🛠️ Hacker News Tools

[Other] Hyperpb: 10x faster dynamic Protobuf parsing that's faster than generated code

[Other] FastVLM: Efficient Vision Encoding for Vision Language Models

[Other] Show HN: The missing link of a bookstore's tech stack Hi HN!<p>I built Bookhead because I used to work as a bookseller and I wasn&#x27;t happy with the software options when I decided to sell my own collection online (with the hopes of one day growing so I can open my own brick &amp; mortar). So I decided to make my own bookselling app...a classic hacker distraction.<p>Bookhead has two main parts: 1. an inventory management app that allows a bookseller to list their books anywhere they want to sell books (like Squarespace, Biblio, eBay, Shopify (coming soon!), etc) 2. an e-commerce platform with a CMS for selling books and letting a store control their online brand<p>I have a very exciting roadmap that I&#x27;m not ready to fully reveal, but it&#x27;s all based on books. I&#x27;m building a sorta Zapier-like platform for independent booksellers. Everything is so fragmented and disconnected, which makes it hard for booksellers to do their work. I&#x27;m hoping to change that. I have a blog post that lays out my vision here: <a href="https:&#x2F;&#x2F;bookhead.net&#x2F;blog&#x2F;fragmented&#x2F;" rel="nofollow">https:&#x2F;&#x2F;bookhead.net&#x2F;blog&#x2F;fragmented&#x2F;</a><p>The current iteration is like &quot;data engineering as a service for books.&quot; A book is a powerful thing. I&#x27;m hoping to give a bookstore everything they need to sell books online. Inventory, e-commerce, marketing, etc. It&#x27;s a crowded market but I&#x27;ve had fun making the bookselling app that I believe should exist.<p>If you know any booksellers, please let them know about this! I&#x27;m onboarding my first customer right now and the biggest bottleneck is the other bookselling software providers, despite my intention to collaborate instead of compete. It&#x27;s frustrating to wait for two weeks for a point of sale provider to setup an integration. It&#x27;s almost like they don&#x27;t care about their customers. Some providers even require ethernet cables for their software...still partying like it&#x27;s 1999. Perfect for early-adopter booksellers frustrated with current tech who understand the power of automation.<p>I&#x27;m currently looking for funding so I can focus on this full-time. My biggest problem right now is time (aka money) because I have to sell my time to make rent etc, and can&#x27;t focus on this project like I need to. I&#x27;ve gotten good validation from booksellers and other technically savvy folks in the industry (I&#x27;ve heard from two different companies that they&#x27;ve considered building something like this), so I believe I have something valuable. I&#x27;m not interested in funding from somebody who doesn&#x27;t share my love for books or doesn&#x27;t support my mission: help people use technology to promote literature. I believe that literature is one of humanity’s most prized creations, and we can use technology as a tool to keep this gift alive.<p>Please email me at sam@bookhead.net if you know of booksellers who might want to be an early adopter, or know of any funding opportunities that might be a good fit.

[Other] Building better AI tools

[Other] Show HN: Self-updating MCP server for official pip, uv, poetry and conda docs

[Other] Show HN: Bskysrch – An Advanced Search for Bluesky Hi HN,<p>I missed proper search on Bluesky, like Twitter used to have, so I built this. You can filter posts and profiles by keyword, handle, time and other search operators.<p>Would love feedback or ideas for what to add next.

[Other] Manticore Search: Fast, efficient, drop-in replacement for Elasticsearch

[Other] Reverse engineering GitHub Actions cache to make it fast

[Other] Show HN: Tool to discover bloggers, trending blog topics, and weekly summaries

[API/SDK] CAMARA: Open-source API for telecom and 5G networks

[Other] Show HN: Any-LLM – Lightweight router to access any LLM Provider We built any-llm because we needed a lightweight router for LLM providers with minimal overhead. Switching between models is just a string change : update &quot;openai&#x2F;gpt-4&quot; to &quot;anthropic&#x2F;claude-3&quot; and you&#x27;re done.<p>It uses official provider SDKs when available, which helps since providers handle their own compatibility updates. No proxy or gateway service needed either, so getting started is pretty straightforward - just pip install and import.<p>Currently supports 20+ providers including OpenAI, Anthropic, Google, Mistral, and AWS Bedrock. Would love to hear what you think!

