Software should understand the moment.
Aurora is an intent-aware execution platform: digital systems adapt in real time, based on mission, urgency, and context. Holmes is the inference engine: deterministic and heuristic algorithms and configurable business rules infer what people are trying to do. No generative AI. Deterministic intent inference that's fast, controllable, explainable, and private, without LLM cost or dependency. Your storefronts and workflows respond to the moment.
Aurora, Holmes, and the example templates are built by Purple Napkin.
Where human state affects outcome, software should adapt.
Commerce. Government. Healthcare. Travel. Wherever outcomes depend on understanding what people need right now, Aurora is the infrastructure. Static software becomes situational intelligence.
Digital Systems Are Static. Human Context is Not.
Mission states change. Your platform should too.
Reactive Interfaces
Interfaces that adapt based on live context.
Holmes Intelligence
Rules-based intent inference. Deterministic scoring, heuristics, and business rules. No generative AI.
Static Journeys
Traditional flows that Holmes can transform.
Dynamic Adaptation
Execution switches when mission state changes.
Holmes Infers Intent in Real Time
Deterministic intent inference that's fast, controllable, explainable, and private, without LLM cost or dependency.
From snack runs to urgent tasks, Aurora adapts instantly.
Marc missed his flight
Marc is rushed to buy new tickets. Holmes infers urgency from his typing speed, referral from the airline, recent search results, and location near the airport. Express checkout and next-available options are prioritised. No upsells. No friction.
Johanna's first time Paella
Johanna has never made Paella before. She starts searching for recipes and saffron. Holmes infers a first-time cooking mission and surfaces a beginner bundle: bomba rice, saffron, seafood, and a paella pan, with a recipe. One click. Everything she needs.
Jason's emergency gift
Jason did his monthly grocery shop this morning on the same site. Now he returns, different moment. Holmes doesn't suggest milk or cereal. It infers a gift purchase: flowers for his friend's birthday. Pairs chocolates (high stock) with the bouquet at a reduced rate. Quick. Personal.
Cart becomes a meal plan
She adds rice, peppers, and chorizo. Holmes offers two or three combo options (paella, stuffed peppers, or a tapas night). She picks one, and categories gently steer toward what’s missing. Same inference stack; clearer choice than a single guessed recipe.
Digital systems are static. Human context is not.
Customers, citizens, and users do not behave as profiles. They operate in states.
Snack run. Emergency repair. Visa renewal. Crisis rebooking. Project planning.
Traditional systems optimise history. Aurora optimises intent.
Built on Adaptive Infrastructure
Event-Driven Architecture
Domain events flow through a reliable outbox. Workflows and listeners react in real time.
Contextual Engine
Structured domain modelling. Holmes infers mission from clean behavioural signals.
Launch-Ready Templates
Holmes-native templates. Deploy store, site, or integrate Holmes into your stack.
Start Fast or Integrate Seamlessly
Launch with Intelligent Templates
Holmes-native templates. Start adaptive from day one. Store, site, or custom app.
Plug Into Your Existing Stack
Event adapters and behaviour hooks. Integrate Holmes without rebuilding.
Aurora provides everything Holmes needs
Situational intelligence infrastructure. Event-native. Measurable.
Holmes mission inference
Infers intent in real time. No asking. It notices.
Execution directives
Mark regions with data-holmes attributes; Holmes hides, highlights, or expands UI in real time based on inferred mission.
Predictive fragment prefetch
Holmes prefetches likely-next content before navigation. Checkout summaries, recommendations, and catalogue lists snap in instantly when users land. Measured via prefetch-hit analytics.
Event-native backbone
Every action is an event. Holmes consumes signals.
Structured domain model
Clean data Holmes can interpret.
Workflow branching
Workflows that adapt to mission state.
Observability
Mission distribution, confidence, time-to-completion, prefetch hit rate. Control Dashboard shows impact metrics.
Adapters & integration
Plug into commerce, email, existing stacks.
Multi-tenant & secure
Enterprise-ready. Production from day one.
Public App
End-user forms, views, and reports. App User role. No Studio access.
Holmes combos & recipes
Combo and recipe caching, mission-aware home, cart bundle discovery and picker, ingredient collages. APIs: contextual hints (with cart fingerprinting), combos-for-cart, select-combo, combo-products. Guardrails teach composability (e.g. egg noodles vs spaghetti for stir fry).
Holmes intent alignment
Mission-driven by default when confidence is high: active mission bar, mission-first command surface, catalogue narrowing. Users can dismiss mission chrome; the roadmap keeps signal-driven prioritisation, suggestions, bundles, and directives underneath. Quick actions, missions, shopping list templates.
Guardrail rules & micro-learning
Deterministic cooking intelligence. When cart matches conditions (e.g. stir fry + spaghetti), suggest egg noodles with a short "why." Contextual hints surface micro-learning insights.
Where human state affects outcome
Commerce first. Then government, travel, healthcare, enterprise. Aurora belongs wherever outcomes depend on understanding the moment.