vLLM's Automatic Prefix Caching feature stores and reuses the key-value (KV) cache of common prefixes across requests. This avoids redundant recomputation of identical prompt prefixes, reducing latency and computational overhead, especially for shared system prompts or multi-turn conversations.
#documentation
30 items
Scalar has revamped its documentation for ScalarDB and ScalarDL, aiming to help users start faster, stay updated with changes, and navigate cleaner, more organized docs. The update includes improved structure, clearer guidance, and better version management for developers.
LMDB 1.0 is a memory-mapped, high-performance transactional database library with a B-tree-based structure, offering full ACID semantics, multi-reader and single-writer concurrency, and a simple key-value API. It is designed for efficiency and reliability, supporting zero-copy reads, crash recovery, and extremely small binary size.
Wikicube is a tool that instantly generates a wiki page for any GitHub repository. The featured example highlights the Granian framework, a Python web server alternative to Uvicorn and Gunicorn.
Hica is a functional-first language emphasizing immutability, pure functions, and algebraic data types, inspired by Haskell and OCaml. It supports pattern matching, higher-order functions, and type inference for expressive code. Core concepts include function composition, recursion, and lazy evaluation.
None of six major aerospace documentation portals are "AI agent-ready," according to a student-built scoring engine called AeroScore. The best score was ICAO with 69/100; no site had an llms.txt or supported URL variants. The author warns this hinders potential productivity gains from AI in aerospace.
MkDocs 2.0 introduces significant updates focused on performance, plugin architecture, and configuration improvements. The new version aims to streamline documentation workflows while maintaining backward compatibility. Key changes include faster build times and a more modular plugin system for enhanced customization.
A free Chrome extension called AI Score uses AFDocs in the backend to analyze how well AI agents can read, navigate, and use a website's documentation. It provides a numeric and letter grade score, with detailed feedback that can be copied and sent to an AI agent for fixing issues.
Engineering teams lose context as knowledge gets scattered across Confluence, Notion, Git, and Slack. The real bottleneck isn't code generation but the loss of the "why" behind decisions, leading to slow onboarding and AI suggestions that are correct locally but wrong architecturally.
Leanstral 1.5 is a Mixture-of-Experts (MoE) model with approximately 26 billion total parameters and 3 billion active parameters, designed for efficient inference on consumer GPUs. It outperforms larger models like Mistral 7B and Llama 3.1 8B on standard benchmarks while offering faster generation speeds and a 128k-token context window. The model is available under the Apache 2.0 license.
Claude Tag
3.0Claude Tag is a feature that allows users to tag and organize conversations with Claude for easier retrieval and management.
The article argues that Markdown's primary value lies in being human-readable and writable in plain text, not in its rendering to HTML. It contends that focusing on visual output misses the point, as Markdown is designed for writing and editing efficiency, not for producing pixel-perfect documents.
The Guardian framework treats software architecture as an executable constraint that can be enforced programmatically, rather than relying on documentation alone. It allows developers to define and verify architectural rules as code, ensuring compliance automatically during the development process.
Rhombus is a new programming language that builds on Racket's foundation, offering a cleaner syntax while maintaining access to Racket's libraries and tools. The getting-started guide covers installation, basic syntax, and how to write and run simple Rhombus programs.
A new project introduces Yocto/BitBake skills that enable AI agents to consult official documentation, aiming to improve their accuracy and reliability when working with embedded Linux build systems.
The article outlines best practices for writing effective software design documents, emphasizing clarity, structure, and audience awareness. It explains how to define goals, describe system architecture, and document trade-offs with decisions. The guide aims to help engineers create documents that facilitate alignment, review, and implementation.
The jank programming language book provides documentation and reference material for jank, a new programming language designed to be a pragmatic, functional-first language that compiles to native code and leverages Clojure-inspired syntax with C++ performance.
Chris Krycho outlines his ongoing maintenance work on "The Rust Programming Language" book, focusing on updates for new Rust editions, fixes for inaccuracies, and improvements to clarity and consistency across chapters.
This guide outlines the key components of an effective software design document, emphasizing clarity, structure, and purpose. It explains how to define the problem, propose a solution, detail architecture, and address trade-offs to align teams and guide implementation.
The article provides guidance on writing effective software design documents, covering key components such as context, goals, design decisions, trade-offs, and open questions. It emphasizes structuring the document to clearly communicate technical decisions and rationale to team members and stakeholders.
The AGENTS.md file in the Ghostty repository provides guidelines and instructions for AI agents and automated tools contributing to the Ghostty terminal emulator project, outlining expected behavior, coding standards, and communication protocols.
The article describes how the author's years of experimentation with various content creation tools ultimately led them to adopt Markdown as the most efficient and durable format for writing, especially for technical documentation and web content.
Stewie Reflect is a new tool that generates owner's manuals for software repositories, linking claims and documentation directly to the relevant source code. It aims to make codebases more understandable by providing human-readable explanations tied to specific code sections.
The author criticizes Auth0's poor documentation and provides a guide on manually authenticating a JWT idToken after user authentication. Only the idToken is needed, not the accessToken.
Show HN: Dk
2.0Dk is a command-line tool for organizing and documenting command line tools to make them easier to discover.
Treedocs is a documentation tool that automatically detects and alerts when documentation becomes outdated or stale relative to the code it describes, helping maintain accuracy over time.
The page provides release notes and documentation for system prompts used with Claude API, detailing updates, changes, and best practices for effectively configuring system prompts to guide model behavior across different use cases.
Brady Stroud shares how to create a private documentation vault for AI write-ups that can be published to a URL without exposing client-specific content, using static site generation and selective publishing techniques.
PreactPress is a documentation site generator designed for Preact projects, styled similarly to VitePress.
The article discusses the subjective nature of judging "beautiful" documentation, explores growing fatigue with AI-generated content in technical writing, and critiques the rise of "tool slop"—poor-quality output produced by over-reliance on software tools without critical human oversight.