add_profile
ChatGPTCreates a one-time link, valid for 30 minutes, for another person to add their own profile to this ChatGPT connection. They open it, sign in with their own Google account and approve. Only when the user explicitly asks; it never creates a profile or gives access to anyone without their approval.
add_to_favs
ChatGPTSave one user-approved favorite food with its usual portion, the five supplied per-serving nutrition values, and optional short aliases (for example обычный творог). The item is a reusable snapshot compatible with record_meal. Use stated values, a saved meal snapshot, or values looked up from a reliable source for the stated product and portion. This creates a template only, without recording consumption. Search get_favs first to avoid unintended duplicates. A fresh requestId creates a new favorite; identical retries replay the saved result.
delete_fav
ChatGPTArchive one owner-scoped favorite after get_favs identifies its favoriteId and current revision. Pass expectedRevision to reject stale or concurrent changes. Historical meals remain unchanged; archived favorites disappear from get_favs. Identical requestId retries replay the original result.
delete_goals
ChatGPTClear the complete current daily calorie/protein goal plan. Read get_goals first and pass expectedRevision. This advances the revision so concurrent chats cannot overwrite the deletion. Meal history is unchanged; identical requestId retries replay the result.
delete_meal
ChatGPTLogically delete exactly one owner-scoped food entry after list_meals identifies its entryId and revision and read_day supplies the day revision. Ambiguous entries must be clarified; this never deletes an entire multi-item meal implicitly. The day remains even when its last entry is deleted. Identical requestId retries replay the result.
delete_measurement
ChatGPTLogically delete exactly one owner-scoped weight/body-fat event. Call list_measurements first, resolve ambiguous readings, and pass its measurementId and current revision. The event disappears from lists, daily projections, and weight progress. Identical requestId retries replay the result.
delete_record
ChatGPTPermanently deletes one meal (with its entries), entry, usual product, recipe, target or weight. Logged meals keep their values when a product or recipe is deleted.
delete_weight_plan
ChatGPTClear the complete current weight plan, including all loss, maintenance, and gain phases. Read get_weight_plan first and pass expectedRevision. Body measurements remain intact and the revision advances to protect against concurrent edits. Identical requestId retries replay the result.
get_favs
ChatGPTRead this user's favorites, optionally searching title and aliases by a short product name. For 'мой обычный творог и йогурт', search творог and йогурт separately. Omit query to list all; follow nextCursor to inspect every candidate. Returns exact saved portions and complete item snapshots for record_meal. Do not choose the first ambiguous match: ask the user. If a narrow search is empty, list all before concluding absence. If no favorite matches a “usual” food, ask which product and portion the user means; for ordinary foods just look up nutrition and record. Offer add_to_favs afterwards. Never invent a match or add a favorite without user agreement.
get_goals
ChatGPTRead this user's optional daily calorie budget (kcal), protein target (g), and revision. Unset goals are null. Offer optional setup on first use, but never block logging or invent targets.
get_weight_plan
ChatGPTRead the current user-chosen weight plan, derived consecutive phase dates and target weights, and revision. An absent plan is null with revision 0 initially. Read before setting or editing a plan.
link_food
ChatGPTReuse a food or recipe from another profile already approved on this connection. Pass its source profile and library record ID from list_saved/read_food, and a name for the selected person's library. Creates a personal link to the same shared revision; never copies private notes or changes an existing chosen revision. Recipe ingredient definitions are retained too. For logging, use new_foods type=link in save_meal instead to make linking and logging atomic.
list_meals
ChatGPTList persisted meals for one explicit diaryDate in bounded pages. Results include stable entry IDs and entry revisions needed by update_meal, are owner-scoped, and order unknown consumedAt values first followed by UTC timestamps and entry ID. Follow nextCursor while reading an unchanged dataset; cursors are signed for this owner and date and support null consumedAt positions.
list_measurements
ChatGPTRead an explicit inclusive startDate/endDate range of at most 366 days. Use raw events or the daily latest-event projection; follow signed nextCursor pages. Missing body fat is never copied from another event. Measurement dates are independent of open food diary days.
