March 2024 Summaries
3 posts from Replit
Filter
Month:
Year:
Post Summaries
Back to Blog
Today, Replit has released the first version of rippkgs, a CLI utility for indexing and searching Nix expressions, designed to improve search performance and accuracy for its users. The motivation behind rippkgs is to provide better tools for users who are not familiar with Nix, as existing solutions such as nix-env, nix search, and search.nixos.org have limitations that hinder their effectiveness. To achieve this, rippkgs was created with a focus on speed and correctness, using an index database generated by its sibling CLI, rippkgs-index. The tool allows users to quickly search the available Nix expressions, providing accurate results without the need for manual evaluation of attributes, and can be used in conjunction with Replit's command-not-found handler to provide better information about installed packages.
Mar 29, 2024
722 words in the original blog post.
Today, Replit is launching Scheduled Deployments, a service that allows users to schedule their applications to run at predetermined time intervals seamlessly. This feature addresses the difficulty of configuring schedulers on Replit, enabling developers to focus on building without unnecessary overhead and costs. The new service has several cost levers, including per-second costs for running the machine, scheduler costs, and outbound data transfer charges. Developers can build a variety of applications with Scheduled Deployments, such as automated analytics reporting and system health checks, using a flexible credit system that comes with Replit Core membership. Future plans include adding configuration options and a robust email notification feature to enhance user experience.
Mar 26, 2024
442 words in the original blog post.
At Replit, the company aims to ensure a high session success rate, with a goal of always successfully connecting legitimate incoming requests from clients to their target Repl. To achieve this, they introduced a new reverse WebSocket proxy called Eval, which abstracts away intermittent infrastructure failures by sitting between the client and Conman VMs. This separation of concerns addresses several issues in their previous architecture, including frequent Conman updates causing multiple disconnect events for clients, Conman's sole job being container hosting making autoscaling complicated, and diagnosing session failure rates being harder due to shared functionality. Eval provides a more reliable connection experience with fewer reverse proxy failures, easier observability, and scalability based on active client connections.
Mar 18, 2024
1,042 words in the original blog post.