Show HN: Reverse Proxy as a Service https://ift.tt/gXAdlPQ

Show HN: Reverse Proxy as a Service https://SnowOwl.co March 9, 2023 at 03:13AM

Show HN: BBC “In Our Time”, categorised by Dewey Decimal, heavy lifting by GPT https://ift.tt/wZ9oFCp

Show HN: BBC “In Our Time”, categorised by Dewey Decimal, heavy lifting by GPT I'm a big fan of the BBC podcast In Our Time -- and (like most people) I've been playing with the OpenAI APIs. In Our Time has almost 1,000 episodes on everything from Cleopatra to the evolution of teeth to plasma physics, all still available, so it's my starting point to learn about most topics. But it's not well organised. So here are the episodes sorted by library code. It's fun to explore. Web scraping is usually pretty tedious, but I found that I could send the minimised HTML to GPT-3 and get (almost) perfect JSON back: the prompt includes the Typescript definition. At the same time I asked for a Dewey classification... and it worked. So I replaced a few days of fiddly work with 3 cents per inference and an overnight data run. My takeaway is that I'll be using LLMs as function call way more in the future. This isn't "generative" AI, more "programmatic" AI perhaps? So I'm interested in what temperature=0 LLM usage looks like (you want it to be pretty deterministic), at scale, and what a language that treats that as a first-class concept might look like. https://ift.tt/DEueKh6 March 9, 2023 at 12:58AM

Show HN: SearQ, RSS are still useful https://ift.tt/zk3iwJU

Show HN: SearQ, RSS are still useful https://searq.org March 8, 2023 at 07:28PM

Show HN: Construct Animate – our new browser-based animation tool https://ift.tt/sONqHG2

Show HN: Construct Animate – our new browser-based animation tool https://ift.tt/2ivsYQ1 March 8, 2023 at 03:47PM

Show HN: Co-locating Debian Bullseye with an evil maid https://ift.tt/q3pd4If

Show HN: Co-locating Debian Bullseye with an evil maid In order to facilitate the secure co-location of a server, I looked into protecting a Debian Bullseye system from evil maid attacks. In addition, since I've enjoyed using ZFS for some time, I decided to rely on a natively encrypted ZFS root file system. Basically... I'd like to take a system containing sensitive information, box it up, and drop it in the mail without worrying about losing it or having it wind up in the wrong hands. A couple of things became clear while researching how to do this. First, there should be little chance that a rogue data-center admin can insert malicious software. When the system reaches the data center and gets powered on we should be confident that it's running our software completely unmodified. As I understand things, Secure Boot is designed to help with this and therefore should be enabled. However, by relying on Secure Boot alone, there will be no remote method of knowing that it hasn't been disabled until after the ZFS pass-phrase is provided to the initramfs via dropbear. At that point it's too late. An evil maid could have already subverted dropbear, for example, and just now stolen the pass-phrase. To avoid this I realized that a second requirement of using a TPM device to automatically unlock the ZFS root was in order. TPM devices have the ability of "sealing" data to so-called Platform Configuration Registers (PCR). This feature allows the data to be accessed only if the "measured" system state matches some original expected state. The TPM can fully start the system unattended but, if anything's unexpectedly meddled with, act like a tripwire requiring the pass-phrase to be typed in manually. If we ssh in and reach dropbear requesting the pass-phrase, we'll know that we either need to update our sealed data after a grub/kernel/initramfs update... or someone's been messing with our start up code. This window of opportunity will be too small for an evil maid to take practical advantage of. This sounded like the right track and I set out to try and configure both, Secure Boot and TPM unlocking of an encrypted ZFS root. I thought it'd take a few hours at most but it actually turned out to be a fair challenge. After a few failed attempts I started tenaciously documenting every avenue. Ultimately I developed helper scripts that can reproduce the configuration should the time come to actually ship a machine out the door. I'm reasonably satisfied with the outcome. However, the scripts haven't been reviewed and neither has the overall process itself. There were a lot of guides I followed that contained typos, bugs, dubious information or simply different requirements. I'm not sure everything is exactly "bullet-proof" for this show HN. For example, I'm beginning to wonder if Secure Boot is necessary and if the TPM alone is sufficient. So naturally, comments and criticisms regarding everything are greatly appreciated. The script files can be found here: https://ift.tt/NRnJrEf and here: https://ift.tt/2tk1TDS Finally, I hope this effort will be useful to others facing similar needs. March 8, 2023 at 02:49PM

Show HN: Salesforce Announces Einstein GPT https://ift.tt/RsuAxob

Show HN: Salesforce Announces Einstein GPT Salesforce Announces Einstein GPT, the World’s First Generative AI for CRM https://ift.tt/7C83FWH March 8, 2023 at 11:38AM

