Show HN: Basic Memory – Build a knowledge graph from Claude conversations Basic Memory is an open-source tool that enables Claude to build and navigate a persistent knowledge graph based on your conversations. It solves the problem of lost context in AI interactions by storing knowledge in standard Markdown files on your computer. I built this because I found myself constantly repeating information to LLMs and wanted a system where my knowledge grew naturally through conversations while maintaining complete control over my data. Demo video: https://ift.tt/4FQHyPv Key features: - Continue conversations exactly where you left off without repetition - All knowledge stays in local Markdown files you can edit anytime - Works with Claude Desktop via the Model Context Protocol - Seamless integration with Obsidian for visualization and editing - Fully open source (AGPL) The system works by creating structure from simple markdown patterns: - Observations with categories: `- [category] fact #tag` - Relations between documents: `- relation_type [[WikiLink]]` or plain `[[Wikilinks]]` - These patterns emerge naturally during conversations When you chat with Claude, you can simply say "Let's continue our conversation about X" and it will build context from your knowledge base, without needing to upload files every time. GitHub: https://ift.tt/HbPA2Md Docs: https://ift.tt/lFn4v9a Website: https://ift.tt/Zl1Fp86 Requires Claude Desktop or other MCP host and Python 3.12+ I'd love feedback from the HN community, particularly from those interested in knowledge management or AI applications. https://ift.tt/HbPA2Md March 15, 2025 at 11:49PM
Show HN: Psyllium, a Ruby Gem to make Fibers behave more like Threads Hi everyone! I created this small Ruby Gem to add some convenient methods to the Fiber class to make it easier to use in the same way a Thread object can be used. This was born out of my frustration that the current implementation of the Fiber class makes it difficult to retrieve the final value of a block passed to a Fiber, especially when creating a fiber via the `schedule` class method. I appreciate any feedback anyone has. https://ift.tt/Z32hXRE March 15, 2025 at 12:09AM
Show HN: OCR Benchmark Focusing on Automation OCR/Document extraction field has seen lot of action recently with releases like Mixtral OCR, Andrew Ng's agentic document processing etc. Also there are several benchmarks for OCR, however all testing for something slightly different which make good comparison of models very hard. To give an example, some models like mixtral-ocr only try to convert a document to markdown format. You have to use another LLM on top of it to get the final result. Some VLM’s directly give structured information like key fields from documents like invoices, but you have to either add business rules on top of it or use some LLM as a judge kind of system to get sense of which output needs to be manually reviewed or can be taken as correct output. No benchmark attempts to measure the actual rate of automation you can achieve. We have tried to solve this problem with a benchmark that is only applicable for documents/usecases where you are looking for automation and its trying to measure that end to end automation level of different models or systems. We have collected a dataset that represents documents like invoices etc which are applicable in processes where automation is needed vs are more copilot in nature where you would need to chat with document. Also have annotated these documents and published the dataset and repo so it can be extended. Here is writeup: https://ift.tt/a6ejtim Dataset: https://ift.tt/crNBC5o Github: https://ift.tt/rRJVzgs Looking for suggestions on how this benchmark can be improved further. https://ift.tt/a6ejtim March 13, 2025 at 02:19AM
Show HN: Pi Labs – AI scoring and optimization tools for software engineers Hey HN, after years building some of the core AI and NLU systems in Google Search, we decided to leave and build outside. Our goal was to put the advanced ML and DS techniques we’ve been using in the hands of all software engineers, so that everyone can build AI and Search apps at the same level of performance and sophistication as the big labs. This was a hard technical challenge but we were very inspired by the MVC architecture for Web development. The intuition there was that when a data model changes, its view would get auto-updated. We built a similar architecture for AI. On one side is a scoring system, which encapsulates in a set of metrics what’s good about the AI application. On the other side is a set of optimizers that “compile” against this scorer - prompt optimization, data filtering, synthetic data generation, supervised learning, RL, etc. The scoring system can be calibrated using developer, user or rater feedback, and once it’s updated, all the optimizers get recompiled against it. The result is a setup that makes it easy to incrementally improve the quality of your AI in a tight feedback loop: You update your scorers, they auto-update your optimizers, your app gets better, you see that improvement in interpretable scores, and then you repeat, progressing from simpler to more advanced optimizers and from off-the-shelf to calibrated scorers. We would love your feedback on this approach. https://build.withpi.ai has a set of playgrounds to help you quickly build a scorer and multiple optimizers. No sign in required. https://code.withpi.ai has the API reference and Notebook links. Finally, we have a Loom demo [1]. More technical details Scorers: Our scoring system has three key differences from the common LLM-as-a-judge pattern. First, rather than a single label or metric from an LLM judge, our scoring system is represented as a tunable tree of metrics, with 20+ dimensions which get combined into a final (non-linear) weighted score. The tree structure makes scores easily interpretable (just look at the breakdown by dimension), extensible (just add/remove a dimension), and adjustable (just re-tune the weights). Training the scoring system with labeled/preference data adjusts the weights. You can automate this process with user feedback signals, resulting in a tight feedback loop. Second, our scoring system handles natural language dimensions (great for free-form, qualitative questions requiring NLU) alongside quantitative dimensions (like computations over dates or doc length, which can be provided in Python) in the same tree. When calibrating with your labeled or preference data, the scorer learns how to balance these. Third, for natural language scoring, we use specialized smaller encoder models rather than autoregressive models. Encoders are a natural fit for scoring as they are faster and cheaper to run, easier to fine-tune, and more suitable architecturally (bi-directional attention with regression or classification head) than similar sized decoder models. For example, we can score 20+ dimensions in sub-100ms, making it possible to use scoring everywhere from evaluation to agent orchestration to reward modeling. Optimizers: We took the most salient ML techniques and reformulated them as optimizers against our scoring system e.g. for DSPy, the scoring system acts as its validator. For GRPO, the scoring system acts as its reward model. We’re keen to hear the community’s feedback on which techniques to add next. Overall stack: Playgrounds next.js and Vercel. AI: Runpod and GCP for training GPUs, TRL for training algos, ModernBert & Llama as base models. GCP and Azure for 4o and Anthropic calls. We’d love your feedback and perspectives: Our team will be around to answer questions and discuss. If there’s a lot of interest, happy to host a live session! - Achint, co-founder of Pi Labs [1] https://ift.tt/Dgok0t4 https://ift.tt/oKT5wYL March 14, 2025 at 07:07PM
Show HN: Bubbles, a vanilla JavaScript web game Hey everybody, you might remember my older game, Lander! It made a big splash on Hacker News about 2 years ago. I'm still enjoying writing games with no dependencies. I've been working on Bubbles for about 6 months and would love to see your scores. If you like it, you can build your own levels with my builder tool: https://ift.tt/jTkDaIU and share the levels here or via Github. https://ift.tt/gNfZPm7 March 13, 2025 at 11:18PM
Show HN: Time Portal – Get dropped into history, guess where you landed Hi HN! I love imagining the past, so I made Time Portal, a game where you are dropped into a historical event and see AI video footage from that moment. You have to guess where you are in time and on the map. It’s like GeoGuessr (and heavily inspired by it!) but for historical events. The videos are all created with AI. It’s a pipeline of Flux (images), Kling (video), and mmaudio (audio). The videos aren’t always historically accurate to the last detail. They might incorporate elements of folklore or have details from popular beliefs about the way things looked rather than the latest academic research on how they looked. I’m thinking a lot about how to make the game more interactive. One thing that makes Geoguessr so fun for me is that you can move infinitely and always find more details to help you pinpoint the location. I want Time Portal to have a similar quality. I have a few ideas to try soon that will hopefully make the game more interactive and infinite. https://ift.tt/l3YatFV March 13, 2025 at 01:53AM
Show HN: Translate Japanese Manga and Korean Manhwa with Chrome Extension If you are a manga or manhwa lover, you must understand the feeling of waiting for your favourite series being translated into English or sometimes your native language. Now, you can translate them in real-time with Fakey Chrome extension! https://ift.tt/GFYNKzH March 9, 2025 at 07:55PM
Show HN: We built a Plug-in Home Battery for the 99.7% of us without Powerwalls Hi HN! I’m Cole Ashman, founder of Pila Energy. I’ve spent my career working on home energy systems—first as an engineer on Tesla’s Powerwall, where I focused on the Backup Gateway, Solar Inverter, and metering systems. More recently, I led Product at SPAN, where we built the Smart Electrical Panel and integrated with most major home solar, EV, and battery systems. Pila ( https://pila.energy/ ) is a home battery that plugs into a standard wall outlet, provides smart backup power, energy shifting, and grid services. It’s more than a power bank—it’s a distributed energy system that can scale across multiple rooms, entire buildings, and work together in real time as a coordinated system. We built Pila to be local first with an open API to allow developers to build use cases on top of our hardware (Home Assistant, etc). Big batteries like Tesla Powerwall and Enphase are great if you own a home and can afford a $10K+ electrical project, but they require permanent installation, electricians, and panel upgrades—which makes them inaccessible for renters, apartments, and cost-conscious homeowners. Over 50% of the cost of installing a Powerwall isn’t even the battery itself—it’s soft costs: labor, permitting, etc. We wanted to create an entry point for more people to access energy security at home. How does it work? Plug Pila into any 120V wall outlet, and power passes through to connected devices and appliances. The inverter, LFP battery, BMS, grid disconnection, controller, and wireless connectivity are all built in. (details at https://ift.tt/FxDUNRq ) When an outage happens, the onboard inverter detects the power loss within 20ms and automatically disconnects from the grid (islanding). Whether you’re home or away, backup kicks in instantly. A built-in cellular radio ensures you get a notification even if your home WiFi is out. Pila is 1.6kWh. That will