Launch scope
MLB-first
NBA remains visible but explicitly blocked until ETL catches up.
User companion
DelphAI xLM
Standalone user-facing runtime for chat, saved prompts, schedule-driven briefings, and league-grounded sports assistance.
User companion MVP
This app is the concrete deployment boundary for the DelphAI user companion. It gives the frontend, orchestration, and MCP work a stable place to meet without mixing in the admin Meta-XLM system.
The app now has an explicit route model, concrete user-facing surfaces, and a compile-time boundary that the remaining API and orchestration work can target.
Launch scope
MLB-first
NBA remains visible but explicitly blocked until ETL catches up.
Primary mode
Guided chat
Users move between live chat, saved prompts, and scheduled briefings.
Delivery boundary
Standalone app
This package can deploy behind /companion without coupling to the public site.
Route inventory
These are the pages the rest of the stack needs to support first: immersive chat, history, profiles, prompt assets, scheduled runs, and settings.
Chat surface
The main workflow balances live answers, citations, and follow-up prompts without collapsing into a support chatbot.
View chat canvasRetention
Users should be able to revisit prior briefings, recover decisions, and promote high-value conversations into reusable prompt assets.
See history modelAutomation
Morning recaps, late lineup checks, and weekend planning need the same orchestration boundary as interactive chat.
Open schedule controlsProfiles
The user companion must expose tone, league, notification, and citation preferences without leaking admin-only configuration.
Review agent profilesWhy now
If the package compiles cleanly, the route contract, UI structure, and user-companion naming are concrete enough to let the next tracks plug in without broad ambiguity.
Deployment boundary
An empty folder is now replaced with a shaped app that can be proxied behind/companion.
Audience safety
All copy, navigation, and env naming stay explicit about the app-user audience.
Integration target
MCP only needs to satisfy tool queries, and the orchestration API only needs to power these pages and flows.