Show HN: Postcard Bot – Send Any Photo as a Postcard with a Text Message https://ift.tt/KQ506Ju

Show HN: Postcard Bot – Send Any Photo as a Postcard with a Text Message Hey HN! This is a fun little project that I built over a weekend back in 2016, and then as life happened I let it dwindle and die. Per request from friends and family, I’ve brought it back to life. Check it out if you want! It’s surprisingly fun receiving physical postcards in the mail. I’ve also got WhatsApp integration and some more fun things in the works, so stay tuned! https://ift.tt/2FfJ416 March 8, 2023 at 05:55AM

Show HN: ChatGPT and Document Parser = Ghost https://ift.tt/VTb5DaF

Show HN: ChatGPT and Document Parser = Ghost I've always wanted to just upload a whole book to ChatGPT and ask questions. Obviously with the char limit that's impossible... So some buddies and I built Ghost. We have it limited to 5 pages for uploads for now, but plan on expanding the limit soon. Let me know what you guys think! https://ift.tt/RJ0ZBMK March 8, 2023 at 12:56AM

Show HN: Summarizing long form videos into easy to follow essay https://ift.tt/xKmtsb0

Show HN: Summarizing long form videos into easy to follow essay Introducing ClipRecaps, the ultimate tool for summarizing long-form videos. With advanced algorithms and natural language processing techniques, ClipRecaps provides concise summaries of key points from videos, saving users time and enabling them to make informed decisions about which videos to watch in full. Say goodbye to lengthy videos and stay informed with ClipRecaps. Founded by a team of researchers from NUS, ClipRecaps has become an essential tool for students, professionals, and anyone looking to stay up-to-date in a fast-paced world. Try ClipRecaps today and experience the future of video summarization. https://cliprecaps.com/ March 7, 2023 at 01:45PM

Show HN: Hello World Java Polyglot https://ift.tt/dGOsezP

Show HN: Hello World Java Polyglot I wanted to see how much simpler native code integration has become with GraalVM's polyglot, when compared to JNI. It's less than 100 lines all up (including the POM). (I feel kinda stupid posting a "hello world" to HN, but this is a hell of an improvement.) Linux only at the moment. https://ift.tt/jfIc7X3 March 7, 2023 at 11:31AM

Show HN: Roastedby.ai – Talk some trash, have some fun https://ift.tt/CZ6fjJd

Show HN: Roastedby.ai – Talk some trash, have some fun https://ift.tt/3N2MD6T March 7, 2023 at 06:25AM

Show HN: Simple Log Alerts to Slack https://ift.tt/7c1lVLn

Show HN: Simple Log Alerts to Slack There are many log alerting systems on the market. The best known is probably Datadog. There’s also Logtail, Papertrail, Splunk, Logstash and others. These are well put together products with a host of great features, such as excellent UIs, sophisticated live searching via web interfaces and sometimes query languages and alerting. They require various levels of installation and they have costs, either through volume-based tiered systems or monthly payments. For a bootstrapped business, this can be problematic, for instance when a surge of logs - indicating a possible important problem that needs to be solved - pushes volume on to another tier. Should the “log ransom” be paid? Instead, I recalled from earlier times surely the simplest log watcher: Swatchdog [1]. It is rather venerable software. Its file history from its source download shows dates in 2015, but it was written much earlier - the 90s or possibly 80s by Todd Atkins [2]. We wanted to have alerts in Slack - the blog explains how we did it. In short: *very simply*. The code is available [3]. [1]: https://ift.tt/sar46B8 [2]: https://ift.tt/CS19NA0 [3]: https://ift.tt/2rzjfVn https://ift.tt/rpeQUf5 March 6, 2023 at 03:40PM

Show HN: Total.js – Low-code development (Node-RED alternative) https://ift.tt/P69B47a

Show HN: Total.js – Low-code development (Node-RED alternative) https://ift.tt/Im5hVP4 March 7, 2023 at 12:39AM