backup a standard fridge for over a day. One key challenge we faced with a distributed architecture was coordination between batteries, for things like solar-following and managing real-time draw from your utility connection. Unlike large garage systems, where you can run a wired CAN bus, our batteries are spread across the home. We’re solving this with a sub-GHz wireless mesh network—self-healing, coordinator-less, and designed to make setup and expansion as simple as plugging in another unit. Long-term, we’d love to open up this protocol to provide a more reliable communication layer for energy products in noisy built environments—reducing reliance on consumer Wi-Fi. We want to deliver the value you’d expect from a whole-home battery like Powerwall, in a plug-in format. That means going beyond a basic lead acid UPS with real home energy management, useful insights about power use, power larger loads like sump pumps, and even deliver grid services. Most portable batteries are missing the functionality that makes a home battery useful: no bidirectional power, no integration with solar or smart home systems, and no ability to manage home energy dynamically. They tend to be boxy, ruggedized, meant to be moved around, not seamlessly integrated into your living space. On top of that, many use e-mobility battery chemistries, which are great for delivering high power on demand but wear out faster when cycled daily for home energy use. As a renter myself, I started Pila because these awesome energy products aren’t accessible enough. And frankly, generators are loud, expensive, and a pain to deal with. Even many Powerwall owners I’ve talked to say they really care about keeping the fridge, WiFi, and a sump pump running—so why does energy resilience have to be so complicated and expensive? As the grid struggles to keep up with demand, we believe modular, renter-friendly batteries can make home energy resilience more accessible. What's been your experience with home batteries? What recent power outages have you had, and how were you affected? https://pilaenergy.com March 11, 2025 at 09:18PM
Show HN: Hot Design – Like Hot Reload, but a Runtime Visual Designer Hi HN, Nick here, from the open-source Uno Platform team. You are likely familiar with Hot Reload , pioneered by Flutter. We’ve taken that concept further and built Hot Design , let me introduce it to you. Architecturally, Hot Design idea is simple: 1. In your IDE, pause the live, running app at runtime, turning it into a designer. 2. Modify the UI directly on the designer —add elements, adjust layouts, tweak bindings etc. 3. Resume the app without restarting or losing state. We built Hot Design to address the frustration of slow iteration cycles when building and tweaking UI or debugging data bindings in apps targeting multiple platforms. Here’s a detailed explanation and a video of Hot Design in action: https://ift.tt/mNz5Kk1 I can see potential criticism: It will get killed by AI, it’s another abstraction over code, it is .NET etc. Happy to respond to those comments if they come; we put a lot of thought into Hot Design and would love to hear it challenged! Nick https://ift.tt/mNz5Kk1 March 11, 2025 at 07:40AM
Show HN: Chrome Extension for ChatGPT to organize conversations into folders Hi HN, I'm Alex, a full-stack developer from Toronto, Canada. I recently built a Chrome extension that organizes ChatGPT conversations into folders, allowing users to sort and save important information for easy reference. The idea for this extension came from a friend who highlighted the lack of good (and affordable) ChatGPT organizers. Many existing tools were either low-quality or overpriced, so I decided to create one that was both reliable and accessible. I built the extension using plain JavaScript and developed a backend with Express to handle Google authentication. For storage, I used MongoDB, enabling all users with an account to save their folder structures and conversation data. Initially, I planned to charge $5 per month to cover costs since originally this extension was intended as a portfolio project addressing a real-world problem. However, just as I finished the main functionality and was about to implement payments, ChatGPT announced an official feature similar to one my extension was providing. Rather than continue competing in a market with an "official" solution, I decided to stop development. But I didn't want my work to go to waste, so I chose to release it for free, motivated by a desire to share it with the community. I made some changes to eliminate the backend. Now the extension stores all folder structures and content locally in Chrome storage. Luckily, I had some old code to reuse for this. The extension is now live on the Chrome Web Store. This project introduced me to a lot of new challenges with technologies I hadn’t used before, but I’m grateful for the experience and the skills I gained along the way. I hope you find it useful! Links to the extension and its website: https://ift.tt/itznvfe... https://ift.tt/1XDOlYi If you have any questions or suggestions, feel free to reach out in the comments or via email at georgepozdman@gmail.com. https://ift.tt/1XDOlYi March 11, 2025 at 04:41AM
Show HN: SQLite vs. GoatDB: Surprising Benchmark Results for a New Realtime NoDB We introduced GoatDB just three weeks ago and have been blown away by the community’s response. Your feedback and excitement genuinely exceeded our expectations—so first and foremost, thank you from all of us! For anyone just hearing about it: GoatDB is a real-time, version-controlled NoDB for Deno and React that’s edge-native, meaning it requires only minimal backend infrastructure without heavy server components. It’s designed for prototyping, self-hosting, single-tenant apps, and even ultra-light multi-tenant setups if you want to keep your backend minimal. One of the biggest requests we heard was, “Where are the benchmarks?” We’re thrilled to share them now. The numbers tell an interesting story: in some tests, our distributed-commit-graph architecture can be significantly slower than SQLite; in others, it’s surprisingly faster. This is what happens when you put synchronization and collaboration first (instead of disk I/O). But let’s be crystal clear: GoatDB isn’t a drop-in SQLite replacement. It has a fundamentally different architecture designed for real-time distributed scenarios and cryptographic auditing, so it comes with its own unique tradeoffs. Key Takeaways: - Simple reads and incremental queries can be blazingly fast , especially with concurrency and real-time syncing. - Opening large repositories can take longer if everything stays in memory (we’re exploring a zero-copy format to address that). - It’s not just a SQLite wrapper—this is a fundamentally different approach with its own unique tradeoffs. We’ve documented how to run these same benchmarks in our documentation if you’re curious. Once again, thank you so much for the excitement and support. We’re a small team on a mission to reimagine what a lightweight database can do, and your feedback keeps us inspired. We can’t wait to see what you build with GoatDB! Checkout our Github Repo: https://ift.tt/6deXMam https://ift.tt/P9ZzREm March 10, 2025 at 10:34PM
Show HN: I built a Figma plugin for quick data calculations I lead design on a B2B SaaS product. It's quite data-heavy in places. Using placeholder content in data tables, checkout summaries and dashboards is a big no-no for us. It might seem like using random numbers saves time at first, but sooner or later there's documentation to write and plenty of clarifications to be made. It throws off customers during interviews – "hey, that's not really my sales target!". It confuses stakeholders at review time– "what's this data point supposed to be?" I built a Figma calculator plugin for my team so that they spend less time doing mental maths. It calculates sums, differences averages and percentages, and makes it easy to use real-looking data in designs. https://ift.tt/DwWGPlc March 10, 2025 at 07:11PM
Show HN: Evolving Agents Framework Hey HN, I've been working on an open-source framework for creating AI agents that evolve, communicate, and collaborate to solve complex tasks. The Evolving Agents Framework allows agents to: Reuse, evolve, or create new agents dynamically based on semantic similarity Communicate and delegate tasks to other specialized agents Continuously improve by learning from past executions Define workflows in YAML, making it easy to orchestrate agent interactions Search for relevant tools and agents using OpenAI embeddings Support multiple AI frameworks (BeeAI, etc.) Current Status & Roadmap This is still a draft and a proof of concept (POC). Right now, I’m focused on validating it in real-world scenarios to refine and improve it. Next week, I'm adding a new feature to make it useful for distributed multi-agent systems. This will allow agents to work across different environments, improving scalability and coordination. Why? Most agent-based AI frameworks today require manual orchestration. This project takes a different approach by allowing agents to decide and adapt based on the task at hand. Instead of always creating new agents, it determines if existing ones can be reused or evolved. Example Use Case: Let’s say you need an invoice analysis agent. Instead of manually configuring one, our framework: Checks if a similar agent exists (e.g., a document analyzer) Decides whether to reuse, evolve, or create a new agent Runs the best agent and returns the extracted information Here's a simple example in Python: import asyncio from evolving_agents.smart_library.smart_library import SmartLibrary from evolving_agents.core.llm_service import LLMService from evolving_agents.core.system_agent import SystemAgent async def main(): library = SmartLibrary("agent_library.json") llm = LLMService(provider="openai", model="gpt-4o") system = SystemAgent(library, llm) result = await system.decide_and_act( request="I need an agent that can analyze invoices and extract the total amount", domain="document_processing", record_type="AGENT" ) print(f"Decision: {result['action']}") # 'reuse', 'evolve', or 'create' print(f"Agent: {result['record']['name']}") if __name__ == "__main__": asyncio.run(main()) Next Steps Validating in real-world use cases and improving agent evolution strategies Adding distributed multi-agent support for better scalability Full integration with BeeAI Agent Communication Protocol (ACP) Better visualization tools for debugging Would love feedback from the HN community! What features would you like to see? Repo: https://ift.tt/iYA13Oq https://ift.tt/iYA13Oq March 9, 2025 at 10:21PM
Show HN: Simple Certificate Decoder Tool Sometimes I need to quickly check certificates, especially key details like SANs, expiration dates, issuer info, etc. I know there are dozens (if not hundreds) of certificate decoders out there already, but I built my own—mostly for fun, but also because I prefer tools that are clean, simple, and straightforward to use. Would appreciate your feedback! https://ift.tt/YherBq8 March 8, 2025 at 11:09PM
