User companion

DelphAI xLM

Standalone user-facing runtime for chat, saved prompts, schedule-driven briefings, and league-grounded sports assistance.

Agent profiles

User-tuned behavior without admin leakage

The user companion can offer distinct working styles, but those styles stay inside the user-facing solution family and do not expose operational or admin configuration.

What belongs here

Tone, response length, citation posture, and tool preferences belong in this screen. Model deployment, secret rotation, and operational analytics do not.

Profile count

Three core modes

Enough variety to expose value without turning settings into model-ops controls.

Audience

User-owned

These controls tune output style and tool posture, not administrator behavior.

Fallback

Capability-aware

Each profile still needs to disclose when a league or tool is blocked.

Profiles

The app can surface differentiated companion styles

These cards give orchestration a stable set of user-facing profile names and expectations to honor.

Fast, concise, scoreboard-first

Game Desk

Optimized for quick game-day answers, schedule pivots, and clear follow-up questions.

Structured, context-heavy, report-ready

Briefing Analyst

Turns prompt assets into polished briefings with caveats, citations, and next-run recommendations.

Sparse, trigger-driven, action-oriented

Notification Runner

Designed for scheduled prompts and follow-up alerts once queue-backed orchestration comes online.