Meilisearch source build and search demo on Debian

This tutorial sets up Meilisearch as a self-hosted search service on a standard Debian installation. The speaker describes it as a component LibreChat uses to search conversations. The walkthrough covers the search service itself; it does not demonstrate a complete LibreChat integration.

The installation starts with Git and curl, followed by Rust and Cargo. The speaker clones the repository and compiles a release binary. After finding that the test machine has one CPU core and 2 GB of memory, he increases it to four cores and 8 GB. He estimates roughly ten minutes for a build with more resources and says he has also built and run it on a Raspberry Pi with one CPU and 2 GB, though compilation took much longer. These are his reported experiences, not hardware requirements or general benchmarks.

Configuration includes a dedicated service user, database storage, dump and snapshot directories, and a TOML file. The example uses production mode, a master key, disabled analytics and disabled SSL options. A systemd unit starts the executable with that configuration and restarts it after failures.

A small HTML client adds movie documents to an index and sends searches as the user types. The speaker also tries Meilisearch's hosted movie demo, which reports four milliseconds for a Wicked search. That result illustrates the example rather than establishing performance for other datasets.