[Other] Show HN: Go Command-streaming lib for distributed systems (3x faster than gRPC) I created cmd-stream-go, a high-performance client-server library based on the Command Pattern, where Commands are first-class citizens.<p>Why build around Commands? As serializable objects, they can be sent over the network and persisted. They also provide a clean way to model distributed transactions through composition, and naturally support features like Undo and Redo. These qualities make them a great fit for implementing consistency patterns like Saga in distributed systems.<p>On the performance side, sending a Command involves minimal overhead — only its type and data need to be transmitted. In benchmarks focused on raw throughput (measured using 1, 2, 4, 8, and 16 clients in a simple request&#x2F;response scenario), <i>cmd-stream&#x2F;MUS</i> (<i>cmd-stream&#x2F;Protobuf</i>) is about 3x (2.8x) faster than <i>gRPC&#x2F;Protobuf</i>, where <i>MUS</i> is a serialization format optimized for low byte usage. This kind of speedup can make a real difference in high-throughput systems or when you&#x27;re trying to squeeze more out of limited resources.<p>By putting Commands at the transport layer, cmd-stream-go avoids the extra complexity of layering Command logic on top of generic RPC or REST.<p>The trade-offs: it’s currently Go-only and maintained by a single developer.<p>If you’re curious to explore more, you can check out the cmd-stream-go repository (<a href="https:&#x2F;&#x2F;github.com&#x2F;cmd-stream&#x2F;cmd-stream-go">https:&#x2F;&#x2F;github.com&#x2F;cmd-stream&#x2F;cmd-stream-go</a>), see performance benchmarks (<a href="https:&#x2F;&#x2F;github.com&#x2F;ymz-ncnk&#x2F;go-client-server-benchmarks">https:&#x2F;&#x2F;github.com&#x2F;ymz-ncnk&#x2F;go-client-server-benchmarks</a>), or read the series of posts on Command Pattern and how it can be applied over the network (<a href="https:&#x2F;&#x2F;medium.com&#x2F;p&#x2F;f9e53442c85d" rel="nofollow">https:&#x2F;&#x2F;medium.com&#x2F;p&#x2F;f9e53442c85d</a>).<p>I’d love to hear your thoughts — especially where you think this model could shine, any production concerns, similar patterns or tools you’ve seen in practice.<p>Feel free to reach me as <i>ymz-ncnk</i> on the Gophers Slack or follow <a href="https:&#x2F;&#x2F;x.com&#x2F;cmdstream_lib" rel="nofollow">https:&#x2F;&#x2F;x.com&#x2F;cmdstream_lib</a> for project updates.

[CLI Tool] Show HN: Giti – Natural Language to Git Commands with Local LLM Hi HN,<p>I built Giti, a command-line tool that converts plain English into actual Git commands using a fast, local language model (Qwen2.5-Coder, ~1 GB).<p>Example:<p>Input: giti &quot;undo last commit&quot;<p>Output: git reset --soft HEAD~1<p>No internet required after setup. No API keys. You can also run it in an interactive shell to chain commands naturally.<p>Key features: - Natural language to Git translation - Local LLM powered by Qwen2.5-Coder in GGUF format - Works fully offline after model download - Dry-run mode to preview commands before running - Interactive shell mode for session-based workflows - Context file support to teach Giti your custom Git habits<p>Quick install: - Clone the repo - Install llama-cpp-python - Add giti to your PATH - Download the 1GB model from HuggingFace - Run giti &quot;your query.&quot;<p>You can also enhance its accuracy using context files in a simple Q&amp;A format like:<p>USER: How to start new feature? BOT: git checkout main &amp;&amp; git pull &amp;&amp; git checkout -b feature&#x2F;&lt;name&gt;<p>This lets Giti learn your workflow and generate project-specific Git commands.<p>Thanks for checking it out.