Show HN: Story Jam, a music composition tool for Storytellers https://ift.tt/nVi9PZj Hello! My name is Cortland Mahoney. I'm a music researcher, software engineer, and producer. I made Story Jam. This doc is intended to inform you of not just the product, but the centuries of work that have led up to its implementation. Are you tired of the barriers in traditional music composition? Story Jam is here to break them down. Designed for anyone with creative ideas — from poets to film directors — our tool offers a new way to create and edit chord progressions, powered by cutting-edge music theory. *Who Story Jam is for: Storytellers* Story Jam makes music composition accessible and meaningful to anybody, with or without musical training. It is designed for those who crave musical control but struggle with traditional composition methods. This includes film directors, slam poets, and self-taught musicians. Story Jam is not music production software. Do not expect fancy sounds or synthesizers. It's purely a composition tool, designed to spark your creative process. *Try it out!* The demo is free on the homepage, no login required! This is an MVP, so it has an "introductory" feature set. Feature requests welcome; help me build the product you want. *The chord progression suggestion logic* This service is built on a novel new music theory I have developed called Monic Theory. Monic Theory is a rigorous proof for music. Not "Western music": music. Monic Theory describes the tonal space of any conventional music on earth (except noise music. For that just use `Math.random()`). It describes the static and transient function of chords, instantaneously and differentially over time. This model enables empirical measurement of chords and the relationship between chords. (hint: It is nothing you have seen in Xenharmonic Alliance. This is a new approach I have been developing over the past 10 years.) Therefore, Monic Theory enables us to describe (or "predict" if you will) a chord progression to invoke a certain feeling. *Music Composition* Three people who helped set up the environment for Monic Theory are composers Paul Hindemith and Harry Partch , and music theorist Heinrich Schenker. These folks independently contributed new ideas to music composition and analysis. All of these people lived without access to rapid computation. This is critical for the Partch case, as he computed many tables of frequencies by hand to support his compositional technique. Partch recognized the human-math-music relation in "Genesis of a Music." He includes in this text some samples of his hand-computed tables of frequency values of overtones and (importantly) undertones which support the basis is technique. Partch's techniques were so far-fetched that he had to construct new instruments to perform his scores. Similarly, I had to build a digital synthesizer to render the output of Monic Theory. (See: https://ift.tt/rhp023Y ). *About me* I was a working composer and violinist from 2007 until 2017, and I have been a software engineer for the past 7.5 years. I was a volunteer organizer for Livecode.NYC, an NYC livecode community; and am the volunteer creator of Data Dancers, Atlanta's livecode community. I am passionate about algorithmic art and have provided about a dozen workshops over four years on the topic. https://ift.tt/UN518ig thank you for reading. May the flow of Spices be with you :) naltroc March 5, 2025 at 11:16PM
Show HN: Microvoxel FPS game test in browser Microvoxel streaming FPS game test with half a kilometer scanned play area made of 1.5cm voxels. In-browser with no signup - just click and you're in. You can shoot holes in anything, dig up new items out of the ground, and generate 3D objects from images you drag into the 3D view from your desktop or other browser windows. All changes are persistent, and you can see/shoot/build with other users in there at the same time, or build/destroy things that have been have made in the past. https://ift.tt/8oEZVei March 5, 2025 at 10:37PM
Show HN: Time travel debugging AI for more reliable vibe coding Hi HN, I'm the CEO at https://replay.io . We've been building a time travel debugger for web apps for several years now (previous HN post: https://ift.tt/Lx3Ef7Y ) and are combining our tech with AI to automate the debugging process. AIs are really good at writing code but really bad at debugging -- it's amazing to use Claude to prompt an app into existence, and pretty frustrating when that app doesn't work right and Claude is all thumbs fixing the problem. The basic reason for this is a lack of context. People can use devtools to understand what's going on in the app, but AIs struggle here. With a recording of the app its behavior becomes a giant database for querying using RAG. We've been giving Claude tools to explore and understand what happens in a Replay recording, from basic stuff like seeing console messages to more advanced analysis of React, control dependencies, and dataflow. For now this is behind a chat API ( https://ift.tt/dBkL4An ). We recently launched Nut ( https://nut.new ) as an open source project which uses this tech for building apps through prompting (vibe coding), similar to e.g. https://bolt.new and https://v0.dev . We want Nut to fix bugs effectively (cracking nuts, so to speak) and are working to make it a reliable tool for building complete production grade apps. It's been pretty neat to see Nut fixing bugs that totally stump the AI otherwise. Each of the problems below has a short video but you can also load the associated project and try it yourself. - Exception thrown from a catch block unmounts the entire app: https://ift.tt/Qogy4Y6 - A settings button doesn't work because its modal component isn't always created: https://ift.tt/4QxFImC - An icon is really tiny due to sizing constraints imposed by other elements: https://ift.tt/hTuJWOP - Loading doesn't finish due to a problem initializing responsive UI state: https://ift.tt/21aQ7L3 - Infinite rendering loop caused by a missing useCallback: https://ift.tt/ARpODNJ Nut is completely free. You get some free uses or can add an API key, and we're also offering unlimited free access for folks who can give us feedback we'll use to improve Nut. Email me at hi@replay.io if you're interested. For now Nut is best suited for building frontends but we'll be rolling out more full stack features in the next few weeks. I'd love to know what you think! https://nut.new March 5, 2025 at 12:23AM
Show HN: Vidformer – Drop-In Acceleration for Cv2 Video Annotation Scripts Hi HN, this is a project I've been working on as part of my PhD. Vidformer is a system that makes video annotation or transformation scripts practically instant. Traditional scripts that render full videos can take minutes—Vidformer speeds this up by optimizing execution and using on-demand rendering, so results appear immediately instead of waiting for entire videos to render. It works as a drop-in replacement for OpenCV's cv2, meaning most scripts can adopt it by simply changing "import cv2" to "import vidformer.cv2 as cv2"—no need to rewrite code or sacrifice flexibility. Vidformer is written in Rust and uses FFmpeg libraries for low-level video access. Under the hood, Vidformer runs code with symbolic references to frames and tracks frame modifications to build a declarative representation of the task. Then, when rendering, it can transparently distribute the workload across many cores and efficiently use additional memory for caching frames. Further, it can expose a Video on Demand endpoint and only render segments once requested; this lets playback begin instantly. Repo: https://ift.tt/fd5BiHI The "Open in Colab" notebook is a great place to start. Would love to hear feedback! https://ift.tt/fd5BiHI March 4, 2025 at 11:05PM
Show HN: Firebender, a simple coding agent for Android Engineers Hey HN, I made a simple coding agent plugin in Android Studio called Firebender. Here’s an unedited 5-minute video where it writes tests for an Android app and iterates against the Gradle task output on its own ( https://ift.tt/5sg46l0 ). You can use the plugin for free, no sign up needed, on the jetbrains marketplace. The agent can edit multiple files, run gradle tasks like tests, and use the output to improve its changes. At the end, it reports a git diff of all changes that can be accepted or rejected. Under the hood, the agent relies on Claude 3.7 sonnet and a fast code apply model to speed up edits. We built tools to give deeper access throughout the IDE like IntelliJ’s graph representation of kotlin/java code, “everywhere search” for classes, and have more integrations planned. The goal is for the agent to have access to all the IDE goodies that we engineers take for granted, to improve the agent's responses and ability to gather correct context. In order to improve the agent, there are internal evals like “tasks” and simulate the IDE which serves as a gym for the agent. This is heavily inspired by SWE-bench. Whenever tools, prompts, subagents, or models are changed, this gym helps find regressions quickly. Building the UI was surprisingly hard. I had the great pleasure of becoming proficient in Java Swing (released in ‘96 by Netscape) to get this done right. Things like markdown streaming, or streaming git diffs are prone to layout flickering where Swing tries to recalculate where elements should go. We had to write our own markdown parsing and rendering engine that repaints Swing components only when changed portions of the markdown nodes. The UI tends to focus on simplifying reviewing AI changes, something I have a feeling we’ll be doing much more in the coming years. If you’re an Android engineer, please let me know if you run into any bugs or want anything improved in the plugin! https://ift.tt/5sg46l0 March 3, 2025 at 11:18PM
Show HN: Mmar – open-source, zero-dependancy, cross-platform HTTP tunneling Hey HN! For the past couple of months, I've been working on and off on a cool project I'm excited to share. mmar (pronounced "ma-mar") is an open-source, zero dependency, cross platform and self-hostable HTTP tunnel built in Go. It allows you to easily expose your localhost to the world on a public URL. You can easily create an HTTP tunnel right away for free on a randomly generated subdomain on "*.mmar.dev" if you don't feel like self-hosting. This isn't something new, in fact there's quite a few of alternative HTTP tunneling tools out there. mmar is my attempt to optimize for a super easy developer experience and simplified implementation. None the less, I had a blast building it and I think developers could find it pretty useful. Additionally, I documented the whole process of building mmar through devlogs. You can read about the thought process and implementation details here ( https://ift.tt/n5yRSiO ). If I would suggest one devlog to read, I highly recommend devlog 5 ( https://ift.tt/SI9rROK ). I describe how I built a (very) basic DNS server just to run simulation tests for mmar (a bit of an overkill, but a fantastic learning experience). I dive deep into the DNS protocol and explain why I needed to implement it. Finally, I would love to hear your thoughts and feedback. If you try mmar out, let me know! https://ift.tt/zIw09E4 March 3, 2025 at 01:28AM
Show HN: Crop images into square, circle, heart, oval for free I developed an Image Cropper, which supports cropping images into square, circle, heart, and oval shapes. It also supports customizing the width and height for arbitrary cropping, which is very simple. https://cropimage.co March 2, 2025 at 10:10PM
Show HN: I built an app to convert ChatGPT Deep Research to PDFs with footnotes Whilst ChatGPT Deep Research is very useful for generating in-depth reports, it's time consuming to copy, reformat the text (thousands of words) and clean referenced hyperlinks for use in a professional context. Out of frustration, I built deep research docs to help save time by automating the reformatting, cleaning links, footnote references, and conversion to shareable PDF format. Hopefully this helps you save time to focus on meaningful work. Let me know your feedback. https://ift.tt/JIyCK0Y March 1, 2025 at 06:22PM
