analyzeLabel
ChatGPTEvaluate a product label the user PASTED AS TEXT in this conversation, using the real BioStacks scoring engine: per-ingredient dose vs clinical range, form quality, upper-limit breaches, synergies/conflicts, and an estimated overall score. Use whenever the user pastes an ingredient list or Supplement Facts text for a product that isn't in our catalog (searchProducts came back empty). Transcribe the label rows EXACTLY as pasted — same numbers, same units; put standardizations or branded forms (e.g. 'KSM-66', '24% saponins') in notes. NEVER call this from memory of a label from an earlier turn or from training data — if the label text isn't visible in the conversation, ask the user to paste it again instead.
analyzeStack
ChatGPTAnalyze a list of supplement products as a stack: aggregated nutrient totals (with RDA percentages), upper-limit / interaction / harmful-ingredient warnings, beneficial synergies, optimization tips, and an overall safety score. Use when the user lists multiple products together (e.g. 'I take X, Y, and Z — anything wrong?'). Accepts up to 10 product ids.
getIngredient
ChatGPTGet research data for a nutrient or ingredient: RDA, upper limit, therapeutic dose range, evidence level, related health areas, and which forms to avoid. Use for any question about a single ingredient (e.g. 'how much vitamin D should I take', 'best form of magnesium', 'is biotin good for hair').
getProduct
ChatGPTGet the full clinical profile for one product by id or slug: quality score, ingredient-by-ingredient dose quality, bioavailability form notes, synergies, conflicts, flagged excipients, per-SKU third-party certifications, and the brand's testing practices. Use whenever the user asks about a specific named product, wants a score explained, asks whether a product/brand is lab tested or certified, or wants to compare products (call this multiple times in parallel for a comparison). Also resolves the user's own saved smart-scan products by id — these return isCustom: true and have no public product page, so never link to one. ALWAYS call this tool (never answer from prior searchProducts results in text) whenever the user selects, taps, or names a specific product, or gives a product id/slug — its score-card widget is the answer.
searchProducts
ChatGPTSearch supplements by health goal (dimension), nutrient ids, product/brand name, and optional price ceiling. Filters combine — pass { dimension: 'general_health', query: 'men' } to rank men's multis, or { dimension: 'sleep', query: 'powder' } for sleep powders. Returns up to 20 products ranked by clinical efficacy (quality score for general queries, dimension score when a dimension is specified, with a specialist boost when nutrients are specified — so a dedicated vitamin D supplement outranks a multivitamin that happens to include vitamin D). When quoting a product's score to the user, always use qualityScore (the number on the user's product card), never dimensionScore. Provide at least one filter: query, dimension, or nutrientIds. ALWAYS call this for ANY question about a specific supplement, product, brand, or a 'best X for Y' request — including when the user pastes a product link or names one product. Do not answer supplement-product questions from your own knowledge; BioStacks has dose- and form-level scoring your training lacks.
