Build a vector search engine from scratch. Create a vector index with exact search, filters and a simplified HNSW graph for approximate search.
Build it in your language
The tester treats your program as a black box: any language works. These are just the most convenient ones for this project.
Project in preparation: we are completing and verifying the implementation path and its tests. The capabilities below describe the project goal; starting the project is not enabled yet.
Vector search retrieves items by mathematical proximity instead of text matching. Here you will build the engine beneath many semantic retrieval systems.
The program receives NDJSON commands, keeps identified vectors, and answers queries with deterministic ordering. You will first establish an exact baseline and then build a measurable approximate index.
5 stages across 3 phases. Every stage comes with its guide and its tests: you move on when they pass.
Define the store, distance metrics, and a trustworthy exhaustive search.
Add filters, replacements, and deterministic query rules.
Build and evaluate a simplified HNSW graph to accelerate retrieval.
Build a key-value database from scratch. Implement a storage engine with a write-ahead log, indexes and compaction to understand data persistence.
Go · Rust
Build a Redis-compatible server from scratch. Implement RESP, an event loop and key expiry, then connect with the official redis-cli.
Go · Rust · Python
Build a SQLite reader from scratch. Open a real database file, walk its B-trees and run SELECT queries by decoding the documented binary format.
Go · Python · Java