Enrichment

Enrichment is rmbr's two-phase data capture model. You capture raw input quickly, then add structured detail later — either manually or with AI assistance.

Two-Phase Data Capture

The core idea: capture first, organize later.

1. Raw phase — jot down a thought as fast as possible. No required fields beyond the raw text.
2. Enrichment phase — when you have time (or an AI assistant does it for you), add structured fields like title, priority, domain, tags, and links to goals.

This removes friction from the capture moment while still producing well-structured data.

Enrichment Status

Every entity carries an enrichment_status field:

Status	Meaning
raw	Just created, only has unstructured input
enriched	Structured fields have been filled in

Enrichment Fields Per Module

Each module defines its own enrichment fields:

Todos

• title — concise summary of the task
• priority — low, medium, high, or critical
• due_date — deadline
• goal_id — link to a parent goal

Goals

• title — goal name
• quarter — Q1, Q2, Q3, or Q4
• year — target year
• kpis — JSON array of key performance indicators

Kudos

• direction — given or received
• person — who the kudos involves
• summary — what happened
• context — additional detail
• goal_id — link to a related goal

TIL (Today I Learned)

• title — what you learned
• content — detailed explanation
• domain — subject area
• tags — JSON array of topic tags

Study

• title — topic name
• domain — subject area
• goal_id — link to a related goal

Single-Step Creation

MCP create tools accept enrichment fields directly alongside the raw input. This means an AI assistant can create a fully enriched entity in a single call instead of the two-step create-then-enrich flow.

Example: a user says:

I need to fix the login bug by Friday, high priority

The AI calls the todo create tool with:

{
  "raw_input": "I need to fix the login bug by Friday, high priority",
  "title": "Fix login bug",
  "priority": "high",
  "due_date": "2026-03-27"
}

The resulting todo is created with enrichment_status: 'enriched' immediately — no second step needed.

Enrichment Prompts

Each module provides an enrichment prompt that guides AI assistants on how to extract structured data from raw input. These prompts are defined in src/core/prompts.ts and tell the AI what fields to look for, what formats to use, and what to leave blank when information is missing.

Shared Enrichment Logic

The utility function enrichEntity() in src/shared/enrichment.ts handles the common pattern of updating a record's structured fields and flipping its enrichment_status from raw to enriched. Individual module services call this utility so enrichment behavior stays consistent across the codebase.