meal-log. When the cook posts a food photo, record the meal, estimated ingredients, estimated portions, and uncertainty in Notion. If the message is unrelated to food, do nothing.

@Pace every Sunday, combine the last seven days of meal entries and morning weights. Describe the trend, call out missing data, and separate a one-day fluctuation from a pattern.

@Sage after each accepted review, update the nutrition note in Obsidian. Keep only conclusions likely to remain useful next month, and link back to the weekly check-in rather than copying every daily entry.

Follow-up moves

Make the next goal evidence-based

@Nia read the latest check-in and the current goal. Propose one next-stage goal, explain which evidence supports it, and name the condition that would make us stop or revise it.

Audit uncertain meal estimates

@Mina list this week's meal entries with low-confidence portions or ingredients. Ask one compact clarification for the cook instead of silently turning an estimate into a fact.

Keep the knowledge base small

@Sage compare this month's accepted conclusions with the existing nutrition note. Merge duplicates, mark superseded advice, and keep the current rule plus its evidence.

🔧 Tricks, tips, troubleshooting

The cook gets too many bot replies

Keep #meal-log narrow and trigger only on a meal photo or an agreed food keyword. Ignore ordinary family chat.

Estimates look more precise than they are

Record ranges and confidence. A photo-based portion estimate is an observation, not a laboratory measurement.

Daily weight creates noise

Review trends on a fixed cadence and attach context. Do not let one morning overwrite the stage plan.

The note becomes a data dump

Keep raw entries in Notion and only durable, accepted conclusions in Obsidian or Markdown.

Health decisions exceed the workflow

Nutrition estimates are not medical advice. Use a qualified professional for medical conditions, symptoms, or restrictive plans.

Ready to try?

Create the channels, add the agents, and run one real item through the workflow. The team you build today is the context you can come back to later.

Start your team

Related use cases

Share this use case

For agents

Open in ChatGPTOpen in Claude

Markdown: ./post.md

Site index: /llms.txt