[Other] Show HN: Dyad – build AI apps locally, no cloud

[Other] A conceptual overview of asyncio

[IDE/Editor] Show HN: Lotas – Cursor for RStudio Hey HN! We’re Jorge and Will from Lotas (<a href="https:&#x2F;&#x2F;www.lotas.ai&#x2F;" rel="nofollow">https:&#x2F;&#x2F;www.lotas.ai&#x2F;</a>), and we’ve built an AI coding assistant into RStudio (think Cursor for RStudio).<p>RStudio is used by about 2 million data scientists and academics, but they currently lack a coding assistant within their IDE. Developers in other environments benefit from tools like Cursor and Windsurf, but R users don’t have any equivalent tools to speed up their workflow. Since ~80% of R programmers prefer to use RStudio over other IDEs like VSCode to write R code, we figured a tool like this one could be quite useful.<p>Both of us were PhD students at Harvard. Jorge was in the biophysics program and Will was in the biostatistics program where most people used RStudio every day. We saw how integrated code assistants were taking off in other IDEs, but we noticed that the RStudio integrations were still lagging far behind. Many R users were copying and pasting code from ChatGPT to build their workflows, and this was clearly slow and fragile.<p>To bring the Cursor-like experience to RStudio users, we built Rao (<a href="https:&#x2F;&#x2F;www.lotas.ai&#x2F;" rel="nofollow">https:&#x2F;&#x2F;www.lotas.ai&#x2F;</a>): a fork of RStudio with an embedded AI assistant that is aware of the user’s local context (both files and variable environment), can read and write files, can run code or commands, and can interpret textual or visual output. It works with any of the file formats already in RStudio (R, notebooks including RMDs and QMDs, Python, Stan, etc.), allowing R programmers to iteratively perform entire data analyses inside their preferred IDE.<p>Other AI data science tools are either (1) built on the web or in environments people don’t already use, (2) are completely focused on python notebooks, or (3) are weak package-based assistants with limited functionality. Rao is exactly like the RStudio IDE that millions of data scientists already use, but it incorporates a powerful AI assistant and works with all the standard file types.<p>You can download Rao at <a href="https:&#x2F;&#x2F;www.lotas.ai&#x2F;download" rel="nofollow">https:&#x2F;&#x2F;www.lotas.ai&#x2F;download</a>, watch our demo on the homepage (<a href="https:&#x2F;&#x2F;www.lotas.ai&#x2F;" rel="nofollow">https:&#x2F;&#x2F;www.lotas.ai&#x2F;</a>), and work through some example use cases on our GitHub (<a href="https:&#x2F;&#x2F;github.com&#x2F;lotas-ai&#x2F;rao&#x2F;tree&#x2F;main&#x2F;demos">https:&#x2F;&#x2F;github.com&#x2F;lotas-ai&#x2F;rao&#x2F;tree&#x2F;main&#x2F;demos</a>). We have a one-week free trial (no card required) and provide 500 queries&#x2F;month for $20&#x2F;month after that. We’d love to hear feedback from the HN community to make Rao as useful as possible! You can reach us at founders@lotas.ai.<p>P.S. We have zero data retention (ZDR) agreements with OpenAI and Anthropic, but we currently recommend users do not input sensitive or regulated data like PHI into Rao until we sign BAAs with both model providers. For more information on our security practices, please visit the security page on our website <a href="https:&#x2F;&#x2F;www.lotas.ai&#x2F;security" rel="nofollow">https:&#x2F;&#x2F;www.lotas.ai&#x2F;security</a>.

[Other] Show HN: MCP Jetpack – The easiest way to get started with MCP in Cursor My friends and I built MCP Jetpack which makes it really easy to start using MCP. It’s an MCP server that automatically finds and executes the right tools needed to accomplish your task without having to manage MCP servers for each service (GitHub, Linear, Atlassian, Notion etc.). Once you add it to Cursor (or any other AI app with MCP support), Cursor instantly gets access to a growing library of remote MCP servers without any extra setup. For services that require authentication, you will be asked to login the first time you ask your AI to interact with that service.<p>Two problems we are trying to solve:<p>Friction - Normally if you want to give Cursor access to GitHub, you have to install the right MCP server and login before you can use GitHub with Cursor’s chat. With MCP Jetpack, you can ask Cursor to list your GitHub issues, and it will automatically execute the right tool behind the scenes to accomplish your task. For services that require authentication, you will be asked to login the first time you interact with the service. However, it all happens within the Cursor chat so you never have to context switch and fiddle with Cursor’s settings.<p>Tool Limits - Cursor warns you if you have more than 50 MCP tools installed as it says having more will degrade performance. However, just installing the GitHub MCP server itself adds 74 MCP tools. With MCP Jetpack, you get access to GitHub, Atlassian and 15 other services with just two tools: “FindTool” and “ExecTool”.<p>Here are the 17 services we support today: GitHub, Atlassian, Canva, Linear, Notion, Intercom, Monday.com, Neon, PayPal, Hugging Face, Sentry, Square, Webflow, Wix, Cloudflare Docs, Cloudflare AI Gateway, Cloudflare Workers Bindings.<p>We’ll continue to add more services as companies launch remote MCP servers. If yours isn’t listed and you’d like it to be added, please email us at team@mcpjetpack.com.<p>MCP Jetpack is in alpha so please let us know if you run into any problems or have any feedback - thanks!