Show HN: Open-Source Alternative to Loom https://ift.tt/0ouLtwS

Show HN: Open-Source Alternative to Loom Hey HN, we're excited to introduce Sorbay - an open source alternative to Loom for creating and sharing screen recordings. With Sorbay, you can easily record your screen, camera, and microphone all at once. It is a complete solution that comes with its own backend service, allowing you to instantly share a link of your recording as soon as it is finished. The video is streamed directly to the backend service as the recording happens to make this possible. With both founders based in different countries, we needed a tool to quickly share screen recordings to keep us up to date or to ask for feedback. Meetings are cool if you need to discuss something deeply, but for almost everything else a quick recording works better. We had to settle for one of the proprietary solutions because none of the open source tools allowed us to quickly share something with each other. Doing the recording is one aspect, but having the ability to instantly share a link was crucial. Waiting on a 400mb video upload to a Dropbox is just too much interruption if you want to quickly share something. The tipping point for us to actually build this open source tool came via an interaction from one of our day jobs. A third party provider sent a screen recording full of confidential information and to make things worse, all of it was uploaded by them to a different third party service. We strongly believe that information like this should stay within a company, ideally on infrastructure that they control themselves. Having a fully integrated open source solution is the best way to go for this. Our goal with this first public release is to gather feedback. The critical code paths are working, but it is still a bit rough to use. We deliberately cut out all non-essential features, but have a clear roadmap on what we want to release this year. There are a couple of known issues like audio glitches, non-working videos in Safari and crashing binaries that we hope to fix in the coming weeks. Later this year, we plan on releasing a cloud hosted version of Sorbay that would let you connect your own S3 storage provider. Additionally, we will be releasing an on-prem option focused on features for enterprises (SSO, RBAC, compliance). Both the Sorbay Client and the backend service are completely open source. For licensing we choose the AGPLv3 throughout the stack. The client is built with Vue.js on top of Electron. The use of Electron might be a bit controversial here on Hackernews but given the resources we currently have that was the only way that allowed us to get a working client out on all major platforms. The backend service is realized with Django. We use Keycloak for authentication and Minio for S3 compatible storage. All of this is run alongside Postgres and Redis, running on Docker containers which are managed by Docker Compose. We invite you to try Sorbay for yourself and join us on our issue tracker[1][2], Slack channel[3] or here on HN. Thanks for checking out Sorbay! [1]: https://ift.tt/TAjLas0 [2]: https://ift.tt/anJfI56 [3]: https://ift.tt/nr7vRkX... https://sorbay.io/ March 6, 2023 at 09:35PM

Show HN: OpenBB Terminal – Investment research for everyone https://ift.tt/L0AiCsK

Show HN: OpenBB Terminal – Investment research for everyone https://ift.tt/i3nuqhz March 5, 2023 at 03:34PM

Show HN: GraphQL-like full stack typesafe development for REST https://ift.tt/YAI40vL

Show HN: GraphQL-like full stack typesafe development for REST https://ift.tt/JyXRHi3 March 5, 2023 at 01:43PM

Show HN: ChatGPT, No Login Needed, Made with Official OpenAI API https://ift.tt/ubpmxln

Show HN: ChatGPT, No Login Needed, Made with Official OpenAI API https://chatgptbay.com/ March 5, 2023 at 12:45PM

Show HN: Llama-dl – high-speed download of LLaMA, Facebook's 65B GPT model https://ift.tt/7kBMzIi

Show HN: Llama-dl – high-speed download of LLaMA, Facebook's 65B GPT model https://ift.tt/hzdxC8y March 5, 2023 at 09:58AM

Show HN: gpt-graph. A simple, GPT-3 text to entity-relation graph generator https://ift.tt/UsRDa39

Show HN: gpt-graph. A simple, GPT-3 text to entity-relation graph generator Hi HN! This is a simple text to entity-relation graph generator, powered by gpt-3 davinci model. The purpose is to feed it actual written data, to obtain a graph representation of entities and relationships mentioned in the text. Also, being able to identify entity attributes like gender, size, age ... My initial goal, was to make it able to process a large amount of text into a big single graph. The problem being the 4000 token limit the model has, I decided to take the approach of feeding the text in batches, and try to merge the incoming graph with the existing information each time. This is done by comparing the incoming node labels with those already in the graph, adding the new information to the existing nodes. This works somewhat, but sometimes entities get duplicated if they are mentioned slightly differently in the text. The comparation method could use some improvement clearly. A nice feature, is that you get to decide what types you want to extract. So if, for example, you are interested only in people, and companies in the text, you can tell the model to stick to that. You can also leave the types to the model discretion. Also, the application allows for saving / loading graphs to json files. These files can be used with Cytoscape Desktop Application, which is a nice side effect of using cytoscape.js. in the UI. I think tools like this can really be of help when going through dense documentation. To have a visual representation of the concepts, entities or whatever, can be really helpful in education, investigation, legal ... Would love to hear your thoughts on how this could be improved. https://ift.tt/VhwgvHl March 5, 2023 at 05:17AM

Show HN: Tiny Metasearch Engine to Find Software Developers https://ift.tt/O42u3e9

Show HN: Tiny Metasearch Engine to Find Software Developers https://ift.tt/pfvkJSQ March 5, 2023 at 04:54AM