Show HN: Helicone (YC W23) – OSS LLM Observability and Development Platform https://ift.tt/pcrabLe

Show HN: Helicone (YC W23) – OSS LLM Observability and Development Platform Hey HN, we're Justin and Cole, the founders of Helicone ( https://helicone.ai ). Helicone is an open-source platform that helps teams build better LLM applications through a complete development lifecycle of logging, evaluation, experimentation, and release. You can try our free demo by signing up ( https://ift.tt/fzKQB5e ) or self-deploy with our new fully open-source helm chart ( https://ift.tt/Md8CLIZ ). When we first launched 22 months ago, we focused on providing visibility into LLM applications. With just a single line of code, teams could trace requests and responses, track token usage, and debug production issues. That simple integration has since processed over 2.1B requests and 2.6T tokens, working with teams ranging from startups to Fortune 500 companies. However, as we scaled and our customers matured, it became clear that logging alone wasn’t enough to manage production-grade applications. Teams like Cursor and V0 have shown what peak AI application performance looks like and it's our goal to help teams achieve that quality. From speaking with users, we realized our platform was missing the necessary tools to create an iterative improvement loop - prompt management, evaluations, and experimentation. Helicone V1: Log → Review → Release (Hope it works) From talking with our users, we noticed a pattern: while many successfully launch their MVP quickly, the teams that achieve peak performance take a systematic approach to improvement. They identify inconsistent behaviors through evaluation, experiment methodically with prompts, and measure the impact of each change. This observation shaped our new workflow: Helicone V2: Log → Evaluate → Experiment → Review → Release It begins with comprehensive logging, capturing the entire context of an LLM application. Not just prompts and responses, but variables, chain steps, embeddings, tool calls, and vector DB interactions ( https://ift.tt/PhvzOxi ). Yet even with detailed traces, probabilistic systems are notoriously hard to debug at scale. So, we released evaluators (either via LLM-as-judge or custom Python evaluators leveraging the CodeSandbox SDK - https://ift.tt/cXz0PaZ ). From there, our users were able to more easily monitor performance and investigate what went wrong. Did the embedding search return poor results? Did a tool call fail? Did the prompt mishandle an edge case? But teams would still edit prompts in a playground, run a few test cases, and deploy based on intuition. This lacked the systematic testing we’re used to in traditional software development. That’s why we built experiments (similar to Anthropic's workbench but model-agnostic) ( https://ift.tt/jNXbJm7 ). For instance, when a prompt generates occasional rude support responses, you can test prompt variations against historical conversations. Each variant runs through your production evaluators, measuring real improvement before deployment. Once deployed, the cycle begins again. We recognize that Helicone can’t solve all of the problems you might face when building an LLM application, but we hope that we can help you bring a better product to your customers through our new workflow. If you're curious how our infrastructure handled our growth: Our initial architecture struggled - synchronous log processing overwhelmed our database and query times went from milliseconds to minutes. We've completely rebuilt our infrastructure with two key changes: 1) using Kafka to decouple log ingestion from processing, and 2) splitting storage by access pattern across S3, Kafka, and ClickHouse. This was a long journey but resulted in zero data loss and fast query times even at billions of records. You can read about that here: https://ift.tt/uWbCDdK... We'd love your feedback and questions - join us in this HN thread or on Discord ( https://ift.tt/32GvrRY ). If you're interested in contributing to what we build next, check out our GitHub. https://ift.tt/geAotJ6 January 23, 2025 at 11:28PM

Show HN: Open-source AI video editor https://ift.tt/GZDWbSA

Show HN: Open-source AI video editor Hey HN community! I'm one of the lead devs of this project at fal.ai and we created an open source lightweight video editor powered by the latest media AI models. The main goal was to tackle some challenges when dealing with complex media handling and encoding on the browser. It all started as an internal experiment but as we tackled some of the issues it was clear there could be some value sharing it with the open source community. Some of the key points and tech stack details: - It uses IndexedDb, so all data is local (i.e. no auth, no cloud db) - Multiple AI models for video, image, music and voice-over. APIs are provided by fal.ai - Built with the typical React+Next.js, Shadcn front-end - Used remotion.dev for the realtime video preview (this is such a great project, without it the codebase would be twice as large) - File uploads so you can bring your own media by uploadthing.gg - ffmpeg for encoding the final video and also some ui tricks, like the audio waveform We deployed a version of it and for now it's free to use. We do plan to add some rate limiting and a bring your own API Key next, but it's open source and I'm curious about what the community will build on top of it, or derive from it. Customize your own video app and if you do, please share. If you have any questions, hit me up! https://ift.tt/nAxdipW January 24, 2025 at 12:04AM

