STAS knowledge base

Training facts and context

STAS stores a canonical Athlete Summary built from your training facts. ChatGPT or Claude receives that summary together with the current profile, goals, rules, strategy, metrics, workouts, reports, notes, wellness, and calendar facts needed for the question; STAS does not generate or store a separate AI Condition assessment.

The Athlete Summary is refreshed from its source facts, while changing workouts, metrics, wellness, reports, notes, and plans remain available directly from their own sources at request time.

What it is

The Athlete Summary is a stored, canonical synthesis of the athlete's factual context. It is not an AI Condition assessment or a replacement for current source data.

Not just workout data

Your watch can show what you did: distance, duration, pace, heart rate, power and load. But training decisions also depend on things your watch does not know: your goals, constraints, preferred schedule, recent reports, long-term strategy and how you actually felt after key sessions.

ChatGPT or Claude brings the relevant layers together for the current question.

When facts are read

The AI requests relevant facts when you ask for analysis, a workout review, or a plan.

Source dates determine freshness. STAS refreshes the canonical Athlete Summary when its source facts change and provides current source data alongside it when needed.

What it includes

The available context has several factual sources. Some come from your watch and Intervals.icu; others are facts you explicitly save in STAS.

Main layers

When relevant and available, the AI can request:

profile — stable athlete information such as height, weight, resting heart rate, training background, personal bests and other sports;
goals — races, dates, target results and longer-term training goals;
rules — schedule limits, rest days, preferences and constraints that should shape future plans;
strategy — the long-term logic of your preparation and whether it still fits the latest context;
sport profile — your main sport and recent training volume across running, cycling, swimming and other activities;
running benchmarks — current VDOT, estimated training paces, recent best efforts and confidence in those numbers;
current training metrics — training load, fatigue, form and recent trends;
weekly history and season context — how your training has developed over recent weeks and across the season;
recent workouts — the latest sessions with key load data and report status;
wellness — sleep, HRV, resting heart rate, weight, readiness, mood and fatigue when available.

Why the extra context matters

Metrics alone do not explain the whole athlete. Two people can have similar training load and completely different goals, constraints and recovery patterns.

The more useful context you add, the less the AI has to guess. The summary starts to describe your actual training situation, not just your latest numbers.

What may be missing

Missing facts remain missing and should be stated as such.

What STAS checks

STAS can see when important pieces are missing. For example:

your profile is empty or only partly filled;
there are no saved goals or training rules yet;
you have many workouts, but no post-workout reports. This helps ChatGPT and Claude avoid treating missing information as known.

Detailed workout data is loaded separately

Detailed workouts and planned events are requested separately when the question needs them.

This keeps each answer focused without relying on a generated summary.

Where the data comes from

Context comes directly from training sources and facts you saved yourself.

Data sources

STAS can use:

completed workouts from Intervals.icu: sport, duration, distance, pace, heart rate, power and training load;
Fitness, Fatigue and Form, recalculated by STAS from completed training;
post-workout reports written in Telegram or on the workout page;
calendar notes about illness, travel, missed sessions, rescheduled workouts or other important context;
profile, goals, rules and strategy saved in STAS;
wellness data such as sleep, HRV, resting heart rate, weight, readiness, mood and fatigue when available.

Numbers plus human context

Workout data gives the facts. Human context explains what those facts mean.

A high heart rate might be heat, fatigue, stress, poor sleep, terrain, a hard group run or a sensor issue. A workout report or note can make that clear. Without that context, AI has to infer too much from the numbers alone.

How it reaches ChatGPT and Claude

ChatGPT or Claude requests the needed facts through STAS when you ask a training question.

How to ask for it

At the start of an important conversation, ask:

“Refresh my data.”

STAS exposes the relevant profile, goals, rules, strategy, metrics, history, wellness, reports, notes, workouts, and calendar facts. The AI then answers the task.

What the AI receives

From those direct facts, ChatGPT or Claude can understand:

who you are as an athlete;
what you are training for;
what rules and constraints matter;
what long-term strategy should guide the plan;
what pace and performance benchmarks are available;
what is happening now with load, fatigue, form and wellness.

One context across chats

Saved profile, goals, rules, and strategy stay consistent because they are explicit records, not chat memory or a generated summary.

ChatGPT and Claude can request the same source facts from STAS.

Why it matters

Without requested source facts, AI only sees the current message. STAS makes the relevant training facts available on demand.

No more rebuilding the context every time

Save stable facts once, keep reports and calendar notes where they belong, and ask the AI to load the facts needed for the task.

Better input leads to better answers

Useful saved facts reduce guessing, while current workouts and metrics keep the answer grounded in what actually happened.

Limits

Access to training facts does not make STAS a doctor, human coach, or autonomous decision-maker.

Important limits

STAS does not diagnose medical conditions and does not replace a doctor.
The facts are context for analysis, not the final decision.
STAS does not invent missing data. If a source did not record a metric, the AI should say it is missing.
One number should not drive a strong conclusion by itself. Training history, trends, reports, plan and recovery all matter.
Running benchmarks and metrics are estimates based on available data, not guaranteed results.
STAS does not read full chat history. Only explicit saved records and requested training sources are available.

Still only a model

Even detailed source data is not complete knowledge of your body. Check dates, add reports when useful, and review AI conclusions.

FAQ

Short answers about training facts and request-time context.

How is Athlete Summary different from Condition?

STAS stores a canonical Athlete Summary built from profile, goals, rules, strategy, workouts, reports, notes, current metrics, wellness, and other factual layers. It does not generate or store a separate AI Condition assessment.

Do I need to fill in everything at once?

No. Start with the parts that affect training most: goals, important rules and basic profile information. Strategy and more detailed context can come later.

Does STAS remember my ChatGPT or Claude conversations?

No. STAS does not store your full chat history. If something should be available in future conversations, it needs to be saved in STAS as profile information, a goal, a rule, strategy, a calendar note or a workout report.

How often should I refresh it?

There is no artifact to refresh. Ask the AI to load current facts when freshness matters and check source dates.

What if my profile, goals and rules are empty?

The AI can still use workout data and current metrics. Missing goals, rules, profile facts, or strategy remain missing context.

Can I inspect the underlying facts?

Yes. Saved personal facts are visible in your profile; workouts, reports, notes, and plans remain attached to their source records.