agent-db

name	agent-db
description	Manage structured data in Railway PostgreSQL — emails, calendar events, tasks, facts, reminders, projects, and notifications. Use when tracking emails, scheduling reminders, managing tasks/todos, recording important facts from conversations, syncing calendar events, generating daily digests, or querying historical data.
type	shell-script
category	data
entry	scripts/email-sync.py

Agent DB

Shared Railway PostgreSQL database for structured data management. Each agent uses agent_id to isolate data.

Setup

• Credentials: ~/.agent-bridge/runtime/credentials/railway-db.json
• Python: import psycopg2 (already installed)
• Agent ID: use your own agent_id in all queries

Quick Access

import os
import sys
import psycopg2

sys.path.insert(0, os.path.expanduser("~/.agent-bridge/runtime/scripts"))
from creds import load_creds

def get_db():
    c = load_creds("railway-db.json")
    return psycopg2.connect(
        host=c["db_host"], port=c["db_port"],
        dbname=c["db_name"], user=c["db_user"], password=c["db_password"]
    )

Tables

emails

Track and manage email messages synced from Gmail.

• agent_id, account, message_id, sender, subject, snippet, received_at
• importance (critical/high/medium/low/spam), category (legal/business/personal/newsletter/spam)
• status (new/notified/handled/ignored/archived), notified_at, notified_to, agent_note

events

Calendar events synced from Google Calendar.

• agent_id, calendar_id, event_id, summary, description, start_at, end_at, location, attendees, all_day
• reminder_30min_sent, reminder_1h_sent, briefing_included

tasks

Track todos, requests, and action items from conversations.

• agent_id, title, description, status (open/in_progress/done/cancelled)
• priority (critical/high/medium/low), category, assigned_to, due_date

facts

Record important facts, decisions, and preferences from conversations.

• agent_id, category (person/project/decision/preference/deadline)
• subject, content, source (conversation/email/calendar), source_date, expires_at

reminders

Schedule reminders for users.

• agent_id, title, description, remind_at, sent, sent_at, target
• recurring (daily/weekly/monthly/null)

projects

Track ongoing projects and their status.

• agent_id, name, description, status (active/paused/completed/archived), owner

notifications

Log all sent notifications to prevent duplicates.

• agent_id, type (email/calendar/system/walk/weather), reference_id, sent_to, channel, message

daily_digests

Track daily summary reports sent to users.

• agent_id, date, type (morning/evening), sent_to, content, email_count, event_count

agent_state

Key-value store for agent operational state.

• agent_id, key, value (JSONB)

Scripts

Email Sync

Sync Gmail -> DB: python3 ~/.agent-bridge/runtime/skills/agent-db/scripts/email-sync.py [hours] [agent_id]

• Gmail API backend is gws
• default lookback: 2 hours
• default agent_id: main

DB Query Helper

Quick queries: python3 ~/.agent-bridge/runtime/skills/agent-db/scripts/db-query.py <query_type> [agent_id]

• new-emails
• today-events
• open-tasks
• pending-reminders
• stats

Privacy

• Always filter by your own agent_id
• Never query another agent's data
• Credentials file is shared but data is logically isolated by agent_id

Usage Guidelines

• Email arrives -> sync, judge importance, update status, notify if needed
• Reply needed -> set reply_needed=true, remind after 24h if no action
• Meeting invite in email -> set has_meeting_invite=true, check calendar
• User says "remember X" -> insert into facts
• User says "remind me" -> insert into reminders
• User assigns task -> insert into tasks

Example Queries

SELECT * FROM emails WHERE agent_id='main' AND status='new' AND importance IN ('critical','high');

SELECT * FROM events
WHERE agent_id='main'
  AND start_at::date = (NOW() AT TIME ZONE 'Asia/Seoul')::date
ORDER BY start_at;

SELECT * FROM tasks
WHERE agent_id='main' AND status IN ('open','in_progress')
ORDER BY priority;