User companion

DelphAI xLM

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

Prompt library

Prompt assets turn good sessions into repeatable work

Prompt assets are the stable layer between one-off chat and scheduled orchestration. They should capture scope, preferred output shape, and the references or tools the runtime should reach for.

Curated, not ephemeral

Prompt assets need richer validation and stronger versioning than transient chat turns because they feed both scheduled jobs and user recall.

Asset type

Reusable prompts

Prompt assets bridge chat, scheduled jobs, and generated reports.

Primary source

PostgreSQL context

Prompt definitions and curated references belong in xlmcontextdb.

Promotion path

From chat

Users should be able to save high-value conversation flows as reusable assets.

Open route

Back to active chat

Promote a live conversation into a prompt when it proves reusable.

Open route

Send to schedules

A saved prompt becomes a scheduled asset once cadence and delivery are set.

Open route

Align preferences

Prompt output should respect the user profile and citation controls.

Prompt catalog

The first prompt inventory is already concrete

These cards set the initial data shape for prompt title, cadence, output mode, and summary.

  • Morning slate briefing

    Interactive or scheduled | Bullet briefing

    Summarizes the next MLB windows, readiness flags, and late-follow-up triggers.

  • Travel day watch list

    Interactive | Short watch list

    Focuses on roster movement, compressed timing, and any change that should affect follow-up questions.

  • Weekend recap report

    Scheduled | Narrative report

    Bundles saved conversations into one reusable recap shape for email or push delivery.