The best Hacker News stories from Show from the past day
Latest posts:
Show HN: Knowledge graph of restaurants and chefs, built using LLMs
Hi HN!<p>My latest side project is knowledge graph that maps the French culinary network using data extracted from restaurant reviews from LeFooding.com. The project uses LLMs to extract structured information from unstructured text.<p>Some technical aspects you may be interested in:<p>- Used structured generation to reliably parse unstructured text into a consistent schema<p>- Tested multiple models (Mistral-7B-v0.3, Llama3.2-3B, gpt4o-mini) for information extraction<p>- Created an interactive visualization using gephi-lite and Retina (WebGL)<p>- Built (with Claude) a simple Flask web app to clean and deduplicate the data<p>- Total cost for inferencing 2000 reviews with gpt4o-mini: less than 1€!<p>You can explore the visualization here: [Interactive Culinary Network](<a href="https://ouestware.gitlab.io/retina/1.0.0-beta.4/#/graph/?url=https://gist.githubusercontent.com/theophilec/351f17ece36477bc48438d5ec6d14b5a/raw/fa85a89541c953e8f00d6774fe42f8c4bd30fa47/graph.gexf&r=x&sa=re&ca[]=t&ca[]=ra-s&st[]=u&st[]=re&ed=u" rel="nofollow">https://ouestware.gitlab.io/retina/1.0.0-beta.4/#/graph/?url...</a>)<p>The code for the project is available on GitHub:
- Main project: <a href="https://github.com/theophilec/foudinge">https://github.com/theophilec/foudinge</a>
- Data cleaning tool: <a href="https://github.com/theophilec/foudinge-scrub">https://github.com/theophilec/foudinge-scrub</a><p>Happy to get feedback!
Show HN: Agents.json – OpenAPI Specification for LLMs
Hey HN, we’re building an open specification that lets agents discover and invoke APIs with natural language, built on the OpenAPI standard. agents.json clearly defines the contract between LLMs and API as a standard that's open, observable, and replicable.
Here’s a walkthrough of how it works: <a href="https://youtu.be/kby2Wdt2Dtk?si=59xGCDy48Zzwr7ND" rel="nofollow">https://youtu.be/kby2Wdt2Dtk?si=59xGCDy48Zzwr7ND</a>.<p>There’s 2 parts to this:<p>1. An agents.json file describes how to link API calls together into outcome-based tools for LLMs. This file sits alongside an OpenAPI file.<p>2. The agents.json SDK loads agents.json files as tools for an LLM that can then be executed as a series of API calls.<p>Why is this worth building?
Developers are realizing that to use tools with their LLMs in a stateless way, they have to implement an API manually to work with LLMs. We see devs sacrifice agentic, non-deterministic behavior for hard-coded workflows to create outcomes that can work. agents.json lets LLMs be non-deterministic for the outcomes they want to achieve and deterministic for the API calls it takes to get there.<p>We’ve put together some real examples if you're curious what the final output looks like. Under the hood, these LLMs have the same system prompt and we plug in a different agents.json to give access to different APIs. It’s all templatized.<p>- Resend (<a href="https://demo.wild-card.ai/resend">https://demo.wild-card.ai/resend</a>)<p>- Google Sheets (<a href="https://demo.wild-card.ai/googlesheets">https://demo.wild-card.ai/googlesheets</a>)<p>- Slack (<a href="https://demo.wild-card.ai/slack">https://demo.wild-card.ai/slack</a>)<p>- Stripe (<a href="https://demo.wild-card.ai/stripe">https://demo.wild-card.ai/stripe</a>)<p>We really wanted to solve real production use cases, and knew this couldn’t just be a proxy. Our approach allows you to make API calls from your own infrastructure. The open-source specification + runner package make this paradigm possible. Agents.json is truly stateless; the client manages all memory/state and it can be deployed on existing infra like serverless environments.<p>You might be wondering - <i>isn’t OpenAPI enough?</i> Why can’t I just put that in the LLM’s context?<p>We thought so too, at first, when building an agent with access to Gmail. But putting the API spec into LLM context gave us poor accuracy in tool selection and in tool calling. Even with cutting down our output space to 5-10 endpoints, we’d see the LLMs fail to select the right tool. We wanted the LLM to just work given an outcome rather than having it reason each time which series of API calls to make.<p>The Gmail API, for example, has endpoints to search for threads, list the emails in a thread, and reply with an email given base64 RFC 822 content. All that has to happen in order with the right arguments for our agent to reply to a thread. We found that APIs are designed for developers, not for LLMs.<p>So we implemented agents.json. It started off as a config file we were using internally that we slowly started adding features to like auth registration, tool search, and multiple API sources. 3 weeks ago, Dharmesh (CTO of Hubspot) posted about the concept of a specification that could translate APIs for LLMs. It sounded a lot like what we already had working internally and we decided to make it open source. We built agents.json for ourselves but we’re excited to share it.<p>In the weeks since we’ve put it out there, agents.json has 10 vetted API integrations (some of them official) and more are being added every day. We recently made the tool search and custom collection platform free for everyone so it’s even easier for devs to scale the number of tools. (<a href="https://wild-card.ai">https://wild-card.ai</a>)<p>Please tell us what you think! Especially if you’re building agents or creating APIs!