Show HN: Built a "Story Chatbot Arena" to Crowdsource AI Story Preferences Hi HN, I’m a university student interested in Gen AI, and over the holidays, I built a project in public: Who Rates the Rater? – a crowdsourced dataset for benchmarking AI-generated storytelling. The Idea Inspired by Chatbot Arena, this lets users compare AI-generated stories and provide feedback. The goal is to refine LLMs for creative writing using real human preferences. How It Works - Live Demo: https://ift.tt/2zFvPnL - Tech Stack: Built with Streamlit + Supabase - Open Source: https://ift.tt/AYeOC1c Get Involved - Try it & star the repo if you find it interesting - Bug reports & feature requests welcome on Twitter - Follow me for future AI & storytelling projects Would love to hear your thoughts! https://ift.tt/AYeOC1c March 1, 2025 at 11:19PM
Show HN: Betting game puzzle (Hamming neighbor sum in linear time) In Spain, there's a betting game called La Quiniela: https://ift.tt/FTyvAZs Players predict the outcome of 14 football matches (home win, draw, away win). You win money if you get at least 10 correct, and the prize amount depends on the number of winners. Since all bets are public, the number of winners and the corresponding payouts can be estimated for each of the 3^14 possible outcomes. We can also estimate their probabilities using bookmaker odds, allowing us to compute the expected value for each prediction. As a side project, I wanted to analyze this, but ran into a computational bottleneck: to evaluate a prediction, I had to sum the values of all its Hamming neighbors up to distance 4. That’s nearly 20,000 neighbors per prediction (1+28+364+2912+16016=19321): S_naive = sum from k=0 to r of [(d! / ((d-k)! * k!)) * (q-1)^k] (d=14, q=3, r=4) This took days to run in my first implementation. Optimizing and doing it with matrices brought it down to 20 minutes—still too slow (im running it in GAS with 6 minutes limit). For a while, I used a heuristic: start from a random prediction, check its 28 nearest neighbors, move to the highest-value one, and repeat until no improvement is possible within distance 3. It worked surprisingly well. But I kept thinking about how to solve the problem properly. Eventually, I realized that partial sums could be accumulated efficiently by exploiting overlaps: if two predictions A and B share neighbors, their shared neighbors can be computed once and reused. This is achieved through a basic transformation that I implemented using reshape, roll, and flatten (it is probably not the most efficient implementation but it is the clearest), which realigns the matrix by applying an offset in dimension i. This transformation has two key properties that enable reducing the number of summations from 19,321 to just 101: - T(T(space, d1), d2) = T(T(space, d2), d1) - T(space1, d) + T(space2, d) = T(space1+space2, d) Number of sums would be the result of this expression: S_PSA = 1 + (d - (r-1)/2) * r * (q-1) I've generalized the algorithm for any number of dimensions, elements per dimension, and summation radius. The implementation is in pure NumPy. I have uploaded the code to colab, github and an explanation in my blog. Apparently, this falls under Hamming neighbor summation, but I haven't found similar approaches elsewhere (maybe I'm searching poorly). If you know or you've worked on something similar, I'd love to hear your thoughts! colab: https://ift.tt/MoVN6Af... github: https://ift.tt/4KqaUNE blog: https://ift.tt/V7PAsRM... March 1, 2025 at 02:03AM
Show HN: News-briefing-generator – Local LLM-powered news digest Hey HN, I created a tool to generate personalized news briefings from RSS/Atom feeds using local LLMs through Ollama. It currently works in two modes: fully automatic or with an interactive review where you can select which "main topics of the day" to include in your briefing. The result is a HTML document with summaries for each topic. https://ift.tt/mjiPnNb February 28, 2025 at 10:45PM
Show HN: Superglue – open source API connector that writes its own code Hi HN, we’re Stefan and Adina, and we’re building superglue ( https://superglue.cloud ). superglue allows you to connect to any API/data source and get the data you want in the format you need. It’s an open-source proxy server which sits between you and your target APIs. Thus, you can easily deploy it into your own infra. If you’re spending a lot of time writing code connecting to weird APIs, fumbling with custom fields in foreign language ERPs, mapping JSONs, extracting data from compressed CSVs sitting on FTP servers, and making sure your integrations don’t break when something unexpected comes through, superglue might be for you. Here's how it works: You define your desired data schema and provide basic instructions about an API endpoint (like "get all issues from Jira"). superglue then does the following: - Automatically generates the API configuration by analyzing API docs. - Handles pagination, authentication, and error retries. - Transforms response data into the exact schema you want using JSONata expressions. - Validates that all data coming through follows that schema, and fixes transformations when they break. We built this after noticing how much of our team's time was spent building and maintaining data integration code. Our approach is a bit different to other solutions out there because we (1) use LLMs to generate mapping code, so you can basically build your own universal API with the exact fields that you need, and (2) validate that what you get is what you’re supposed to get, with the ability to “self-heal” if anything goes wrong. You can run superglue yourself ( https://ift.tt/16T9mSV - license is GPL), or you can use our hosted version ( https://ift.tt/M2CEuy3 ) and our TS SDK (npm i @superglue/client). Here’s a quick demo: https://www.youtube.com/watch?v=A1gv6P-fas4 You can also try out Jira and Shopify demos on our website ( https://superglue.cloud ) Excited to share superglue with everyone here—it's early so you'll probably find bugs, but we'd love to get your thoughts and see if others find this approach useful! https://ift.tt/16T9mSV February 27, 2025 at 10:50PM
Show HN: Instantly Translate Manga – TranslateManga Since I was young, I've loved anime, and over the years, manga has brought me joy, given me courage, and sparked excitement in my heart. However, as I read more, I realized that many of these manga weren't translated at all. I also came across some AI-based translation tools, but the results often fell short. So, I decided to create a tool that allows manga fans to read and enjoy their favorite manga, no matter the language or whether a translation team is available. This product has just been launched, and there are certainly areas that can be improved. However, with time, I'm confident it will only get better. You're welcome to try it out and share your valuable feedback! https://ift.tt/FHqJcIA February 24, 2025 at 08:09PM
Show HN: Yaak – An Open Source, Git-Friendly Desktop API Client Hi HN! I'm Greg and I've been working on Yaak for just over a year. I usually describe it as a Postman alternative since most developers are familiar with that, but it's targeted more toward users that just want an API client (no mocking, testing, spec design, etc). Having also created Insomnia for a similar purpose, I never thought I'd build another API client again. But, after selling Insomnia in 2019 and watching it expand into the broader feature set of Postman, I was left wanting a simpler tool again. Yaak was my answer to that. It's hard to describe how it's better than the 100 other API clients, since its main benefit is design, but there are a few stand-out features: - Optionally sync data to a local directory as plain-text, for use with Git/etc - Build and install plugins for authentication, template functions, or generic actions - Support for REST, GraphQL, WebSocket, server sent events, and gRPC Yaak requires a license for commercial use but this only applies when using the prebuilt binaries. If building from source, a license is not required. It's available for Mac, Windows, and Linux and the source is at https://ift.tt/e1hGn2V I'd love to hear you think! https://yaak.app February 26, 2025 at 10:54PM
Show HN: I built a PR listener and ruleset to detect malicious code in CI/CD I built a GitHub app that detects it in pull requests, notifies or blocks them. Alongside it, I published a Semgrep ruleset for any stage of the CI/CD. I started this after getting frustrated by all the FUD around malicious code - lots of noise, little effort to solve it. Having said that, it's still a major attack vector - a stored RCE, with the codebase itself as the sink. Feedback is appreciated. The app, PRevent - https://ift.tt/e5itWw8 The ruleset: https://ift.tt/VBouAmY The research: https://ift.tt/I7XvkKB... https://ift.tt/e5itWw8 February 26, 2025 at 12:52AM
Show HN: Txtl – Fast static website of text utilities I often need simple text tools (e.g. diffing, epoch time conversion, number base conversion) and was frustrated that I couldn't find a simple option that was as fast as I wanted. Many (SEO-optimized) websites exist, but they unnecessarily send your input to their servers. I used tools like DevToys but their diff has been broken on Mac for a while [1] and it requires installing an app, and CyberChef is good but requires too many interactions (yes, I'm picky). So, I decided to build my own. txtl tries to autodetect what mode you want based on the input, and in general is fast and gets out of your way as much as possible. For example, if you paste in a number then it will show base conversions, but if it looks like an epoch timestamp then it will display the corresponding date. And of course it's open source at https://ift.tt/XqQBC3Z This was partly also a way for me to try out LLMs on a frontend/new project. This could probably be a whole blog post, but I see why some people can say that LLMs can 10x their productivity (I certainly didn't find this true in my day job) - I was really impressed with what o1 can do, it probably got me 70% of the way there in one shot. And as someone not familiar with frontend development, iterating was way faster. But, it still requires plenty of feedback and requires handholding - I had to completely rewrite the diff rendering logic, despite what I tried I couldn't get an LLM to get it right. Would love feedback on the autodetection and overall usability! [1] https://ift.tt/GrMJIYf https://ift.tt/5zwdgBA February 25, 2025 at 11:30PM
Show HN: Bracket City – A daily, exploded (?) crossword puzzle Hi hn - I co-own a diner where I co-host a puzzle night that is kind of like a diner-themed escape room. At the last one, I made a puzzle that was crossword-like clues nested in brackets. People at the diner seemed to like it, so I resolved to make it a real game and Bracket City was born: https://bracket.city . I love crosswords, so it's been fun to write crossword-like clues: [it contains MSG] as well as clues that would not make it into a crossword: [___ <=== you ===> hard place] I write all the puzzles and post a new one at midnight ET every day of the week. Still working on a lot of features/fixes. I'm aware that scoring based on keystrokes is pretty unfair, especially given not-ideal custom keyboard on mobile! Still thinking through the best solution there. Also fun fact: if you sign up for the email list, you get a special "Word of the Day" email written by James Somers (of https://jsomers.net ). The only way to sign up for the email list is to finish a puzzle! ** (answer key: NYC, ROCK) https://bracket.city February 24, 2025 at 08:50PM