Show HN: I Made an Open-Source Laptop from Scratch https://ift.tt/ZvQOMSU

Show HN: I Made an Open-Source Laptop from Scratch Hello! I'm Byran. I spent the past ~6 months engineering a laptop from scratch. It's fully open-source on GH at: https://ift.tt/HA8TQJ7 https://ift.tt/6PxzYlE January 23, 2025 at 02:11AM

Show HN: Responding to SMS Spam with Ollama https://ift.tt/wngWzKk

Show HN: Responding to SMS Spam with Ollama I've been working on a side project to generate responses to spam with various funny LLM personas, such as a millenial gym bro and a 19th century British gentleman. By request, I've made a write-up on my website which has some humorous screenshots and made the code available on Github for others to try out [0]. A brief outline of the system: - Android app listens for incoming SMS events and forwards them over MQTT to a server running Ollama which generates responses - Conversations are whitelisted and manually assigned a persona. The LLM has access to the last N messages of the conversation for additional context. [0]: https://ift.tt/JVdzcAv I'm aware that replying can encourage/allow the sender to send more spam. Hopefully reporting the numbers after the conversation is a reasonable compromise. https://ift.tt/igrbj4M January 23, 2025 at 12:53AM

Show HN: RAG Web UI – Possibly the Most Beginner-Friendly RAG Knowledge Base https://ift.tt/7UxAwnd

Show HN: RAG Web UI – Possibly the Most Beginner-Friendly RAG Knowledge Base RAG Web UI is designed to be the most straightforward way to build your own knowledge-based Q&A system. While other RAG (Retrieval-Augmented Generation) projects might be complex, we focus on making it super easy to understand and use. Why It's The Most Beginner-Friendly: Simple Document Management - Just upload your documents (PDF, DOCX, Markdown, Text) - System handles all the complex processing automatically - No need to worry about document chunking or vectorization - Documents update automatically in the background Easy-to-Use Chat Interface - Ask questions in plain language - Get accurate answers from your documents - See exactly which parts of your documents the answers come from - Natural back-and-forth conversations just like chatting Professional Architecture Made Simple - Clean, modern web interface - Rock-solid backend design - Built for reliability with distributed storage - High-performance search using ChromaDB/Qdrant - Easy to switch databases without touching code Get Started in Minutes: 1. Clone the repo 2. Follow our beginner-friendly setup guide 3. Upload your documents 4. Start chatting with your knowledge base Whether you're building a company knowledge base or a personal Q&A system, you don't need to be a RAG expert - we handle the complexity for you. Looking forward to your feedback on how we can make it even more beginner-friendly! https://ift.tt/yptH5N3 January 22, 2025 at 09:35PM

Show HN: CloudCoil – Production-ready Python client for cloud-native ecosystem https://ift.tt/tWZkQRy

