The sandbox is empty. Load a sample dataset to create tables first.
The sandbox is empty. Load a sample dataset to create tables first.
pgvector is an open-source PostgreSQL extension for storing embeddings and running exact or approximate nearest-neighbour search with vector distance operators.
Use this playground to test vector columns, cosine or L2 distance, similarity ranking, and HNSW indexes using the pre-enabled pgvector extension.
This page sends your script to a hosted, network-isolated pgvector 0.8.6 runtime. Every Run executes the whole editor contents against a fresh database, then discards its database state and files. It does not run the database in your browser and it never connects to your own database.
Available output: HTML result tables, psql-style text, and CSV. The schema panel shows a bounded catalogue captured after successful execution. Large output may be truncated, and package installation, remote network access, persistent storage, and external database connections are not supported.
Build a pgvector script in the editor with syntax highlighting and editor assistance.
Run the complete script against a fresh isolated database sandbox.
Review rows, messages, errors, and execution time in the results panel.
Your draft stays in this browser, and Share creates a separate snapshot link.
CREATE TABLE items (
id BIGINT GENERATED ALWAYS AS IDENTITY PRIMARY KEY,
embedding VECTOR(3) NOT NULL
);
INSERT INTO items (embedding) VALUES ('[1,0,0]'), ('[0.8,0.2,0]');
SELECT id, embedding <=> '[1,0,0]' AS cosine_distance
FROM items
ORDER BY cosine_distance;The editor uses Monaco, the engine behind VS Code, with language-aware editing tools and familiar keyboard controls.
Write your query in the editor, then click Run (or press Ctrl/Cmd + Enter) to execute it.
Results appear instantly in the output panel below, along with any errors and how long the query took.
Nothing to install and no account to create — queries run in a hosted, isolated database sandbox.
Smart autocomplete and syntax highlighting for your query language
Real-time error underlines, bracket matching, and code folding
Multi-cursor editing plus find & replace (Ctrl/Cmd + F)
One-click formatting to tidy up long queries
Copy to clipboard and download your script
Built on Monaco — the same editor that powers VS Code
Run the whole editor script and see its native results
Output shows returned rows, messages, errors, and execution time
Start from a clean database every run, so experiments never clash
Switch between light and dark themes
Adjust font size and other editor settings
Toggle side-by-side or top/bottom layouts
Enter distraction-free fullscreen mode
Generate a shareable link to your exact query in one click — no login needed
Your work autosaves to your browser, so you can pick up where you left off
Copy your query to the clipboard or download it as a file
Run query: Ctrl/Cmd + Enter
Find & replace: Ctrl/Cmd + F
Open the full shortcut list from the keyboard icon in the toolbar