Show HN: I built an app to stop me doomscrolling by touching grass i wanted to change the habit of reaching for my phone in the morning and doomscrolling away an hour so i built an app to help me. now i have to literally touch grass before accessing my most distracting apps the app is built in swiftui, uses the screen time apis provided by apple and google vision to recognise grass or not i'd love to get your thoughts on the concept. https://touchgrass.now/ February 24, 2025 at 05:45PM
Show HN: Cardog – AI interface for vehicle ownership Hi HN! I wanted to share something I've been working on after an interesting pivot. Last year I built a vehicle search tool that ran into legal issues with major listing sites. That experience led me to tackle a different problem - making the entire car ownership experience more accessible and data-driven. Ended up building an AI interface that helps research any vehicle, access documentation, and manage ownership - think having a car expert, market analyst, and personal assistant rolled into one. Core features: - Natural language interface to research any vehicle, parse manuals, and search relevant web/YouTube content (think perplexity for cars) - Monitor market values and listings across North America - Track maintenance, service records, registration dates for your garage - Store ownership documents, recall info, service bulletins Live demo: https://cardog.ai | Example: Ask about reliability ratings for the latest SUVs or "What should I look for when buying a used Model 3?" Would love to hear what aspects of car ownership you find most frustrating. https://cardog.ai February 24, 2025 at 01:55AM
Show HN: OmniTools–A Self-Hosted Suite of Open Source Tools for Everyday Tasks It’s 2025, and somehow, there’s still no good self-hosted alternative to sites like ILoveIMG.com or OnlineTools.com... until now. OmniTools is here to fill that gap! It’s a free web-based, open-source self-hosted platform that brings together all your favorite online tools in one place—fully self-hosted and ad-free. Project: https://ift.tt/h3aUVMd Why Omni Tools? Completely FREE & Open-Source – No hidden fees, ever. Self-Hosted – Keep control of your data, no tracking, no nonsense. All Your Favorite Tools in One Place – Image, coding, file utilities and more! Beta Version – Just launched, and I need your feedback to make it even better! https://ift.tt/h3aUVMd February 23, 2025 at 08:58PM
Show HN: Mapping historical markers around the world I saw the 'map of torii' post yesterday and thought y'all might like to see the small app I've been working on that uses HMDB.org data to map historical markers around the world. HMDB has been aggregating markers for over 15 years and back when I was living out of my van and traveling full-time I wanted to get notified whenever I passed one, so I built a mobile app around that. I think historical markers are underrated - as a physical marker they make history tangible. Rather than reading about history from a classroom, you get the opportunity to see and engage with it at the source. If you're already nearby, they are often worth the stop to learn more. Since releasing the iOS app a few years ago, I've been able to enhance the markers with summaries (which makes reading the content a lot more palatable), and converting them to audio, so you can listen to markers when you're driving. Yesterday I officially released the android app, with the same features as the iOS app. https://ift.tt/UFIRed9 February 23, 2025 at 10:28PM
Show HN: Course on Building Full-Stack Chrome Extensions with React and Node.js I've been working in the extension space on a variety of products for a number of years now and decided to put together a course on how everything I wish I knew when I first started out. It goes through building an entire "product", meaning UI, API, and extension, all communicating with each other. It covers a lot of topics I get asked about often as well such as extension-level authentication, injecting React apps into web pages via content scripts, and a bunch more. https://ift.tt/L8UsSxk February 22, 2025 at 11:28PM
Show HN: Willpayforthis.com – Ideas people will pay for Ah, there's a dumb easy hack to figure out what ideas people will pay for. Search "I'd pay for" on Twitter and you'll find hundreds of posts from people talking about pain points and products they'd pay for to solve them. Do this enough and you realize you have to filter through a lot of slop. slop. slop. I created willpayforthis.com to accumulate high signal, high quality posts and save you some time. I love thoughts from the community on how I can make it better, save you time, and help you work on the best ideas. https://ift.tt/EQbMJ4h February 23, 2025 at 12:21AM
Show HN: Rhiza – easily create shortcuts and add entries to PATH Rhiza is a Windows-only tool that makes any app easily launchable from both the command line and the Windows Start Menu. It works by creating shortcuts and adding entries to the PATH. Key Features: * Crawl ~ common directories to detect apps and games automatically * Add ~ any app by searching for it across the entire file system * Path ~ search for an executable and add its directory into PATH Rhiza simplifies app launching / calling tools by finding and managing them for you. https://ift.tt/PUgS4sJ February 22, 2025 at 01:03AM
Show HN: Tradofire, a fun way to learn crypto trading risk-free Hey HN, I have been building this for the last couple years and probably spent way too much time. First I wanted to make an automated trading strategy based on crypto coins breaking support and resistance lines but after writing a whole system and backtesting infra I realized it doesnt work :( So here it is now as an app where you can learn to trade crypto based on these signals. Some key features are 1. Real-Time Market Analysis: See when coins break support/resistance levels or suddenly spike. 2. Paper Trading with Leverage: Test your trading strategies without risking real money. 3. Performance and Leaderboards: See your paper trading performance and compare with others. I dont know honestly if this will ever make any money but just sharing and hoping some folks like it. If you like it please tell me what else I can add. Cheers Sumeru PS: The app is iOS only for now (Android can come soon if there is demand) https://ift.tt/xqmeDTf February 22, 2025 at 01:23AM
Show HN: DSBG – A Static Site Generator That Fast-Tracks Your Digital Presence The ethos behind it is to automate your digital presence as much as possible, while retaining control over the created content. To that end, the following features are available: Easy installation; Support for Markdown & HTML source files; Automatic tag generation from paths ; built-in tag filtering; Client-side fuzzy search over all content; Automatic RSS feed generation; Watch mode with automatic rebuild for continuous feedback; 3 different themes, with the ability to add your own via custom CSS; Automatic share buttons for major social networks; Easy to extend with analytics, comments, and more. https://ift.tt/e1SloND February 22, 2025 at 12:09AM
Show HN: Benchmarking VLMs vs. Traditional OCR Vision models have been gaining popularity as a replacement for traditional OCR. Especially with Gemini 2.0 becoming cost competitive with the cloud platforms. We've been continuously evaluating different models since we released the Zerox package last year ( https://ift.tt/U2dYJ3N ). And we wanted to put some numbers behind it. So we’re open sourcing our internal OCR benchmark + evaluation datasets. Full writeup + data explorer here: https://ift.tt/SI2oPGA Github: https://ift.tt/qNlWjn8 Huggingface: https://ift.tt/mUTXGyZ Couple notes on the methodology: 1. We are using JSON accuracy as our primary metric. The end goal is to evaluate how well each OCR provider can prepare the data for LLM ingestion. 2. This methodology differs from a lot of OCR benchmarks, because it doesn't rely on text similarity. We believe text similarity measurements are heavily biased towards the exact layout of the ground truth text, and penalize correct OCR that has slight layout differences. 3. Every document goes Image => OCR => Predicted JSON. And we compare the predicted JSON against the annotated ground truth JSON. The VLMs are capable of Image => JSON directly, we are primarily trying to measure OCR accuracy here. Planning to release a separate report on direct JSON accuracy next week. This is a continuous work in progress! There are at least 10 additional providers we plan to add to the list. The next big roadmap items are: - Comparing OCR vs. direct extraction. Early results here show a slight accuracy improvement, but it’s highly variable on page length. - A multilingual comparison. Right now the evaluation data is english only. - A breakdown of the data by type (best model for handwriting, tables, charts, photos, etc.) https://ift.tt/SI2oPGA February 21, 2025 at 12:19AM
Show HN: WinCse – Integrating AWS S3 with Windows Explorer WinCse is an application that integrates AWS S3 buckets with Windows Explorer. Utilizing WinFsp and the AWS SDK, WinCse allows you to treat S3 buckets as part of your local file system, making file management simpler. The application is currently in development, with plans for additional features and improvements. https://ift.tt/IHLVNrk February 20, 2025 at 11:23PM
Show HN: Make your logo liquid metal (open source) Good morning!! We thought the Apple liquid metal invite was so cool. How fun would it be if everyone could see their logo in liquid? So we made an app to let you make your logo in liquid. Just drag in your logo and see. To play with your logo: https://ift.tt/loYtDha Repo: https://ift.tt/y2Q3CN0 (We think you're gonna love it!) https://ift.tt/loYtDha February 20, 2025 at 01:41AM
Show HN: Subtrace – Wireshark for Docker Containers Hey HN, we built Subtrace ( https://subtrace.dev ) to let you see all incoming and outgoing requests in your backend server—like Wireshark, but for Docker containers. It comes with a Chrome DevTools-like interface. Check out this video: https://www.youtube.com/watch?v=OsGa6ZwVxdA , and see our docs for examples: https://ift.tt/L9lfBTF . Subtrace lets you see every request with full payload, headers, status code, and latency details. Tools like Sentry and OpenTelemetry often leave out these crucial details, making prod debugging slow and annoying. Most of the time, all I want to see are the headers and JSON payload of real backend requests, but it's impossible to do that in today's tools without excessive logging, which just makes everything slower and more annoying. Subtrace shows you every backend request flowing through your system. You can use simple filters to search for the requests you care about and inspect their details. Internally, Subtrace intercepts all network-related Linux syscalls using Seccomp BPF so that it can act as a proxy for all incoming and outgoing TCP connections. It then parses HTTP requests out of the proxied TCP stream and sends them to the browser over WebSocket. The Chrome DevTools Network tab is already ubiquitous for viewing HTTP requests in the frontend, so we repurposed it to work in the browser like any other app (we were surprised that it's just a bunch of TypeScript). Setup is just one command for any Linux program written in any language. You can use Subtrace by adding a `subtrace run` prefix to your backend server startup command. No signup required. Try for yourself: https://ift.tt/L9lfBTF https://ift.tt/GhYXgRx February 19, 2025 at 04:59AM