Show HN: CloudCoil – Production-ready Python client for cloud-native ecosystem Show HN: CloudCoil – Production-ready Python client for the cloud-native ecosystem I built CloudCoil ( https://ift.tt/g0pYqlj ) to make cloud-native development in Python feel first-class, starting with a modern async Kubernetes client. Frustrated with existing tools that felt like awkward ports from Go/Java, I focused on creating an API that Python developers would actually enjoy using. Installation is as simple as: uv add cloudcoil[kubernetes] # Using uv (recommended) pip install cloudcoil[kubernetes] # Using pip Key features: - Elegant, truly Pythonic API that follows Python idioms - Async-first with native async/await (but sync works too!) - Full type safety with MyPy + Pydantic - Zero-config pytest fixtures for K8s integration testing Quick taste of the API: # It's this simple to work with resources service = k8s.core.v1.Service.get("kubernetes") # Async iteration feels natural async for pod in await k8s.core.v1.Pod.async_list(): print(f"Found pod: {pod.metadata.name}") # Create resources with pure Python syntax deployment = k8s.apps.v1.Deployment( metadata=dict(name="web"), spec=dict(replicas=3) ).create() The ecosystem is growing! We already have first-class integrations for: - cert-manager (cloudcoil.models.cert_manager) - FluxCD (cloudcoil.models.fluxcd) - Kyverno (cloudcoil.models.kyverno) Missing your favorite operator? I've made it super easy to add new integrations using our cookiecutter template and codegen tools. I'd especially love feedback on: 1. The API design - does it feel natural to Python devs? 2. Testing features - what else would make k8s testing easier? 3. Which operators/CRDs you'd most like to see integrated next Check out https://ift.tt/g0pYqlj or try it out with PyPI: cloudcoil https://ift.tt/g0pYqlj January 22, 2025 at 03:26AM

Show HN: Pytest-evals – Simple LLM apps evaluation using pytest https://ift.tt/0WsQuIz

Show HN: Pytest-evals – Simple LLM apps evaluation using pytest https://ift.tt/safyUEw January 21, 2025 at 11:33PM

Show HN: SupGen, an model-free program synthesizer by examples / dependent types https://ift.tt/qmenwAB

Show HN: SupGen, an model-free program synthesizer by examples / dependent types https://www.youtube.com/watch?v=bEP88ucXga January 21, 2025 at 12:33AM

Show HN: Morse Man – Learn Morse Code While Playing Hangman https://ift.tt/QtHOWI0

Show HN: Morse Man – Learn Morse Code While Playing Hangman I built a Hangman variant that teaches you Morse code while you play. Instead of typing letters directly, you enter them using dots and dashes. Features include: - Interactive Morse code input with visual feedback - Built-in cheatsheet (costs an attempt) - 100+ word dictionary - Animated hangman visualization Play it here: https://ift.tt/WB3v0sb Perfect for anyone interested in learning Morse code through gameplay. Would love feedback from the HN community! https://ift.tt/WB3v0sb January 20, 2025 at 11:27PM

Show HN: Zippd – Deploy static sites in seconds (OSS) https://ift.tt/WEAkKUX

Show HN: Zippd – Deploy static sites in seconds (OSS) I built a static site deployment tool similar to GitHub Pages or Firebase Hosting And It's Open Source, Link - https://ift.tt/VCyXERW https://zippd.app/ January 20, 2025 at 08:15AM

Show HN: TikTok Video Downloader https://ift.tt/YkD4n9g

Show HN: TikTok Video Downloader We just built a small tool to download all your tiktok videos by just providing your tiktok username. You can try it out in https://ift.tt/8OjQZHV Even though it's reinstated, with all the ban and no-ban conversation it's better to download all your videos and back it up. This is primarily aimed at creators who have a large number of videos. Please feel free to drop any feedback! https://ift.tt/Agova7F January 20, 2025 at 12:05AM

Show HN: We built an Anime Recommendation and streaming Website https://ift.tt/6IcSGzE

Show HN: We built an Anime Recommendation and streaming Website Me and my friend built an unique content based Recommendation System, where user can just select Anime or write synopsis and our system will find the most similar anime available. We used Qdrant Vector Database for the Recommendations. Other Features includes, Streaming, Custom watchlist creation and sharing of watchlists. We update our Database regularly and plan to introduce new features in future. https://aniversehd.com/ January 19, 2025 at 11:27PM

Show HN: Float Gallery, visualizations for various floating point formats https://ift.tt/ED2HZlQ

Show HN: Float Gallery, visualizations for various floating point formats https://ift.tt/u78oHDO January 19, 2025 at 09:19PM

Show HN: I built a simple Cron Expression Generator https://ift.tt/y2WldMx

Show HN: I built a simple Cron Expression Generator https://cronevery.day/ January 18, 2025 at 11:24PM