Show HN: Agents.json – OpenAPI Specification for LLMs
Hey HN, we’re building an open specification that lets agents discover and invoke APIs with natural language, built on the OpenAPI standard. agents.json clearly defines the contract between LLMs and API as a standard that's open, observable, and replicable.
Here’s a walkthrough of how it works: <a href="https://youtu.be/kby2Wdt2Dtk?si=59xGCDy48Zzwr7ND" rel="nofollow">https://youtu.be/kby2Wdt2Dtk?si=59xGCDy48Zzwr7ND</a>.<p>There’s 2 parts to this:<p>1. An agents.json file describes how to link API calls together into outcome-based tools for LLMs. This file sits alongside an OpenAPI file.<p>2. The agents.json SDK loads agents.json files as tools for an LLM that can then be executed as a series of API calls.<p>Why is this worth building?
Developers are realizing that to use tools with their LLMs in a stateless way, they have to implement an API manually to work with LLMs. We see devs sacrifice agentic, non-deterministic behavior for hard-coded workflows to create outcomes that can work. agents.json lets LLMs be non-deterministic for the outcomes they want to achieve and deterministic for the API calls it takes to get there.<p>We’ve put together some real examples if you're curious what the final output looks like. Under the hood, these LLMs have the same system prompt and we plug in a different agents.json to give access to different APIs. It’s all templatized.<p>- Resend (<a href="https://demo.wild-card.ai/resend">https://demo.wild-card.ai/resend</a>)<p>- Google Sheets (<a href="https://demo.wild-card.ai/googlesheets">https://demo.wild-card.ai/googlesheets</a>)<p>- Slack (<a href="https://demo.wild-card.ai/slack">https://demo.wild-card.ai/slack</a>)<p>- Stripe (<a href="https://demo.wild-card.ai/stripe">https://demo.wild-card.ai/stripe</a>)<p>We really wanted to solve real production use cases, and knew this couldn’t just be a proxy. Our approach allows you to make API calls from your own infrastructure. The open-source specification + runner package make this paradigm possible. Agents.json is truly stateless; the client manages all memory/state and it can be deployed on existing infra like serverless environments.<p>You might be wondering - <i>isn’t OpenAPI enough?</i> Why can’t I just put that in the LLM’s context?<p>We thought so too, at first, when building an agent with access to Gmail. But putting the API spec into LLM context gave us poor accuracy in tool selection and in tool calling. Even with cutting down our output space to 5-10 endpoints, we’d see the LLMs fail to select the right tool. We wanted the LLM to just work given an outcome rather than having it reason each time which series of API calls to make.<p>The Gmail API, for example, has endpoints to search for threads, list the emails in a thread, and reply with an email given base64 RFC 822 content. All that has to happen in order with the right arguments for our agent to reply to a thread. We found that APIs are designed for developers, not for LLMs.<p>So we implemented agents.json. It started off as a config file we were using internally that we slowly started adding features to like auth registration, tool search, and multiple API sources. 3 weeks ago, Dharmesh (CTO of Hubspot) posted about the concept of a specification that could translate APIs for LLMs. It sounded a lot like what we already had working internally and we decided to make it open source. We built agents.json for ourselves but we’re excited to share it.<p>In the weeks since we’ve put it out there, agents.json has 10 vetted API integrations (some of them official) and more are being added every day. We recently made the tool search and custom collection platform free for everyone so it’s even easier for devs to scale the number of tools. (<a href="https://wild-card.ai">https://wild-card.ai</a>)<p>Please tell us what you think! Especially if you’re building agents or creating APIs!
