{ "llms_version": "1.5.0", "last_updated": "2026-03-06", "site": { "name": "Recipe Alchemy", "canonical_url": "https://recipealchemy.ai", "contact": { "email": "support@recipealchemy.ai" } }, "product": { "category": ["recipe_generation", "nutrition_analysis", "ai_governance"], "primary_differentiator": "claim_authorization_architecture_with_usda_validation", "summary": "Reference implementation of 'Simulators propose. Reality vetoes.' architecture. AI generates recipes through governed envelopes; 7-phase deterministic pipeline computes nutrition using USDA FoodData Central with completeness gates and provenance enforcement.", "architecture_pattern": "Claim Authorization Architecture (CAA)" }, "ai_models": { "router": "openrouter", "total_models": 7, "providers": ["openai", "anthropic", "google", "perplexity"], "categories": { "research": { "description": "Real-time web search and deep research models (Perplexity)", "models": [ { "id": "perplexity/sonar-deep-research", "use_case": "Deep ingredient identity research (CFPO v2)" }, { "id": "perplexity/sonar-pro-search", "use_case": "Ingredient research, deep knowledge/nutrition research, web search" } ] }, "image_generation": { "description": "Visual content creation models", "models": [ { "id": "google/gemini-3-pro-image-preview", "use_case": "Recipe hero images, meal presentation visuals, plating suggestions" } ] }, "text_anthropic": { "description": "Anthropic Claude models for structured reasoning", "models": [ { "id": "anthropic/claude-opus-4.5", "use_case": "Recipe generation (primary). Highest-fidelity structured recipe output." }, { "id": "anthropic/claude-sonnet-4.5", "use_case": "Leftover suggestions, content generation fallback." } ] }, "text_openai": { "description": "OpenAI GPT models for diverse tasks", "models": [ { "id": "openai/gpt-5-mini", "use_case": "Nutrition analysis, content generation, insights, knowledge, shopping lists, fast tasks" } ] }, "text_google": { "description": "Google Gemini models for fast multimodal", "models": [ { "id": "google/gemini-2.5-flash", "use_case": "Recipe previews, enrichment preview teasers" } ] } }, "primary_models_by_service": { "recipe_generation": "anthropic/claude-opus-4.5", "nutrition_analysis": "openai/gpt-5-mini", "deep_research": "perplexity/sonar-deep-research", "image_generation": "google/gemini-3-pro-image-preview", "fast_tasks": "openai/gpt-5-mini" } }, "authority_claims": { "nutrition_numbers_generated_by_llm": false, "nutrition_source_is_usda_fdc_only": true, "missing_data_policy_is_unknown_not_guess": true, "per_value_provenance_required": true, "manual_verification_for_public_recipes": true }, "authority_model": { "principle": "simulators_propose_reality_vetoes", "pattern_name": "Claim Authorization Architecture", "trust_domains": { "llm_allowed": [ "recipe_text", "instructions", "cooking_tips", "pairings", "explanations" ], "llm_forbidden": [ "nutrition_values", "nutrient_panels", "micronutrient_amounts", "calorie_counts", "protein_fat_carb_grams", "nutriscore", "nova_classification" ] }, "enforcement": { "type": "architectural", "description": "Nutrition values are emitted only by deterministic services, not by the LLM.", "hard_rules": [ "NO_LLM_NUTRITION_NUMBERS", "NO_PROVENANCE_NO_OUTPUT", "MISSING_REQUIRED_IDENTITY_BLOCKS_CALCULATION" ] } }, "nutrition": { "oracle": { "name": "USDA FoodData Central", "oracle_type": "external_dataset", "datasets": ["Foundation", "SR_Legacy", "FNDDS", "Branded"] }, "provenance": { "required": true, "required_fields_per_value": ["fdc_id"], "policy_on_missing": "return_unknown", "reject_if_missing": true }, "identity_requirements": { "canonical_identity_layer": true, "required_attributes": ["prep_state", "form", "fat_tier_if_applicable"], "completeness_gate": { "enabled": true, "policy_on_missing": "ask_or_block" } }, "cooking_method_adjustments": { "enabled": true, "approach": "retention_factors", "examples": [ "boiling_reduces_vitamin_c", "steaming_retains_minerals_better_than_boiling" ] }, "metrics": { "nutrients_tracked_minimum": 60, "derived_metrics": ["nutri_score", "nova_classification"] } }, "human_verification": { "public_recipes_manually_verified": true, "scope": "all_public_recipes" }, "features": { "recipe_generation": { "natural_language_input": true, "dietary_preferences": true, "ingredient_substitution": true, "portion_scaling": true, "quality_assurance_workflow": true, "governance_envelopes": true, "seven_phase_pipeline": true }, "nutrition_analysis": { "usda_fooddata_central": true, "nutrient_count": 60, "provenance_tracking": true, "cooking_method_adjustments": true, "nutri_score": true, "nova_classification": true, "completeness_gates": true, "identity_authority": true, "yield_factor_adjustments": true }, "ai_governance": { "claim_authorization_architecture": true, "execution_envelope_compilation": true, "drift_detection": true, "ontic_turbulence_instrumentation": true, "observer_mode_governance": true, "wire_harness_preview": true, "three_orthogonal_runtime_facts": true }, "smart_search": { "enabled": true, "modes": ["ingredient", "cuisine", "dietary", "free_text"] }, "dietary_filters": ["keto", "paleo", "mediterranean", "dash", "vegetarian"], "recipe_count_claim": { "minimum": 500, "type": "ai_generated" }, "recipe_of_the_day": true }, "technical_architecture": { "pipeline_phases": [ "recipe_generation", "quality_assurance", "identity_resolution", "nutrition_cascade", "calculation_safety", "human_curation", "content_enrichment" ], "governance_pattern": "simulators_propose_reality_vetoes", "completeness_enforcement": "requires_specification_on_incomplete_identity", "authority_hierarchy": [ "human_locks_p1.0", "canonical_targets_p1.5", "curated_mappings_p1.75", "bayesian_matching_p2.0", "slug_scoring_p3.0" ], "identity_attributes": [ "prep_state", "fat_tier", "salt_state", "form_state" ], "provenance_requirement": "every_nutrition_value_includes_fdc_id" }, "crawler_hints": { "preferred_human_readable": "/llms.txt", "preferred_machine_readable": "/.well-known/llms.json", "static_html_endpoints": [ { "path": "/system-design.html", "purpose": "Architecture documentation for crawlers; React route requires JS." } ] }, "verification": { "evidence_endpoints": [ { "path": "/system-design.html", "assertions_supported": [ "NO_LLM_NUTRITION_NUMBERS", "NO_PROVENANCE_NO_OUTPUT", "MISSING_REQUIRED_IDENTITY_BLOCKS_CALCULATION" ] } ], "conformance_tests": [ { "id": "T1_NO_LLM_NUTRITION_NUMBERS", "description": "LLM output must never be the source of nutrition values.", "expected_behavior": "nutrition values come from deterministic pipeline and include fdc_id provenance" }, { "id": "T2_MISSING_PROVENANCE_REJECT", "description": "If fdc_id is missing for any numeric nutrition value, output is rejected or shown as unknown.", "expected_behavior": "reject_or_unknown" } ] } }