Visualizers

Learn by moving things

Interactive explainers for the ideas behind retrieval, embeddings, and agents. A formula tells you what is true; dragging a vector shows you why. Each one also appears inside the article it belongs to, and runs entirely in your browser, so drag, poke, and break it.

Portrait of Sachin Gupta rendered in binary

A poison message that fails every time would redeliver forever and burn workers. After a set number of receives (the maxReceiveCount), it is moved to a dead-letter queue for inspection, and the main queue keeps flowing.

The ticket nobody can makeinteractive
main queueorder 999workerthrows errorreceive count: 1 / 5dead-letterqueue

Order 999 points at a deleted product, so the worker throws every time. Attempt 1 of 5: it fails and reappears on the queue.

Fan-out: one event, many listeners

From: Pub/Sub: One Event, Many Listeners

A publisher posts one event to a topic, and every subscriber gets its own copy at the same moment: email, SMS, push, in-app. The publisher never knows who is listening, so adding a new channel is just one more subscriber.

One event, many listenersinteractive
publisherorder shippedtopicorder-eventsemailSMSmobile pushin-app

An event happens: an order ships.

The same event, two patterns side by side. A work queue hands it to exactly one worker (the work gets done once). A pub/sub topic hands a copy to every subscriber (everyone who cares hears it).

Distribute vs broadcastinteractive
work queuequeueworker 1worker 2worker 3one message, one workerpub/sub topictopicsubscriber 1subscriber 2subscriber 3one event, every subscriber

The same event arrives at both a work queue and a pub/sub topic.

Visualizers — Sachin Gupta