Show HN: Agents.json – OpenAPI Specification for LLMs
Hey HN, we’re building an open specification that lets agents discover and invoke APIs with natural language, built on the OpenAPI standard. agents.json clearly defines the contract between LLMs and API as a standard that's open, observable, and replicable.
Here’s a walkthrough of how it works: <a href="https://youtu.be/kby2Wdt2Dtk?si=59xGCDy48Zzwr7ND" rel="nofollow">https://youtu.be/kby2Wdt2Dtk?si=59xGCDy48Zzwr7ND</a>.<p>There’s 2 parts to this:<p>1. An agents.json file describes how to link API calls together into outcome-based tools for LLMs. This file sits alongside an OpenAPI file.<p>2. The agents.json SDK loads agents.json files as tools for an LLM that can then be executed as a series of API calls.<p>Why is this worth building?
Developers are realizing that to use tools with their LLMs in a stateless way, they have to implement an API manually to work with LLMs. We see devs sacrifice agentic, non-deterministic behavior for hard-coded workflows to create outcomes that can work. agents.json lets LLMs be non-deterministic for the outcomes they want to achieve and deterministic for the API calls it takes to get there.<p>We’ve put together some real examples if you're curious what the final output looks like. Under the hood, these LLMs have the same system prompt and we plug in a different agents.json to give access to different APIs. It’s all templatized.<p>- Resend (<a href="https://demo.wild-card.ai/resend">https://demo.wild-card.ai/resend</a>)<p>- Google Sheets (<a href="https://demo.wild-card.ai/googlesheets">https://demo.wild-card.ai/googlesheets</a>)<p>- Slack (<a href="https://demo.wild-card.ai/slack">https://demo.wild-card.ai/slack</a>)<p>- Stripe (<a href="https://demo.wild-card.ai/stripe">https://demo.wild-card.ai/stripe</a>)<p>We really wanted to solve real production use cases, and knew this couldn’t just be a proxy. Our approach allows you to make API calls from your own infrastructure. The open-source specification + runner package make this paradigm possible. Agents.json is truly stateless; the client manages all memory/state and it can be deployed on existing infra like serverless environments.<p>You might be wondering - <i>isn’t OpenAPI enough?</i> Why can’t I just put that in the LLM’s context?<p>We thought so too, at first, when building an agent with access to Gmail. But putting the API spec into LLM context gave us poor accuracy in tool selection and in tool calling. Even with cutting down our output space to 5-10 endpoints, we’d see the LLMs fail to select the right tool. We wanted the LLM to just work given an outcome rather than having it reason each time which series of API calls to make.<p>The Gmail API, for example, has endpoints to search for threads, list the emails in a thread, and reply with an email given base64 RFC 822 content. All that has to happen in order with the right arguments for our agent to reply to a thread. We found that APIs are designed for developers, not for LLMs.<p>So we implemented agents.json. It started off as a config file we were using internally that we slowly started adding features to like auth registration, tool search, and multiple API sources. 3 weeks ago, Dharmesh (CTO of Hubspot) posted about the concept of a specification that could translate APIs for LLMs. It sounded a lot like what we already had working internally and we decided to make it open source. We built agents.json for ourselves but we’re excited to share it.<p>In the weeks since we’ve put it out there, agents.json has 10 vetted API integrations (some of them official) and more are being added every day. We recently made the tool search and custom collection platform free for everyone so it’s even easier for devs to scale the number of tools. (<a href="https://wild-card.ai">https://wild-card.ai</a>)<p>Please tell us what you think! Especially if you’re building agents or creating APIs!
