- Classical Music (like [0])
    - Progressive Rock / Symphonic Rock (like [1])
    - Rock Opera (like [2])
    - Proto Metal (like [3])

    - Taking papers on music theory?
    - Playing at least 1 instrument?
    - Using software to tinker with sounds?

 I've also considered making my own audio software in C, maybe starting with 8-bit music

[1] - https://youtube.com/clip/UgkxfP0TplV2UPFFUGKWj-PvoAJwz3x3pJov?si=2nU7Ck3eu7mb7ZLA

[2] - https://www.youtube.com/watch?v=9hKYpNpajpI&t=520s

[3] - https://www.youtube.com/watch?v=jX0caf1HvNs

compose2nix is a CLI tool that automatically converts your Docker Compose project into a NixOS configuration[1].

* Why does this even exist?

Well, NixOS is a Linux distro that is configured using Nix (the language). However, Compose adds a configuration layer on top that is opaque to Nix, which means you lose out on some of the "magic" of NixOS. compose2nix solves this by "lifting" Compose functionality into native Nix/NixOS constructs, which results in a more native integration with NixOS.

In addition, this tool unlocks two primary use-cases:

1. You can run existing non-trivial applications - e.g., Immich - without worrying if they're natively packaged for NixOS.

2. You can bring your existing Compose stack over to NixOS and run it as-is without having to rewrite it in Nix. The auto-generated config can either act as a starting point, or you can keep your Compose file as the source of truth and re-generate it as you go.

* How does it work?

compose2nix maps Compose constructs into systemd services that run native Docker/Podman commands. Under the hood, the tool utilizes the same Compose library[2] used by the Docker CLI, so you also get Compose syntax checking "for free". In a way, it's a thin re-implementation of the Docker Compose CLI. One thing to point out is that compose2nix aims to natively support both Podman and Docker as container runtimes.

The README has a bit more detail. There's also a (somewhat rambly) video that explains what the project is about and includes a short demo of the tool: https://www.youtube.com/watch?v=hCAFyzJ81Pg

Oh, and please feel free to shoot me any questions or feedback (good or bad)!

[1]: In case it's your first time hearing about NixOS, here is a good overview: https://zero-to-nix.com/concepts/nixos

[2]: https://github.com/compose-spec/compose-go

https://www.youtube.com/watch?v=ht-bx7iqgYM

Ross and I met 8 years ago in Shanghai, where we worked at an education startup and organized tech and design events. When we realized Covid was creating a tsunami of webinars, Ross noticed the growing cost of editing all the new content as B2B companies replaced their in-person marketing channels with online events.

Most registrants to online events don't end up attending. They may be interested in the content, but they won’t take time to watch an entire webinar recording. Webinar content has a short shelf life unless it is reworked into a friendlier format. Doing that with traditional video editing software is cumbersome, so it often doesn’t happen. It’s time-intensive to review videos for key moments, ask designers to create appropriate graphics and captions, and receive final approval from managers.

We started out contacting companies organizing webinars, and learned they were stuck in a vicious cycle of constantly having to focus on the next upcoming event. We started manually editing videos for them to better understand how the most engaging bits could be reworked. Doing this manually revealed a glaring problem: the technology interfacing with video has changed dramatically, but the editing software hasn’t. Video editing software is designed for film makers or social media, and businesses creating video content have very different needs.

Milk Video uses a transcript-to-video based interface to review long recordings and minimize the mental effort around editing. We transcribe uploaded videos, present you with the content so you can quickly clip the best parts, and allow you to use templates to compose visually interesting layouts with additional assets, like logos or static text.

We made a drag-and-drop interface for creating short video clips with styled word-by-word captions. In a world where people often don't have their audio on, the timestamp information on a machine-generated transcript is perfect for creating interesting visual elements, such as captions styled one word at a time. This also makes content accessible by default. And because most webinars or Zoom recordings are visually similar, we have the ability to recommend which video templates might be best suited for their uploaded content in the future.

