User companion

DelphAI xLM

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

Scheduled work

The companion should still work when the user is away

Scheduled briefings and follow-up jobs are part of the MVP because the user value is not limited to interactive chat. This screen defines the job inventory and delivery expectations.

Async boundary

This route is the frontend counterpart to a queue-backed worker pipeline. The runtime should expose schedule intent here, then hand execution to asynchronous orchestration.

Execution mode

Queued

Scheduled jobs should move through a queue-backed orchestration path, not the interactive request path.

Delivery targets

App, email, push

The UI owns user intent; the orchestration layer owns the actual delivery fan-out.

Failure posture

Explainable

The user should see why a job skipped or degraded, especially when capability is blocked.

Open route

Choose source prompt

Every scheduled job should start from a reusable prompt asset.

Open route

Adjust quiet hours

Delivery windows and channels belong to user settings, not prompt definitions.

Open route

Review generated runs

Generated artifacts should feed back into recall and saved reports.

Job inventory

The queueable jobs are already small enough to implement

This first pass keeps the async surface tight: daily recap, late lineup follow-up, and a weekly look-ahead report.

  • Morning companion recap

    Daily at 07:15 local | Email and in-app

    Queues a briefing before the earliest MLB window and stores the generated artifact.

  • Late lineup sweep

    Triggered 60 minutes before first pitch | Push notification

    Looks for readiness or lineup movement that materially changes the earlier briefing.

  • Sunday look-ahead

    Weekly on Sunday evening | In-app report

    Packages a reusable prompt with the next week of slate planning and known blockers.