Fuel

Use this to answer macros for an everyday meal description or meal photo with an editable Fuel card. First search_foods for compatible USDA food/preparation IDs. ChatGPT interprets ordinary language or the actual attached photo; this server does not inspect images. Do not require grams to start. Use sourced cups/pieces when available; otherwise declare a reasonable portion estimate with amountOrigin, everyday portionLabel and estimatedPortion. Record material assumptions in the card source details; ask one focused question only for material ambiguity (e.g. cooked or dry rice when context does not resolve it). Unknown recipes/brands stay unresolved. Never invent nutrients. Preserve amount origins and other foods on corrections; changing cooked rice to dry requires the newly stated amount, not reuse of the cooked weight. Only include listed ingredients. If the meal cannot be identified, do not call a calculation tool. Ask for a recipe, label or photo. If the user already says ingredients or amounts are unknown, do not ask for the same information again; offer a photo or recipe as the next step. The card is the answer: never repeat its totals, foods, sources or assumptions in chat. For a complete meal return no additional prose. For an incomplete meal ask one useful short question and stop; do not explain the incomplete status again. If the user cannot identify an ingredient or amount, ask for a photo, recipe or label rather than repeating the same question. Returns reproducible source-backed arithmetic, estimate status and sources. Guest only. For a diary-save request respond only: "This calculator cannot save meals to a food diary." Do not call tools, recap the meal or add suggestions.

calculate_macros

Convenience deterministic parser for a typed list of supported foods with amounts, one per line or comma separated, e.g. 150 g cooked chicken breast, 200 g cooked white rice, 100 g cooked broccoli. Uses a declared 12-record USDA catalog and exact aliases only. Unknown foods, ambiguous state/portions produce a labeled subtotal and clarification, never a guessed full answer. Photos and URLs unsupported. Prefer search_foods then calculate_macros for host-resolved language.

meal_macros

Use query * for 12 quick choices. Search 7,793 offline USDA SR Legacy records from April 2018. Returns candidate food IDs, preparation states, per-100g values and exact source portions. Returns at most 12 candidate descriptions; refine broad queries. Select a compatible record. Compute immediately for an explicit unambiguous supported food, state and portion; ask only when a consequential detail is missing. Historical brand descriptions are not current label verification. This is an offline snapshot, not live USDA search. No result means unsupported: do not invent a food/brand match, nutrient or raw/cooked conversion. Prefer source household portions. A visual/default amount may be declared as an estimate in calculate_macros, never as a sourced conversion. Ask only material ambiguity, without requiring a scale.

search_foods