list_profiles
ChatGPTThe profiles this ChatGPT connection can use, with names. Each is an independent person who joined with their own sign-in. Call when it is unclear whose diary is meant.
list_saved
ChatGPTLists usual products or recipes (q filters by text), targets (every variant, newest first; kind filters) or weights (from/to filter). Paged.
read_catalogue_record
ChatGPTOne catalogue record with its published values, units and bases, plus the records grouped with it for this profile.
read_day
ChatGPTRead one explicit diaryDate and its current revision. hasData distinguishes recorded food from missing dates; revision remains available for empty technical days. Totals include calories, protein, fat, carbs, fiber, and knownItems/totalItems coverage. value is null for incomplete nutrition; knownValue is the separate known sum, and complete describes nutrient coverage; unknown values remain unknown and do not become zero.
read_diary
ChatGPTMeals with entries, weights, the day's target kind and effective target for each day from..to (at most 31 days).
read_food
ChatGPTRead one selected reusable food, including its full nutrients, evidence and recipe ingredients. Search list_saved(record_type=foods) first.
read_goal_progress
ChatGPTCompare recorded intake with CURRENT optional goals for an explicit diaryDate and the seven calendar days ending there. Return daily coverage, remaining/exceeded amounts and weekly averages with explicit eligible-day denominators. Missing dates and unknown nutrients are excluded with separate denominators. Current day is excluded by default; includeCurrentDay counts it once and marks progress provisional. Future dates are always excluded. Coverage describes recorded data, not completeness of actual food intake. Calories are an upper budget; protein is a target to reach. Use after food writes and for progress advice.
read_nutrition_stats
ChatGPTRead live daily nutrition, period totals and equal-weight daily averages for 1–366 inclusive calendar dates. Uses the owner's IANA timezone and local midnight. Missing dates are gaps, unknown nutrients exclude only that nutrient's day, and each denominator is returned. Current day is excluded by default; includeCurrentDay counts its recorded food once and marks the report provisional. Future dates are shown but always excluded. value/total are null for incomplete coverage; knownValue/knownTotal are partial known sums. No completeness of actual eating is implied.
read_profile
ChatGPTRead the selected profile name/version and saved body facts: height, optional date_of_birth (YYYY-MM-DD), dated age, activity and goal. A missing context is null. Read before asking for missing details; meal logging never requires body setup.
read_weight_progress
ChatGPTCompare body measurements with the current weight plan on an explicit date. Report phase, planned target, dated latest actual weight and deviation at its observation date, plus seven-day daily-projection averages with coverage. Missing dates are excluded, never zero-filled; disclose stale readings and sparse coverage. Body-fat percentage comes only from supplied measurements. This reports the user-selected schedule, not a health recommendation.
record_meal
ChatGPTRecord one to twenty food items atomically for an explicit diaryDate. Every item must include per-serving caloriesKcal, proteinG, fatG, carbsG, and fiberG as finite nonnegative numbers; a known zero is valid. consumedAt is optional and remains null when unknown; timezone remains required. For usual or favorite foods, call get_favs first and copy the selected saved item at its default portion. For other foods without stated nutrition, look up typical values for the food, preparation, and portion from a reliable source and record them; mention in the answer that they are estimates. Reuse requestId only to replay the identical normalized result; changing its payload conflicts.
record_measurement
ChatGPTRecord an owner-scoped scale observation with explicit measurementDate and timezone. Supply weightKg and/or bodyFatPercent (0 to 100); unknown values remain null. measuredAt is optional: omit unknown time, never invent it. Multiple readings per day are preserved. source may describe home bioimpedance scales. Supply requestId for identical retry safety.
save_meal
ChatGPTLogs a new meal with entries, or adds entries to an existing meal when meal_id is given. Report success only from the returned record.
save_product
ChatGPTSaves a new usual product. Label values are per the stated basis.
save_recipe
ChatGPTSaves a new recipe with its ingredients and whole-recipe values.
save_target
ChatGPTSaves a normal-day or training-day target variant from its effective date. Only after the person explicitly agreed to it.
save_weight
ChatGPTRecords a dated weight in kg or lb as stated.