Show HN: Recommendarr – AI Driven Recommendations Based on Sonarr/Radarr Media
Hello HN!<p>I've built a web app that helps you discover new shows and movies you'll actually enjoy by:<p>- Connecting to your Sonarr/Radarr/Plex instances to understand your media library<p>- Leveraging your Plex watch history for personalized recommendations<p>- Using the LLM of your choice to generate intelligent suggestions<p>- Simple setup: Easy integration with your existing media stack<p>- Flexible AI options: Works with OpenAI-compatible APIs like OpenRouter, or run locally via LM Studio, Ollama, etc.<p>- Personalized recommendations: Based on what you actually watch.<p>While it's still a work in progress, it's already quite functional and I'd love your feedback!
Show HN: Recommendarr – AI Driven Recommendations Based on Sonarr/Radarr Media
Hello HN!<p>I've built a web app that helps you discover new shows and movies you'll actually enjoy by:<p>- Connecting to your Sonarr/Radarr/Plex instances to understand your media library<p>- Leveraging your Plex watch history for personalized recommendations<p>- Using the LLM of your choice to generate intelligent suggestions<p>- Simple setup: Easy integration with your existing media stack<p>- Flexible AI options: Works with OpenAI-compatible APIs like OpenRouter, or run locally via LM Studio, Ollama, etc.<p>- Personalized recommendations: Based on what you actually watch.<p>While it's still a work in progress, it's already quite functional and I'd love your feedback!
Show HN: Recommendarr – AI Driven Recommendations Based on Sonarr/Radarr Media
Hello HN!<p>I've built a web app that helps you discover new shows and movies you'll actually enjoy by:<p>- Connecting to your Sonarr/Radarr/Plex instances to understand your media library<p>- Leveraging your Plex watch history for personalized recommendations<p>- Using the LLM of your choice to generate intelligent suggestions<p>- Simple setup: Easy integration with your existing media stack<p>- Flexible AI options: Works with OpenAI-compatible APIs like OpenRouter, or run locally via LM Studio, Ollama, etc.<p>- Personalized recommendations: Based on what you actually watch.<p>While it's still a work in progress, it's already quite functional and I'd love your feedback!
Show HN: Robyn – “Batman Inspired” Python Web Framework Built with Rust
Show HN: I built a memory-safe web server in Rust
The web server that I am building is currently in beta, so any feedback is welcome.
Show HN: I built a memory-safe web server in Rust
The web server that I am building is currently in beta, so any feedback is welcome.
Show HN: I built a modern Goodreads alternative
Since 2005, Goodreads has been the default book tracking site, connecting millions of readers. But let’s be real—it’s barely changed in 20 years. It’s the same site it was, just with more ads.<p><pre><code> Still no half-star ratings.
No proper DNF (Did Not Finish) option.
UI still looks like it's from 2005.
Amazon owns it and doesn't care.
</code></pre>
So I built Kaguya, a modern alternative, over the past 9 months.<p>What’s live:<p><pre><code> Custom shelves (Organize however you want)
Rich-text reviews (format your thoughts properly)
10-star rating system (More nuance than 5 stars)
DNF, On-Hold, and other reading statuses
Likes, shares, comments on reviews
Import your library from Goodreads/StoryGraph
A beautiful design that doesn’t make you feel like you’re using an ancient website
Coming next:
Deep tagging system (Genres, moods, character traits, tropes)
Beautiful stats & insights (Visualize your reading habits)
Discussion forums for every book (Think subreddit-style discussions)
</code></pre>
Would love feedback. What do you think?
Show HN: I built a modern Goodreads alternative
Since 2005, Goodreads has been the default book tracking site, connecting millions of readers. But let’s be real—it’s barely changed in 20 years. It’s the same site it was, just with more ads.<p><pre><code> Still no half-star ratings.
No proper DNF (Did Not Finish) option.
UI still looks like it's from 2005.
Amazon owns it and doesn't care.