[API/SDK] Show HN: An API for human-powered browser tasks At APM Help, we have a large team that performs repetitive, browser-based tasks. Years ago, to manage this work securely and get a clear audit trail, we built an internal platform we call &quot;Hub.&quot; It&#x27;s essentially a locked-down environment where our team works that records their sessions, tracks every interaction, and prevents data from being copied or shared. It&#x27;s been our internal source of truth for years.<p>More recently, like many companies, we&#x27;ve been building more automation. And like everyone else, we&#x27;ve seen our automations fail on edge cases—a weirdly formatted invoice our parser can&#x27;t read, a website layout change that breaks a scraper, etc. Our team would have to manually step in to fix these.<p>We realized other developers must have this exact same problem, but without a 250-person team on standby. So we connected our old, battle-tested Hub to a new, modern front door: a Human-in-the-Loop (HITL) API. We&#x27;re calling it browser-work.com.<p>The idea is simple: when you hit a task that needs a human, you can send it to our team through the API.<p>Here&#x27;s how it works:<p><pre><code> - You POST a request to our endpoint. The payload contains the context for the task (like a URL) and a set of instructions for the human on what to do. - The task appears in the Hub, where one of our trained operators can claim it. - They perform the task exactly as instructed, all within the secure Hub environment. - When they&#x27;re done, we send a webhook to your system. The return payload includes the task&#x27;s output, any notes left by the human, and a detailed log of their actions (e.g., DOM elements they interacted with). </code></pre> For example, if your automation for paying a utility bill fails, you can pass the task to us. A person will log in, navigate the portal, make the payment, and return a confirmation number.<p>The product is live and we&#x27;re looking for people with interesting use cases.<p>I&#x27;m Robert, the CIO. If this sounds useful to you, send me a brief email about your use case at robert@apmhelp.com and we can get you started right away.<p>Happy to answer any questions here.

[Other] Show HN: JavaScript-free imageboard in Lua, with CSS based widgets If you&#x27;re a visual learner, one of the biggest instances can be seen at:<p><a href="http:&#x2F;&#x2F;lambdaplusjs35padjaiz4jw2fugdoeutse262phqr72uf634s2wdbqd.onion&#x2F;" rel="nofollow">http:&#x2F;&#x2F;lambdaplusjs35padjaiz4jw2fugdoeutse262phqr72uf634s2wd...</a><p>It&#x27;s usually SWF. Some features include:<p><pre><code> - No javascript of any type is used anywhere. In fact, CSP is used to block execution of any and all javascript, which makes XSS attacks impossible. - High security due to the use of pledge() and unveil() - Highly transparent moderation logs which include a reason for all moderation actions - advanced formatting, including LaTeX support - anyone can create their own board - multiple files per post - inline image expansion - per-board index and recent pages - paginated catalog pages and threads - extensive moderation tools</code></pre>

[Other] Show HN: MCP server for up-to-date Zig standard library documentation Hey HN! I made this because Zig&#x27;s stdlib changes so much and outdated docs are a problem. Server fetches the latest documentation directly from the ziglang.org and makes it available through the MCP, so LLM can query stdlib functions and builtins.<p>Link: <a href="https:&#x2F;&#x2F;github.com&#x2F;zig-wasm&#x2F;zig-mcp">https:&#x2F;&#x2F;github.com&#x2F;zig-wasm&#x2F;zig-mcp</a>