search_catalogue
ChatGPTSearches Ciqual, BLS, USDA and Open Food Facts by words (q) or barcode (gtin). Results are grouped by shared barcode and this profile's confirmed links; each record keeps its own published values and units, and USDA records carry their raw data_type (e.g. branded_food, foundation_food, sub_sample_food). Paged by group.
search_diary
ChatGPTText search over logged entries, usual products and recipes, newest first. Paged.
select_day_target
ChatGPTMarks a diary day as a normal or training day. Days without a choice use the normal-day target. To change an existing choice, pass the day's target_kind_version from read_diary.
set_food_link
ChatGPTRecords (linked: true) or removes (linked: false) this profile's assertion that two catalogue records are the same food. Only after the person confirmed it; never changes logged meals.
set_goals
ChatGPTSet or edit user-chosen daily caloriesKcal and/or proteinG. Omitted fields retain their values; null removes a goal. Both are optional for diary use. Read get_goals for expectedRevision (0 initially). Supply a fresh requestId; identical retries replay the saved response. Targets apply to current-goal comparisons, including past days.
set_weight_plan
ChatGPTSave a complete user-chosen plan with startDate, startWeightKg, and consecutive phases containing name, durationWeeks and signed weeklyChangeKg (negative loss, positive gain, zero maintenance). Read get_weight_plan for expectedRevision and use requestId. plan:null clears the plan. Durations represent elapsed weeks: each phase ends start+7*weeks days, and the next starts on that boundary. Use the stated baseline weight or the latest recorded weight; ask only if neither exists. Clarify ambiguous dates; never prescribe targets. Replacing the plan also changes comparisons for past dates.
update_body_context
ChatGPTSave only supplied height, optional date_of_birth (YYYY-MM-DD), dated age, activity and overall goal after the person asks or agrees. Omitted fields stay unchanged; null clears a whole field. Age includes its as_of date and never ages automatically. Read first: version 0 is only for absent context. Weight history and nutrition targets stay separate and unchanged. Never store body facts in weight notes. consent_required needs the selected owner to review Settings; meal logging stays available.
update_entry
ChatGPTChanges only the given fields of one entry. Values are for the amount eaten and are never rescaled by the app: when quantity changes, send the recalculated nutrients with it. For a value-only correction send just the changed nutrient keys; others are kept. Returns the entry and the meal's new version.
update_fav
ChatGPTReplace one owner-scoped favorite's complete item snapshot and aliases. Call get_favs first and pass its favoriteId and current revision as expectedRevision. Supply all five nutrition values, stated by the user or looked up from a reliable source for the new portion. This is a full replacement, not automatic scaling. Omitting aliases clears them; send the current aliases to keep them. A stale revision is rejected, and identical requestId retries replay the saved result.
update_meal
ChatGPTCorrect exactly one owner-scoped meal entry identified by stable entryId and explicit diaryDate. Read list_meals first and pass both expectedEntryRevision and expectedDayRevision. In replacement mode, title, quantity, unit, caloriesKcal, proteinG, fatG, carbsG, and fiberG are all required; consumedAt/timezone can be omitted to preserve their saved values. In rescale mode, pass only a new quantity; the unit and every saved nutrition value are scaled deterministically at fixed precision, and incomplete legacy nutrition is rejected. requestId retries are idempotent only for the identical payload.
update_measurement
ChatGPTCorrect a measurement by measurementId and expectedRevision from list_measurements. Omitted fields are preserved; null clears nullable fields, but at least weight or body fat must remain. Read and disambiguate the event first. Supply a fresh requestId for each intentional edit.
update_product
ChatGPTReplaces a usual product with the given fields. Meals already logged keep their values.
update_profile
ChatGPTChange only the selected person's saved profile name, after they ask. Read its version with read_profile or list_profiles. Profile facts, goals and weight use their own tools.
update_recipe
ChatGPTReplaces a recipe, including its whole ingredient list. Meals already logged keep their values.
update_target
ChatGPTReplaces a saved target variant. Only after the person explicitly agreed to the change.
update_weight
ChatGPTReplaces a recorded weight.