Preview build — in active development. Account creation is closed and no data you enter here is kept. The full product is open as a demo: try the demo, or talk to Sunsato.

Documentation / current runtime

The architecture behind the answer.

ADE is a modular TypeScript decision platform. These are the current runtime layers, accepted architecture decisions, and protected API surfaces.

Runtime composition

  • Aggregation compacts raw rows into gold, analysis-ready records.
  • Compression reduces and reshapes noisy input before analytics begin.
  • Clustering groups usage patterns with deterministic K-Means.
  • Regression forecasts demand, revenue, and risk deterministically.
  • Fuzzy logic produces explainable risk scores and fired-rule traces.
  • Scenario evaluation compares alternative decisions side by side.
  • Confidence scoring tells you how much to trust each result.
  • Recommendations turn scores into structured operational actions.
  • Decisions track lifecycle: suggested, approved, executed, measured.
  • Genetic optimisation searches for configurations only when explicitly invoked.

Accepted architecture decisions

ADR-001: LLM Schema Generation Layer

Accepted

LLM assistance is limited to schema mapping and never makes runtime decisions.

ADR-002: Prediction Before Decision

Accepted

Recommendations are grounded in a forecast rather than current state alone.

ADR-003: Supabase Persistence

Accepted

Datasets, analyses, and decisions persist per tenant with row-level security.

ADR-004: Bronze / Silver / Gold Data

Accepted

A layered data model keeps the runtime off raw source rows.

API surface

/api/analyze

POST

Run ADE analysis on a provided dataset and compression strategy.

/api/optimize

POST

Run the deterministic genetic optimizer when explicitly requested.

/api/datasets/upload

POST

Upload and validate a CSV, TSV, or JSON dataset scoped to a project.

/api/health

GET

Report startup readiness for the configured Supabase environment.