Cost
Benchmarks, estimates and suggest-cuts across fashion, cosmetics and jewelry spend.
Vihana · Intelligence Engine
The cognition layer of the Mystique Earth ecosystem.
The analytics, ML and reasoning brain behind Mystique Earth — cost, chat, casting, discovery, forecasting, anomaly, research and ambient assist, now with a cognition layer of patterns, reasoning, neural scoring, self-learning and composable pods. Reads the ecosystem read-only; Node stays the system of record.
Version1.0.1
Environmentproduction
Uptime4d 13h
Endpoints266
Capabilities
Everything the service powers
A production surface built for the Mystique Earth ecosystem — composable, observable, and secure by default.
Chat
KB-first answers spoken through VI's own brain — the native mind and our own generative heart, no third party — learning every resolved turn back into the knowledge base.
Casting
Model and talent fit-scoring with ranked casting shortlists for every show and campaign.
Discovery
Rising talent, power graphs and lookalike discovery across the ecosystem graph.
Forecast
Trend and show-success forecasting to plan collections and events with confidence.
Anomaly
Spend, velocity and funnel anomaly detection with isolation forests and reasons.
Research
Live rate and vendor research — hotels, logistics and vendors — feeding budget-aware decisions.
Assist
The ambient Vihana orb: ranked, persona-rendered suggestions on every screen.
Patterns
Association-rule and co-occurrence mining (a light in-house apriori) over baskets, behaviour and casting.
Reasoning
A reasoning layer that chains engine outputs into explained, what-if and causal narratives.
Neural
Neural scoring (scikit-learn MLP) extending lead, churn and success models beyond linear baselines.
Self-learning
Online learning (SGD partial-fit) that updates models from live feedback without full retrains.
Pods
Composable cognition pods that orchestrate patterns, reasoning, neural and learning into one answer.
API Surface
Endpoint groups
Explore the live routes. Full request/response schemas live in the interactive docs.
POST
Ingestion/ingest · behavioural signals, append-only
4 routesCRUDFeatures/features · per-user feature vectors
2 routesPOSTScoring/score · lead / propensity, churn, lookalike
3 routesCRUDSegments/segments · clustering & audiences
4 routesGETInsights/insights · NLP, trends, dashboards
3 routesPOSTAnomaly/anomaly · spend / velocity / funnel outliers
4 routesCRUDCampaigns/campaigns · consent-gated marketing orchestration
3 routesPOSTWhatsApp/whatsapp · transport via the mother API
3 routesCRUDCost/vi/cost · benchmarks, estimates, suggest-cuts
13 routesPOSTChat/vi/chat · KB-first conversation on VI's own brain
7 routesPOSTCasting/vi/casting · fit scoring & casting shortlists
5 routesGETDiscovery/vi/discovery · rising, power graphs, lookalikes
6 routesPOSTForecast/vi/forecast · trend & show-success forecasting
3 routesGETResearch/vi/research · live rate & vendor research
8 routesPOSTAssist/vi/assist · ambient, persona-rendered suggestions
3 routesPOSTPatterns/vi/patterns · association-rule & co-occurrence mining
4 routesPOSTReasoning/vi/reason · explained, what-if & causal chains
5 routesPOSTSelf-learning/vi/learn · online learning from live feedback
6 routesGETModels/vi/models · model & pod registry
4 routesGETPods/vi/pods · composable cognition pods
6 routes
Part of the ecosystem
One spine. Many engines.
Vihana · Intelligence Engine is a distributed engine behind a single unified gateway — sharing identity, catalog, finance, and event context with every other Mystique Earth service.
Requests are authenticated once and routed to the right domain engine, so fashion, cosmetics, and jewelry lines run on the same trusted rails.
412
Models
293
With photos
27
Celebrities
21,728
Events run
1.0.1
Version
4d 13h
Uptime
52
MCP tools
The Self-Reliant Brain
One mind — with its own intelligence, its own model, and its own MCP. It thinks and answers on its own; a third party is only ever the last resort, and can be switched off entirely.
1 · PERCEIVE
Reads who is speaking and the live warehouse
instant
the ground it stands on
2 · NATIVE
Facts & every safety refusal answered on its own
instant · 0 model calls
14 intents · learned model
3 · HEART
Open-ended & empathetic replies, streamed token-by-token
our own model
qwen3:1.7b · standby → native
4 · COUNCIL
The सभा deliberates weighty matters, then rules
on demand
4 luminaries · 30 lenses
Self-reliant
Facts, safety & social answered locally — 14 native intents, zero third-party calls.
native + nlu
Its own model
The generative heart runs on our own server (Ollama · qwen3:1.7b). BRAIN_LOCAL_ONLY forbids any third party.
heart: standby → native
Its own intelligence
A learned intent model (scikit-learn) reads every message — trained on 66 of our own scenarios.
model warming
Its own MCP
The brain is exposed as tools — brain_ask · brain_intent · brain_state — for the whole ecosystem.
3 brain tools
Model routing ● standby
role → model, availability-aware
voice—
council—
fast—
embed—
Audition funnel — S4 141 confirmed/216 · S3 95 confirmed/103 · S2 0 confirmed/88