</code></pre>
So I built Kaguya, a modern alternative, over the past 9 months.<p>What’s live:<p><pre><code> Custom shelves (Organize however you want)
Rich-text reviews (format your thoughts properly)
10-star rating system (More nuance than 5 stars)
DNF, On-Hold, and other reading statuses
Likes, shares, comments on reviews
Import your library from Goodreads/StoryGraph
A beautiful design that doesn’t make you feel like you’re using an ancient website
Coming next:
Deep tagging system (Genres, moods, character traits, tropes)
Beautiful stats & insights (Visualize your reading habits)
Discussion forums for every book (Think subreddit-style discussions)
</code></pre>
Would love feedback. What do you think?
Show HN: Torii – a framework agnostic authentication library for Rust
Show HN: I took over Painteresque, a 14 year old Objective-C app and rebuilt it
Around 3 weeks ago, I came across a reddit post from someone who was looking to see if anyone was interested in taking over a 12-14 year old objective c app which hadn't been on the app store for many years. I responded and decided to do it.<p>The app is called Painteresque. It lets you turn photos to charcoal drawings, paintings etc while doing all the processing locally on device:<p><a href="https://apps.apple.com/ca/app/painteresque-photo-to-sketch/id6742038583">https://apps.apple.com/ca/app/painteresque-photo-to-sketch/i...</a><p>The app is entirely free to use, has no ads, third party analytics, trackers, subscriptions or other junk. It does offer 5 options for tips (which are basically consumable in app purchases), portion of which I will share with the original developer. However, based on my past experiences with tips, they are rarely given by users.<p>Here's a screenshot of the OLD app:<p><a href="https://i.imgur.com/fPhoqlp.jpeg" rel="nofollow">https://i.imgur.com/fPhoqlp.jpeg</a><p>The old app was built like 12-14 years ago in a combination of objective c and vanilla C. The "filters" were built in C while the UI was in objective c.<p>The codebase was so old that it wasn't even using ARC (Automatic Reference Counting).<p>The app would also instantly crash because back in those days, using the camera or adding photos to photo library didn't require declaring permission things like `NSCameraUsageDescription`, `NSLocationWhenInUseUsageDescription`.<p>Also had to remove obsolete frameworks to get it to run on devices iOS 12 onwards.<p>It was also not built for retina displays, nor handling safe area layout guides.<p>After fixing all these things, I got the old app to work on phone. But the UI was very rough and rudimentary. Back in those days, it would get approved by Apple. But nowadays, Apple is very strict about these things.<p>So, I decided to rebuild the app from scratch in Swift with, what I think, is a better UI design. I still used the old vanilla C code for the filters with some small changes and extra filters of my own. But the rest of the UI is all from scratch. I also incorporated things like `NSOperationQueue` for background processing of the image filtration as it can take a few seconds to generate. I incorporated caching of the images, preview thumbnails etc. to disk because the current phones have cameras which take humongous photos that can use a lot of ram if kept in memory.<p>The only package dependency the app uses is SnapKit to make auto layout simpler. The UI is made with code instead of storyboards.<p>The App Store screenshots are all made by me from scratch using Photopea, a browser alternative to Photoshop. The app icon is still the old icon, except I changed the background to be an off-white color instead of the old blue color to make it look a bit more modern.<p>Overall, this process took about 3 weeks. Most of the important work was done in about 1.5-2 weeks. The rest of the time was spent adding things like tipping in app purchases, UI improvements, tweaking the filters, adding feedback button etc.<p>The App Store review was fast. I submitted at 6:30 PM and it was approved at 4 AM. Also got an update approved last night to fix some slowness issue when opening new images.<p>The new app on iOS is only around 1.8 MB, most of it used by images for the buttons and app icon.<p>Ask me anything you'd like about this experience. Have any of you taken over an old app and rebuilt it?