Show HN: A GPU-accelerated binary vector index This is a vector index I built that supports insertion and k-nearest neighbors (k-NN) querying, optimized for GPUs. It operates entirely in CUDA and can process queries on half a billion vectors in under 200 milliseconds. The codebase is structured as a standalone library with an HTTP API for remote access. It’s intended for high-performance search tasks—think similarity search, AI model retrieval, or reinforcement learning replay buffers. The codebase is located at https://ift.tt/cyK9eUS . https://ift.tt/Skwipyg February 17, 2025 at 06:15AM
Show HN: Mind Map Wizard – Generate Infinite Mind Maps with AI, for Free (FOSS) Hi everyone, I want to share my newest project, Mind Map Wizard - a Text to Mind Map AI that transforms your text input into a detailed mind map in seconds. Just type in your topic and go. No account required! I want to get a solid overview of any possible topic in seconds. You don't need an account and there's no paid option. Features: - Simplicity: Only the essentials. Editing, sharing, downloading and creation history. All mind maps are stored locally in your browser. - Focus: Without any ads and an intuitive design, you can fully concentrate on your topics without distractions. -Local first: Data is stored locally, ensuring fast navigation and enhanced privacy. Check out this mind map I generated about Switzerland: https://ift.tt/1Z0C2UT I’d love to hear your thoughts on how to improve Mind Map Wizard. Thanks for your support! GitHub link: https://ift.tt/oG7RE3N https://ift.tt/WSFDATk February 17, 2025 at 10:53PM
Show HN: Hackyournews.com v2 A year and a half after I published https://ift.tt/7xfBiL3 , I've rewritten it to be neater and added support for more news sources. HackYourNews.com v1 had a great response on HN [1] and consistently sees ~2k weekly unique visitors. There were many long-standing requests that I wanted to fulfill (thanks for your patience!): a proper dark mode, correct rendering on mobile devices, and more cogent summaries. This rewrite is the result. gpt-4o-mini reduces the cost of summarization to an absurd degree, so it's now sustainable to keep this free service going! Someday, I hope to use the Batch API [2] to drive down costs even further. Enjoy. [1] https://ift.tt/tDfVlMa [2] https://ift.tt/x7mLb4V February 16, 2025 at 06:16AM
Show HN: The news in the last 30, 14, 7, 3, or 1 days I made this for when I come back from vacation and want to catch up on news. It's a bit of a simplistic LLM transformation on headlines and URLs that I store from RSS feeds. So it bugs out sometimes. But I think it might be useful to me. You can check out some of the prompts in the "debug" links. What do you think? https://ift.tt/VUu37X8 February 16, 2025 at 11:24AM
Show HN: Tech Brief – AI enhanced news reading I built this because I wanted it, and I now use it every day. It's a simple news site that gathers and summarises tech content and discussions, across multiple sources, providing tight, easily digestable summaries along with some simple tooling to support reading workflows. 1) Hourly updated homepage with the latest tech news across the web. 2) A simple < 3 min "News of the Hour", every hour, audio clip. 3) Summaries of HackerNews and Product Hunt, incl. comments and sentiment (more to come). 3) GitHub login with AI summaries of any releases made to your starred repos. 4) Read/Unread article status. 5) Simple swipe interface and keyboard support. 6) Simple Bookmark/Readling List, and Favourite tags (logged in) No Tracking. Fast. Mobile Friendly. Easy sharing. https://tech.brief.page/ February 16, 2025 at 05:58AM
Show HN: I got frustrated with CRMs, so I'm building my own for startups After trying a bunch of CRMs for my startup, I kept running into the same issues—overpriced plans that scale aggressively or bloated features that just slow me down. I wanted something simple, affordable, and actually built for startups, so I decided to build it myself: Leadchee.com. Fixed pricing, no nonsense. Curious—how do you all handle CRMs? Do you stick with the big players, go for niche tools, or build your own? Would love to hear your thoughts! https://leadchee.com February 14, 2025 at 09:25PM
Show HN: Open-Sourcing My LLM Drag and Drop Website Builder Hey HN - OP here. I wrote some about this project in the following link, and there's a video demo as well: https://ift.tt/Zv1PARS... This has been one of my favorite things I've ever worked on - the way the LLM collaborates with the user to accelerate tedious and hard work, the way you can directly edit the code instead of dealing with a panel of visual editing toggles - I think it has a lot of potential but I don't have time to pursue it anymore so open-sourcing it. The idea for this came out of conversations with a few people who were struggling with frontend development. For technical people, strictly using an LLM to write code can be tedious. To combat this, LLM usage is limited to getting started quickly, improving design, and wiring up frontend state. On the other hand, writing frontend code feels less efficient than just moving things around on a screen. Hence the drag and drop interface that makes it fast to build. Finally, I despise the visual editing toggles on Figma / Bubble / Squarespace / etc. The amount of hunt and peck to simply adjust a font a bit and change some colors or add a shadow is a huge time suck. So I built a way to directly edit the underlying React code when styling - just add or remove tailwind classes. IMO the craziest thing is that all of the code is just stored on the frontend in a config language of sorts. It is inflated at runtime and can be updated without any hot reload. There is no "underlying React code" for the app you're building here - in order to edit the code, I convert the config into React code, then convert back to a config, which triggers updates in the dom. Anyways, I think there's a lot of clever stuff in here, but then again I wrote it. Happy to answer any questions and hope this is interesting/helpful to someone else out there. https://ift.tt/tOx7hLo February 14, 2025 at 11:38PM
Show HN: Dockershrink – AI Assistant to reduce the size of Docker images For the past few months, I've been hacking around a project I call Dockershrink. It automates a simple task: Take a Dockerfile and optimize its code with the goal of reducing the size of the final Docker image. People don't realize that we can apply some very basic techniques to reduce, for eg, a 2GB image down to just ~100MB: - Multistage builds with light-weight base image for final stage - Remove unused dependencies - Optimizations specific to the tech stack And I feel like I've already done this optimization for my personal projects and backend apps at my job(s) a couple of times. The project currently uses GPT-4o (open source so you can run it locally) and only works for Nodejs projects. There are a couple of reasons why I think dockershrink can be better than using just Vanilla LLM or Github Copilot/Cursor: - Image optimization can benefit from a lot of custom prompting, especially when you have insights about specific tech stacks. Describing techniques deeply in the prompt gave better results than simply asking the LLM to "optimize code for bloat reduction". - A RAG approach will be truly beneficial. I plan on giving dockershrink access to up-to-date documentations of Docker, Bash and all programming languages out there. Additionally, it can be given a few suitable chunks of code to enhance the context. - Analysing custom base images: most orgs have their customized base images. Adding context about these can further help Dockershrink make better decisions. Try it out - "brew install dockershrink" Happy to hear your thoughts! https://ift.tt/fZ1o8ME February 14, 2025 at 04:15AM
Show HN: SQL Noir – Learn SQL by solving crimes I built SQL Noir, an interactive detective game that challenges you to solve mysteries using real SQL queries. It’s fully open source, designed to give you a practical and immersive way to learn SQL while engaging with a narrative-driven mystery. https://www.sqlnoir.com February 14, 2025 at 03:19AM
Show HN: Dev SSO IdP, a mock single sign-on provider as a development aide Hi HN! This project grew out of my want for the development of my web UIs to not get hung up on integration with OpenID Connect single sign-on. SSO was only available in our stage and prod environments. Getting this integration laid down and tested fast, without having to jury-rig something in stage, would've been huge. And so I decided to build a solution myself. Hence, Dev SSO IdP. The vision for it is to mock all the features of an OIDC SSO server that would be pertinent to the development of web apps. To try it out: 1) Create a file called `.production.env` and paste in it the following 2 lines to start with: DEVSSOIDP_PERCENT_ENCODED_REDIRECT_URIS=http%3A%2F%2Flocalhost%3A5173 DEVSSOIDP_CLIENT_IDS=my_cool_app 2a) (with Node) Clone the repo with `git clone https://ift.tt/Dxi10Qr `, then overwrite the project's `.production.env` with yours, then in the project's folder run `npm install`, then `npm start` 2b) (with Docker) Run `docker run -p 3000:3000 --rm --env-file .production.env bmcase/devssoidp:1.0.0` You can then see it at (and have your app redirect to) http://localhost:3000/authorize?response_type=code&client_id... You can add or change environment variables in `.production.env` in the likely case that its defaults don't apply to you. The GitHub readme goes into more detail on all of this. They say "be flexible in what you accept and strict in what you output". But Dev SSO IdP is intentionally strict in what it accepts so that I could catch issues faster. It raises an alarm in dev so you don't later get one in prod. This version I am comfortable designating v1.0.0. It has all the features needed for the OIDC code flow. I'd appreciate any advice, and in particular am interested in: * Would this actually be useful in your projects? Is there anything else it would need? * Do you use the OIDC implicit flow? I've never had reason to, and I understand it's regarded as a bad practice. But I worry I may be in a bubble and so I want to know if there's in fact a lot of folks out there who use the implicit flow. Aside, I'm open to work, and would be interested in bringing my full stack skills to your team (or the team of someone you want to do a favor for), in the Austin TX area or remotely. I'm happy to hear from you by email (ben@benswords.com) or LinkedIn ( https://ift.tt/CFtjpOg ). https://ift.tt/yR1tNjk February 14, 2025 at 01:53AM
Show HN: Sort lines semantically using llm-sort This is a small plugin I made for Simon Willison's llm utility. You can do things like: cat names.txt | llm sort -q "Which one of these names is best for a pet seagull?" cat books.txt | llm sort -q "Which book is more related to basic vs. advanced CS topics?" I see a lot of potential marrying LLMs with classic UNIX interfaces. https://ift.tt/qBKiYfD February 11, 2025 at 08:55AM
Show HN: Community Detection on Bluesky We ran the Leuven community detection algorithm on popular users on Bluesky (where the graph has edges determined by Jaccard similarity of a users' followers). We identified 118 communities and based on the names and descriptions of the top 10-20 users had LLMs generate title and descriptions for them. There are communities like "Feline enthusiasts", "Web Professionals", a bunch of NSFW ones and quite many communities are many different flavors of progressive/liberal activists. https://ift.tt/qThzLNX February 12, 2025 at 02:09AM
