Pumasi Booking Apache-2.0
Autonomous Commons · Multi-Model Verified

Open-source scheduling, built for absolute truth.

Connect Google Calendar and Microsoft 365 with real-time busy checks, zero double-booking guaranteed inside the database, and 1-minute self-hosting. Stop paying per-seat subscription rent.

No credit card needed Zero telemetry / ad trackers 1-Min Docker / Worker Deploy
booking.pumasi.ai/sarah/arch-review
SC
Sarah Chen
30 Min Technical Architecture Call

Review scheduling architecture, timezone arithmetic, and self-hosted PostgreSQL concurrency.

30 min America/Chicago Google Meet
MON31
TUE1
WED2
THU3
FRI4
09:00 AM
10:30 AM
02:00 PM ✓
03:30 PM
Google Calendar & Microsoft 365 Verified · Zero Conflicts
0 ms
Ambient State Drift
Pure mathematical engine in @pumasi/booking-core
100%
Database Exclusion Lock
PostgreSQL btree_gist + SQLite atomic triggers
250+
Verified Test Cases
Multi-model tested across DST gaps & timezones
$0 / Seat
Per-User License Cost
100% Apache-2.0 open source forever

Why Pumasi Booking is built differently

Most commercial scheduling tools use application-level checks that fail under concurrent bookings. Pumasi locks availability atomically inside the database engine.

Traditional SaaS / Naive App Logic

Check-Then-Act Race Condition

The web server checks if the time is free, pauses, and then sends an INSERT. Two concurrent clicks both pass the check and double-book your calendar.

User A & User B click 2:00 PM simultaneously
App checks DB: "Slot is available" (True for both)
💥 Double Booking Created (Data Corruption)
Pumasi Commons Architecture

SQL-Level Temporal Exclusion

Concurrency protection lives inside the database transaction. Overlapping intervals are physically rejected by index constraints, even under thousands of parallel requests.

User A & User B click 2:00 PM simultaneously
DB evaluates btree_gist interval lock
🛡️ Exactly 1 Confirms · Loser rejected cleanly

Engineered for reliability, privacy, and speed

Live Calendar Truth

Checks Google Calendar and Microsoft 365 busy times in real time before offering slots. If a provider is unreachable, it fails closed to protect your schedule from collision.

Google Calendar OAuth2 real-time sync
Microsoft 365 Graph API conflict check
Fails-closed network protection

Pure Engine Separability

Availability calculation in @pumasi/booking-core is a pure function: zero I/O, zero clock dependencies, byte-identical output. Extract the engine alone with git subtree split.

Spring-forward & fall-back DST precision
Daily caps evaluated on host's local date
Frozen multi-model acceptance test matrix

1-Minute Self-Hosting

Self-hosting is a first-class citizen. Spin up locally with Docker Compose and PostgreSQL, or deploy to Cloudflare Workers with Durable Objects for zero server maintenance.

docker compose up ready
Cloudflare Workers + Durable Objects build
Zero third-party telemetry or ad tracking

How it works

01

Set your availability

Define weekly working hours, custom buffer times between meetings, and event types (15m, 30m, 60m discovery calls).

02

Share your custom link

Send your public booking link (/username/event) directly to clients, embed it in emails, or link it on your website.

03

Automatic sync

When a booker selects a slot, real-time availability locks the time, and email confirmations with .ics invites are delivered instantly.

SEED PREVIEW · ALPHA

Built in the open by autonomous multi-model agents

Pumasi is an open commons. Every line of code is reviewed across competing model families (Claude, Gemini, Grok), active debts are publicly acknowledged, and rejected candidates stay on the record.