Show HN: I took over Painteresque, a 14 year old Objective-C app and rebuilt it
Around 3 weeks ago, I came across a reddit post from someone who was looking to see if anyone was interested in taking over a 12-14 year old objective c app which hadn't been on the app store for many years. I responded and decided to do it.<p>The app is called Painteresque. It lets you turn photos to charcoal drawings, paintings etc while doing all the processing locally on device:<p><a href="https://apps.apple.com/ca/app/painteresque-photo-to-sketch/id6742038583">https://apps.apple.com/ca/app/painteresque-photo-to-sketch/i...</a><p>The app is entirely free to use, has no ads, third party analytics, trackers, subscriptions or other junk. It does offer 5 options for tips (which are basically consumable in app purchases), portion of which I will share with the original developer. However, based on my past experiences with tips, they are rarely given by users.<p>Here's a screenshot of the OLD app:<p><a href="https://i.imgur.com/fPhoqlp.jpeg" rel="nofollow">https://i.imgur.com/fPhoqlp.jpeg</a><p>The old app was built like 12-14 years ago in a combination of objective c and vanilla C. The "filters" were built in C while the UI was in objective c.<p>The codebase was so old that it wasn't even using ARC (Automatic Reference Counting).<p>The app would also instantly crash because back in those days, using the camera or adding photos to photo library didn't require declaring permission things like `NSCameraUsageDescription`, `NSLocationWhenInUseUsageDescription`.<p>Also had to remove obsolete frameworks to get it to run on devices iOS 12 onwards.<p>It was also not built for retina displays, nor handling safe area layout guides.<p>After fixing all these things, I got the old app to work on phone. But the UI was very rough and rudimentary. Back in those days, it would get approved by Apple. But nowadays, Apple is very strict about these things.<p>So, I decided to rebuild the app from scratch in Swift with, what I think, is a better UI design. I still used the old vanilla C code for the filters with some small changes and extra filters of my own. But the rest of the UI is all from scratch. I also incorporated things like `NSOperationQueue` for background processing of the image filtration as it can take a few seconds to generate. I incorporated caching of the images, preview thumbnails etc. to disk because the current phones have cameras which take humongous photos that can use a lot of ram if kept in memory.<p>The only package dependency the app uses is SnapKit to make auto layout simpler. The UI is made with code instead of storyboards.<p>The App Store screenshots are all made by me from scratch using Photopea, a browser alternative to Photoshop. The app icon is still the old icon, except I changed the background to be an off-white color instead of the old blue color to make it look a bit more modern.<p>Overall, this process took about 3 weeks. Most of the important work was done in about 1.5-2 weeks. The rest of the time was spent adding things like tipping in app purchases, UI improvements, tweaking the filters, adding feedback button etc.<p>The App Store review was fast. I submitted at 6:30 PM and it was approved at 4 AM. Also got an update approved last night to fix some slowness issue when opening new images.<p>The new app on iOS is only around 1.8 MB, most of it used by images for the buttons and app icon.<p>Ask me anything you'd like about this experience. Have any of you taken over an old app and rebuilt it?
Show HN: I made a website where you can create your own "Life in Weeks" timeline
Show HN: I made a website where you can create your own "Life in Weeks" timeline
Show HN: Berlin Swapfest – Electronics flea market
Hey HN,<p>After scrounging the local classifieds and our version of Craigslist to find pre-owned equipment and getting frustrated I decided that there needs to be a flea market but for electronics tools and homelab stuff. The Berlin Swapfest follows in the same spirit as the MIT Swapfest!<p>I’ve partners with a long time hacker space in Berlin c-base who will be hosting this event!<p>Would love for local hackers to come by and buy and sell their old gear!<p>Details on the site!
Show HN: Berlin Swapfest – Electronics flea market
Hey HN,<p>After scrounging the local classifieds and our version of Craigslist to find pre-owned equipment and getting frustrated I decided that there needs to be a flea market but for electronics tools and homelab stuff. The Berlin Swapfest follows in the same spirit as the MIT Swapfest!<p>I’ve partners with a long time hacker space in Berlin c-base who will be hosting this event!<p>Would love for local hackers to come by and buy and sell their old gear!<p>Details on the site!
Show HN: Berlin Swapfest – Electronics flea market
Hey HN,<p>After scrounging the local classifieds and our version of Craigslist to find pre-owned equipment and getting frustrated I decided that there needs to be a flea market but for electronics tools and homelab stuff. The Berlin Swapfest follows in the same spirit as the MIT Swapfest!<p>I’ve partners with a long time hacker space in Berlin c-base who will be hosting this event!<p>Would love for local hackers to come by and buy and sell their old gear!<p>Details on the site!