Show HN: Colada for Claude I actually enjoy Claude.ai's interface and artifacts implementation. I didn't want to lift and shift over to another LLM tool. So, to get past the daily limits, I decided to build a simple Chrome extension which continues conversations using your own Anthropic API key (or using our managed key). In short: - Get past Claude.ai conversation limits - Bring your own Anthropic API key - Optionally use our managed unlimited API key - Preserve conversation context - Bring project knowledge context into your extended convos - Reference web search results anytime by asking "Search the web" Let me know what you think. Open to any and all feedback. https://usecolada.com February 11, 2025 at 10:38PM
Show HN: WhisperCat – An Audio Recorder and Transcription Tool Hi HN, I wanted to share my first open-source project with you all: WhisperCat . WhisperCat is a small desktop application for recording audio and transcribing it using OpenAI's Whisper API. I built this because I needed something simple and reliable for my own transcription workflows, and now I'm hoping it might be useful to others as well. It's still pretty early stage, but it works well for basic audio recording and transcription tasks. What It Does: Lets you record audio with your preferred microphone. Transcribes audio files automatically via Whisper (OpenAI's transcription API). Supports global hotkeys for recording (e.g., CTRL + R or a custom sequence like triple ALT). Runs in the background (system tray) when minimized. Has a basic microphone testing feature to help you pick the right device. Shows desktop notifications for events (e.g., when recording starts or errors happen). Platforms: WhisperCat is available for Windows and Linux , and there’s also an experimental macOS build you can try if you’re feeling adventurous: Experimental macOS Build You can download the latest release here: https://ift.tt/3c1ka2E Feedback is welcome! https://ift.tt/RCcf6pr February 11, 2025 at 01:11AM
Show HN: Arelo – A simple, flexible file watcher for auto-restarting commands arelo is a lightweight, language-agnostic file watcher that automatically runs a command when files change. It requires no configuration files; everything is controlled via simple command-line options. Easy to use: arelo -p '**/*.go' -- go run . Flexible file watching: Supports fsnotify (real-time), polling (for environments like WSL2), and fine-grained control with extended globbing (** and {js,ts,json}). Cross-platform and lightweight: Runs on Windows, macOS, and Linux without extra dependencies. Installation: - go install github.com/makiuchi-d/arelo@latest - Or download a prebuilt binary from GitHub Releases: https://ift.tt/1asD0UM https://ift.tt/awq5iZ3 February 10, 2025 at 09:49PM
Show HN: Curatrs – Scheduled Programming for Podcasts Like many of us, I got tired of scrolling endlessly through podcast apps trying to find the right show for my commute. So I built Curatrs (curatrs.com) - it brings radio-style scheduled programming to podcast discovery. Instead of endless scrolling, you get podcasts programmed for specific times and durations. Currently in early MVP, built with Vite/Supabase, focused on making discovery more intentional and time-based. Would appreciate any feedback, especially from regular podcast listeners https://ift.tt/Ppm0DVL February 10, 2025 at 01:23AM
Show HN: Neovim Plugin for iOS and macOS Development Over two years ago, I began exploring whether it was possible to shift my iOS development to Neovim. It took me over six months to resolve all issues and figure out how to connect everything, creating an environment with all the features required for development. After that, I decided to develop my own plugin to enable others to do the same. Since then, I have been developing apps for iOS and macOS using Neovim, for both my professional work and personal projects with no issues. This change has significantly boosted my productivity, and I no longer have to deal with Xcode's flaws. I can accomplish 95% of my work without needing to open Xcode. Before that, it seemed impossible to develop for Apple platforms outside of Xcode. I'm proud that I was stubborn enough to make it work after all those failures along the way. Neovim is an amazing project <3. https://github.com/wojciech-kulik/xcodebuild.nvim February 9, 2025 at 06:16PM
Show HN: Hyloblog – minimal, Git-based SSG for writing (not theming) Hi HN, We're working on Xr0 [0] and have been building a static-site generator that meets our tastes and needs. The basic emphasis is on simplicity and content rather than customisability and feature-richness. We are imitating Jekyll and Hugo (and other SSGs) in their basic generation paradigm and LaTeX in its separating form from content, but attempting to combine these into a unified, minimal philosophy where you can open a repo and start writing without needing a CLI tool to generate your folder for you. The project is broken into two applications: an SSG you can run locally (or in a GH Action etc.) and a platform for easy hosting that bundles in some basic audience interaction features. Both are available on GitHub: [1] and [2]. We've been working on this somewhat sporadically for the past couple of months, and it is very much a WIP, particularly in the themes it offers, but we're keen to hear thoughts on this. [0]: https://xr0.dev [1]: https://ift.tt/618wZAl [2]: https://ift.tt/FNxkqhI https://hyloblog.com/ February 9, 2025 at 02:53AM
Show HN: Chez Scheme txtar port from Go This is a loose port of https://ift.tt/aQuUEe0 , which concatenates files together and allows for a top-level comment. The txtar format is specifically designed to be easy for humans to read and write by hand. It's perfect for test data. The library itself tries to follow scheme standards by depending on only SRFIs. I hope it's a helpful showcase of a scheme library! I'm also new to writing scheme, and if any experienced scheme devs are out there with feedback, that would be much appreciated. https://ift.tt/yDNlHTB February 8, 2025 at 11:54PM
Show HN: Open-source Hacker News apps We at Emerge Tools (YC W21) recently released open source Hacker News apps for both iOS and Android. The apps use the latest SwiftUI and Compose frameworks and are entirely native. Our goal is to help dogfood our products, but more importantly our team just enjoys reading HN every day and wanted an app to hack on and call our own. :) We are still missing some features but should otherwise be pretty solid. And open to any contributions. iOS: https://ift.tt/TmC1QvY... Android: https://ift.tt/Xt34dDG... https://ift.tt/q8WeDMa February 8, 2025 at 02:11AM
Show HN: A configuration management system for minimal *Nix environments I built this thing to scratch my itch for a simpler alternative to Ansible and similar for basic jobs like setting up a dev env, building containers, and provisioning small fleets of servers. Grateful for feedback on the approach! A couple of friends and I have found it useful for day-to-day automation tasks and I'm wondering if there's enough utility in its ability to target minimal environments like Alpine containers or IoT devices to warrant fleshing out a proper 1.0 release. https://ift.tt/CI6uEtQ February 7, 2025 at 11:16PM
Show HN: Heap Explorer I wrote a little LD_PRELOAD library that makes it easy to inspect and interact with a running program's glibc heap. It's fun to pause processes, free a bunch of their allocations, then resume them. Most of the time, the processes continue as though nothing happened, but sometimes they do interesting things :) https://ift.tt/06CVZmT February 6, 2025 at 10:24AM
Show HN: Kindly RSS, a self-hostable RSS app designed for e-ink devices In the last few weeks I've been working on a RSS application designed to be used in e-ink devices such as Kindle, through the device's web browser. It's a self-hostable app optimized for running on low-end hardware (such as Raspberry Pi, I actually run it on a 3b model). The project is in its early stages of development. It is usable, but you may (and probably will :P) encounter bugs from time to time. I did it for myself (I like to read at night before going to sleep but I don't like to use my phone at that time). I thought people could find it useful so I worked on it a little bit more to publish it. At the moment it can only be run by downloading and compiling the source code or using the docker image (in the repo and the landing page there is a curl that executes the script to run the container, manual instructions can be found in the repo's README). Repo: https://ift.tt/DLun96a Dockerhub: https://ift.tt/RrZkF3N Thank you for reading! I'd love to hear your thoughts and suggestions. https://kindlyrss.app/ February 6, 2025 at 02:16AM
Show HN: CoPlay – Enabling In-Room Xbox Gaming for Children's Hospitals Hey everybody, My name is Brady. I'm the creator of CoPlay. Think of it as an MDM/fleet management for xbox accounts and devices. Pediatric hospitals want to allow their patients to play and connect with friends, family and other patients but they have lacked the tools to facilitate and manage this in the past. A friend and I found this problem while volunteering at our local children's hospital. We are now in 6 pediatric hospitals across the US. Yup, it's a niche. But it's a cool niche. I'm sharing this to get feedback, answer questions, contribute to this amazing community and most importantly, HOPEFULLY FIND SOMEONE THAT CAN GET US CONNECTED TO SOME OF THE HIGHER UPS AT MICROSOFT/XBOX. If you have any connections at all please reach out. We believe there is an opportunity for a beautiful partnership there. Sorry for shouting ;) https://ift.tt/2HGADes February 5, 2025 at 02:08AM
Show HN: Haystack Code Reviewer – Perform code reviews on a canvas Hi HN! We’re building Haystack Code Reviewer, a tool that lays out code diffs for a GitHub pull request on an interactive canvas. Instead of scrolling through diffs line-by-line, you can view all changes in a more connected, visual format – similar to viewing a call graph. We hope this will make it easier and less cognitively taxing to understand how different changes across files work together. For a quick overview, check out our short demo video: https://www.youtube.com/watch?v=QeOz70x0WPE . If you would like to give it a spin, head over to https://ift.tt/RTGSDjI , click the “Review pull request button” in the top toolbar, and load any pull request via URL or pick a pull request from a dropdown. We built Haystack Code Reviewer because we found pull requests difficult to review in a pure textual format — especially when hopping between multiple files or trying to break down complex changes. Oftentimes, pull request authors would have to specifically structure their commits so that code reviews would be easier to tackle, which is a time-consuming and error-prone process. Our goal is to make any pull request easy to understand at a glance, and reduce the effort needed from both reviewers and authors to craft a good code review. Haystack Code Reviewer works on private repositories! We have authentication to ensure that someone cannot open the server for your pull request without having access to that pull request on GitHub. For additional security, we plan to build self-hosting soon. Please contact us if you’re interested in this. Alternatively, a completely local option would be to download desktop Haystack and then navigate to your pull request from there. This is great for trying out the feature without exposing any data on the cloud! In the near future, we plan to: 1. Introduce step-by-step navigation to guide reviewers through each part of the changeset 2. Allow for self-hosting We’d love to hear your thoughts, suggestions, and any feedback on our approach or potential features! https://ift.tt/tSGg4f7 February 4, 2025 at 10:32PM
