Platform architecture
The public data flow from owner input to one market relationship or a Demand record.
Component map
Analysis pipeline
The analysis service separates interpretation from construction. One model call parses the business and owner intent; candidate preparation then runs from the selected market generation. A second bounded reasoning step can compare only the prepared shortlist. The server re-grounds contract facts, applies side guards, calculates economics and rebuilds the public brief through explicit schemas and allowlists.
Live analysis moves through four public-safe stages: intent detection, exposure translation, candidate construction and the final brief. Its stream carries progress and completed public objects without exposing private model payloads; the browser can alternatively replay a precomputed case.
Static application and dynamic services
The Next.js application exports static pages. Dynamic behavior lives behind server functions with explicit methods, response schemas, cache rules and origin policy. This keeps the product usable for keyless examples while allowing live analysis, Demand capture, receipts and partner sync to fail independently.
Data stores
| Store | Public role |
|---|---|
| Bundled case and market data | Reproducible examples and safe pre-activation or failure fallback. |
| Immutable market generations | Current analysis and Sports source after activation. |
| Private run capture | Optional operational record; never required for product completion. |
| Private Demand evidence | Customer signals waiting for cohort publication. |
| Partner-safe Demand ledger | Published records, generations and change checkpoints. |
| Private receipt table | Validated public receipt payloads behind opaque identifiers. |
The public browser never receives direct database credentials or unrestricted table access.