Show HN: ZX Spectrum SCR to PNG Converter https://ift.tt/QcLhXRl

Show HN: ZX Spectrum SCR to PNG Converter Scratching my own itch. I had to do this for showing information on ZX Spectrum games. So thought I'd turn it into a useful tool for other people to use. https://ift.tt/sJYvIy3 January 17, 2025 at 04:50PM

Show HN: Discorch – Offline tool to browse and delete your Discord messages https://ift.tt/c8bPShV

Show HN: Discorch – Offline tool to browse and delete your Discord messages Built this to help users manage their Discord message history. Upload your data package to browse messages and generate deletion requests that comply with Discord's requirements, all offline and locally. Discord's bulk deletion process is complex and poorly documented. With their recent push toward monetization and ads, users need better tools to control their data. Discorch makes this accessible by guiding you through the process step by step, with a simple and intuitive interface. Includes a Go CLI for attachment downloads. Search functionality needs improvement and there are some known bugs, but it works well for most use cases. Issues and PRs welcome at https://ift.tt/c04exvp . I'll keep an eye on the comments for feedback and bug reports! https://discorch.org January 18, 2025 at 04:30AM

Show HN: Watchfakenews.com https://ift.tt/akYSEI1

Show HN: Watchfakenews.com Hi everyone, we're democratizing access to deepfakes. Product is live.. try it out If the above url doesn't work, you can find us on https://ift.tt/ez23Y8x https://ift.tt/ez23Y8x January 18, 2025 at 01:58AM

Show HN: Compile C to Not Gates https://ift.tt/tWOkPZJ

Show HN: Compile C to Not Gates Hi! I've been working on the flipjump project, a programming language with 1 opcode: flip (invert) a bit, then jump (unconditionally). So a bit-flip followed by more bit-flips. It's effectively a bunch of NOT gates. This language, as poor as it sounds, is RICH. Today I completed my compiler from C to FlipJump. It takes C files, and compiles them into flipjump. I finished testing it all today, and it works! My key interest in this project is to stretch what we know of computing and to prove that anything can be done even with minimal power. I appreciate you reading my announcement, and be happy to answer questions. More links: - The flipjump language: https://ift.tt/Y7Qn1Ga https://ift.tt/8bjKV2C - c2fj python package https://ift.tt/rZKDeR6 https://ift.tt/CgfJyEh January 18, 2025 at 01:06AM

Show HN: News Minimalist – News ranked by significance https://ift.tt/LsIq3uA

Show HN: News Minimalist – News ranked by significance Hey HN! I'm the author of News Minimalist — a news aggregator where all news is ranked by significance on a scale from 0 to 10. The project was born out of personal pain — I wanted a way to read only significant news, like major humanity milestones, or historical political events, filtering out all the celebrity gossip and smartphone releases. But I couldn't find a way to do that — everywhere I looked, the news was ranked by popularity, coverage, or relevance, not significance. I first tried to solve the problem in the beginning of 2023 with GPT-3 (the top model at that time) by asking it to estimate the significance of some news stories. The results were painfully bad — for some reason, the model preferred tragic, personal stories, completely missing the essence of what makes the news significant. No amount of prompt engineering could fix that. But it all changed in March 2023 when GPT-4 came out. The scores it gave made much more sense. After a month of work, the first version was ready. News Minimalist had its first successful Hacker News post ( https://ift.tt/VXF0Roz ), and I realized that a lot of people had the same problem I had. I've been working on improving the project ever since. As probably most tech founders, I spent too much time on technical improvements, completely ignoring marketing. But I think that work paid off, and I'm finally satisfied with the scores it gives. The results are posted on the site: https://ift.tt/igeE5Hr Let me know what you think! Vadim https://ift.tt/igeE5Hr January 16, 2025 at 02:05AM

Show HN: Image-to-3D-to-Image in real-time https://ift.tt/Yiuo8L7

Show HN: Image-to-3D-to-Image in real-time https://twitter.com/krea_ai/status/1879929607320633870 January 17, 2025 at 01:30AM