Show HN: CLI tool that makes your app self-hostable in one command I've been exploring the self-hosting world for some time now, and I wanted to simplify things with one command and avoid having to jump through hoops to make my application self-hostable. I thought many of these could be automated, so I'm automating everything related to self-hosting, one after the other. Please check the roadmap for what I have in mind. I'm open to anyone who wants to contribute and give feedback. https://kagehq.com/ February 4, 2025 at 11:34PM
Show HN: Mandarin Word Segmenter with Translation I've built mandoBot, a web app that segments and translates Mandarin Chinese text. This is a Django API (using Django-Ninja and PostgreSQL) and a NextJS front-end (with Typescript and Chakra). For a sample of what this app does, head to https://ift.tt/wdZ0Ex6 . This is my presentation of the first chapter of a classic story from the Republican era of Chinese fiction, Diary of a Madman by Lu Xun. Other chapters are located in the "Reading Room" section of the app. This app exists because reading Mandarin is very hard for learners (like me), since Mandarin text does not separate words using spaces in the same way Western languages do. But extensive reading is the most effective way to learn vocabulary and grammar. Thus, learning Mandarin by reading requires first memorizing hundreds or thousands of words, before you can even know where one word ends and the next word begins. I'm solving this problem by allowing users to input Mandarin text, which is then computationally segmented and machine translated by my server, which also adds dictionary definitions for each word and character. The hard part is the segmentation: it turns out that "Chinese Word Segmentation"[0] is the central problem in Chinese Natural Language Processing; no current solutions reach 100% accuracy, whether they're from Stanford[1], Academia Sinica[2], or Tsing Hua University[3]. This includes every LLM currently available. I could talk about this for hours, but the bottom line is that this app is a way to develop my full-stack skills; the backend should be fast, accurate, secure, well-tested, and well-documented, and the front-end should be pretty, secure, well-tested, responsive, and accessible. I am the sole developer, and I'm open to any comments and suggestions: roberto.loja+hn@gmail.com Thanks HN! [0] https://ift.tt/60OTK8Q [1] https://ift.tt/WJSpGEL [2] https://ift.tt/cjzCVm4 [3] https://ift.tt/2crU8ge https://ift.tt/4CXU1kP February 4, 2025 at 11:26PM
Show HN: I Built a Platform to Buy and Sell GitHub Repositories Hey HN, I built a platform that allows developers to buy and sell GitHub repositories using private forking. The idea is to help indie developers, open-source maintainers, and teams monetize their work while ensuring buyers get fully functional projects with minimal hassle. Many developers create great projects but lack the time or resources to maintain them. Instead of letting them fade away, why not sell them to someone who wants to continue the work? Here is how it works: - Sellers list theis GitHub repos in the platform - Buyers purchase repos - Buyers automatically added as collaborators and can fork the repo Check it out here: https://gittrader.com https://ift.tt/yEQcLnz February 3, 2025 at 06:07AM
Show HN: Chrome extension to run DeepSeek, LLMs and Whisper locally in browser Hi all, I open sourced my toy project that runs Generative AI models LOCALLY in the side panel of a Chrome extension. The Chrome extension uses Transformers.js to run models in browser under the hood. I've integrated and tested these models so far. \1. LLM: Llama 3, Phi 3.5, Qwen 2.5, SmolLM2 \2. Reasoning: DeepSeek R1 \3. Multimodal LLM: Janus \4. Speech-to-Text: Whisper On an M1 MacBook, DeepSeek R1 1.5B runs at ~30 tokens/sec If you're interested in, you can download the extension from chrome web store or clone my github repository. \1. chrome web store: https://ift.tt/TYnW9vw... \2. github: https://ift.tt/rCbGxUg \3. demo: https://www.youtube.com/watch?v=MSCDdFG5Lls It's my first react project. Feedback is always welcome! https://ift.tt/rCbGxUg February 2, 2025 at 11:19PM
Show HN: Groundhog AI Spring API For anyone building weather-related AI apps, I am releasing an exciting iteration on last year’s model. My Groundhog API is trained on 130 years of data and makes use of 82 separate data sources. Similar to DeepSeek, it is completely open source and free to use. The primary use case is to make inferences about whether spring will come early or not, using a Mixture of Exports (MoE) approach, but surely others can be found if you are creative. Other use cases: - All predicting groundhogs - Where they all live - Whether they are “real” groundhogs or imposters Excited to see what people do with it! https://ift.tt/xtD0YE9 February 2, 2025 at 10:59PM
Show HN: ESP32 RC Cars This is a projected I started that blends both the fun of playing a split screen multiplayer driving game and controlling real rc cars. The cars can also be controlled via bluetooth gamepads and is meant to be easily hackable. https://ift.tt/cfNoauW February 2, 2025 at 12:21AM
Show HN: I hacked LLMs to work like scikit-learn Working with LLMs in existing pipelines can often be bloated, complex, and slow. That's why I created FlashLearn , a streamlined library that mirrors the user experience of scikit-learn. It follows a pipeline-like structure allowing you to "fit" (learn) skills from sample data or instructions, and "predict" (apply) these skills to new data, returning structured results. High-Level Concept Flow: Your Data --> Load Skill / Learn Skill --> Create Tasks --> Run Tasks --> Structured Results --> Downstream Steps Installation: pip install flashlearn Learning a New "Skill" from Sample Data Just like a fit/predict pattern in scikit-learn, you can quickly "learn" a custom skill from minimal (or no!) data. Here's an example where we create a skill to evaluate the likelihood of purchasing a product based on user comments: from flashlearn.skills.learn_skill import LearnSkill from flashlearn.client import OpenAI # Instantiate your pipeline "estimator" or "transformer", similar to a scikit-learn model learner = LearnSkill(model_name="gpt-4o-mini", client=OpenAI()) data = [ {"comment_text": "I love this product, it's everything I wanted!"}, {"comment_text": "Not impressed... wouldn't consider buying this."}, # ... ] # Provide instructions and sample data for the new skill skill = learner.learn_skill( data, task=( "Evaluate how likely the user is to buy my product based on the sentiment in their comment, " "return an integer 1-100 on key 'likely_to_buy', " "and a short explanation on key 'reason'." ), ) # Save skill to use in pipelines skill.save("evaluate_buy_comments_skill.json") Input Is a List of Dictionaries Simply wrap each record into a dictionary, much like feature dictionaries in typical ML workflows: user_inputs = [ {"comment_text": "I love this product, it's everything I wanted!"}, {"comment_text": "Not impressed... wouldn't consider buying this."}, # ... ] Run in 3 Lines of Code - Concurrency Built-in up to 1000 calls/min # Suppose we previously saved a learned skill to "evaluate_buy_comments_skill.json". skill = GeneralSkill.load_skill("evaluate_buy_comments_skill.json") tasks = skill.create_tasks(user_inputs) results = skill.run_tasks_in_parallel(tasks) print(results) Get Structured Results Here's an example of structured outputs mapped to indexes of your original list: { "0": { "likely_to_buy": 90, "reason": "Comment shows strong enthusiasm and positive sentiment." }, "1": { "likely_to_buy": 25, "reason": "Expressed disappointment and reluctance to purchase." } } Pass on to the Next Steps You can use each record’s output for downstream tasks such as storing results in a database or filtering high-likelihood leads: # Suppose 'flash_results' is the dictionary with structured LLM outputs for idx, result in flash_results.items(): desired_score = result["likely_to_buy"] reason_text = result["reason"] # Now do something with the score and reason, e.g., store in DB or pass to next step print(f"Comment #{idx} => Score: {desired_score}, Reason: {reason_text}") https://ift.tt/KnsYruq February 1, 2025 at 10:09PM
Show HN: I re-designed interface for HN Any suggestions are appreciated, stack utilizes Gluestack UI ,Expo, React Native, and Cloudflare Pages. There is a known bug via touch scroll ability on Android, external keyboard's spacebar or mouse works correctly though, currently. If you know about a solution, let me know. Please note, this is just a prototype, it still has a lot of features to be included. I'd like to learn more about how people use HN and how Hacked could help, where other HN clients failed. https://ift.tt/c6RvFrG February 1, 2025 at 10:41PM
Show HN: Simple to build MCP servers that easily connect with custom LLM calls Hi! After learning about MCP, I'm really excited about the future of provider-agnostic, re-usable tooling. Unfortunately I've found that while it's easy to implement an MCP server for use with tools that support it (such as Claude Desktop), it's not as easy to implement your own support (such as integrating an MCP server into your own LLM application). We implemented a thin MCP wrapper that easily integrates with Mirascope calls so that you can hook up an MCP server and client super easily to any supported LLM provider. Excited to see what people build with this! https://ift.tt/c5qC27Y February 1, 2025 at 06:20AM
Show HN: Ros2_utils_tool, a powerful GUI toolset for ROS2-based utilities Hi Hackernews, over the past few weeks, I've been tirelessly working on a GUI toolset for all sorts of ROS2-based utilites to simplify my tasks with ROS at work. Now I want to present to you the ros2_utils_tool. This tool can do many ROS2-based utilites, for example editing a ROS bag file to remove, rename or crop topics, extracting a video or image sequence out of a ROS bag, creating dummy bag files or just publishing a video as a ROS topic. While being developed to be as simple and lightweight as possible, the toolset supports many advanced options, for example different video and image formats, custom fps values, switching colorspaces and more. I've also heavily optimized the tool to support multithreading or in some cases even hardware-acceleration to run as fast as possible. The tool offers full graphical user interface support for all features, while I've also added additional command line interface support for most of them. As of now, the ros2_utils_tool supports ROS2 humble and jazzy. The application is still in an alpha phase, which means I want to add many more features in the future, for example GUI-based ROS bag merging or republishing of topics under different names, or some more advanced options such as selecting messages for video or image generation. The ros2_utils_tool requires an installed ROS2 distribution, as well as Qt6 or Qt5 for the user interface, the cv_bridge for transforming images to ROS and vice versa, and finally catch2_ROS for unit testing. You can install all dependencies (except for the ROS2 distribution itself) with the following command: sudo apt install libopencv-dev ros-humble-cv-bridge qt6-base-dev ros-humble-catch-ros2 For ROS2 Jazzy: sudo apt install libopencv-dev ros-jazzy-cv-bridge qt6-base-dev ros-jazzy-catch- Install the UI with the following steps: cd path/to/your/workspace/src git clone https://ift.tt/yq65k7R cd path/to/your/workspace/ colcon build Then run it with the following commands: source install/setup.bash ros2 run ros2_utils_tool tool_ui The ros2_utils_tool uses the EUPLv1.2 as license. More information, for example regarding the command line interface tools are shown under [0]. [0] https://ift.tt/NhnglxR https://ift.tt/NhnglxR January 31, 2025 at 09:13PM
संदेश
