# Implementation Notes

The template already includes two reusable foundations that are directly relevant to this product. The `AIChatBox` component provides a production-ready chat surface with markdown rendering, assistant and user message styling, loading states, suggested prompts, and an input workflow that can be extended for Research Flow's verification bot. It does not yet support URL-specific actions, citations-aware rendering, or native file upload controls, so it should be adapted rather than replaced.

The `DashboardLayout` component provides an authenticated application shell with a persistent sidebar, responsive mobile behavior, and profile/logout handling. Its default navigation labels are placeholders and should be replaced with Research Flow-specific sections such as Live Feed, Verify, Saved, and History. Because the product is an authenticated intelligence workspace rather than a public marketing site, this layout is a suitable starting point for the main app experience.

The backend already exposes the platform credentials needed for this build through `ENV.forgeApiUrl` and `ENV.forgeApiKey`. A reusable `callDataApi(apiId, options)` helper exists in `server/_core/dataApi.ts`, which means live data ingestion should use that helper as the primary connection layer for Manus API Hub rather than introducing a separate client. The project also already includes `invokeLLM` for server-side model calls and `storagePut` for persisted uploads, which makes it feasible to build a verification pipeline that stores uploaded PDF and image artifacts, retrieves candidate records from the 2026 corpus, and enforces response generation from retrieved evidence only.