The frontend is a React app based on Redux Toolkit and Recoil.js. Our performant transcript interface is made possible due to Slate.js. Our backend is a Ruby on Rails app and depends on a non-trivial number of serverless functions hosted on Google Cloud and AWS. Our speech-to-text provider is AssemblyAI, who we found were both cheaper, faster and better than Google and Amazon.

We would love your feedback on the tool. We are spending a lot of time working directly with our first users, and would appreciate all of the input we can get. I’m also happy to go into detail around how any specific parts work! We’ll be in the comments and are eager to hear all your thoughts!

I’ve helped build infrastructure tooling from scratch at multiple companies and realized two things: that the shape of the solution with the advent of containerization, Kubernetes, and hyperscalers is quite similar across orgs, and that highly knowledgeable engineers are needed to build and manage this system.

Internal infrastructure teams juggle a multitude of tasks—provisioning cloud infrastructure, configuring runtimes, building code, securing artifacts, running tests, and deploying at scale. Post-deployment, they're also tasked with monitoring app performance, errors, uptime, and cost visibility. It's a lot of work, and having to build this tooling in-house is a deep inefficiency in engineering teams. The root of the problem is that AWS and GCP provide a lower level of abstraction than the entities, such as environments and applications, that developers have to deal with, and a ton of work is getting duplicated, often by underfunded teams, across many orgs. Argonaut’s objective is to be the developer platform and control center that you would otherwise have to build internally.

Argonaut provides an intuitive developer experience that simplifies working with Kubernetes and enables developer self-service, reducing the burden on devops and platform teams. We've productized this workflow orchestration, incorporating best practices to provide a push-to-deploy experience with flexible pipelines and scalable infrastructure, all within minutes.

Our users are startups across various domains like healthcare, IoT, fintech, AI, and SaaS products. Over the last two years, we’ve enabled customers to scale their engineering teams 10x and manage 10+ environments in parallel without needing a dedicated infra/DevOps team, saving them precious time and resources.

Argonaut lets you set up production-ready infrastructure and customize as you scale. We then let you set up automated deployments of your application in minutes. We offer configurable build-and-deploy pipelines, powered by Dagger and ArgoCD, and deep integration with GitHub Actions and GitLab CI. In addition, we support container registries, multiple cloud accounts, observability stacks, cost visibility providers, CDNs, and the entire helm chart ecosystem of Kubernetes, with more integrations on the way.

Key features include: (1) easily create environments encapsulating cloud infrastructure, applications, and deployment pipelines (2) autoscaling deployments for apps and cronjobs to GCP and AWS with a progression across environments (3) compose deployments across multiple environments with our visual pipeline builder (4) get cost estimates before making infra changes, giving you a clear understanding of your expenses; (5) managed Terraform state and pre-built modules that just work, fostering team collaboration on infrastructure.

Argonaut is self-serve, so you can sign up and start using the product right away: https://ship.argonaut.dev. There is a free tier that doesn't require a credit card to get started. We'd be delighted to have you try it, and are happy to help with onboarding.

If your teams work with AWS, GCP, or Kubernetes, I’d love to hear about your experiences, pain points, and what you think a product like Argonaut should be able to do. Looking forward to your comments!

Am I missing a tool or something? Shouldn't I be able to run my server in my IDE and proxy it into a Compose network or Kubernetes namespace, so I get my IDE tools for free? Or at least have my Docker container run in "watch" mode, where a change to one of the files the container is based on restarts the process with the new files?

Commodities are the center of global trade. The food you eat and the coffee you drink all start with agricultural producers. A trader organizes a shipment of it to a receiver near you, who then manufactures and distributes it to your local supermarket or coffee shop.

That sounds simple enough, but any shipment involving commodities involves around 200 processes, at least 50 documents, and extensive coordination through emails and phone calls. The amount of paper and processes is staggering. Errors can be catastrophically costly, ending up with lawsuits over millions of dollars.

Commodity trading companies allocate 40% of their workforce to logistics and operations. It’s a relationship-focused business. These aren't trading firms with Bloomberg terminals doing spot trading with algorithms all day. They are trading actual physical commodities. Many of these deals are agreed on via text message.

Now imagine you're managing thousands of shipments a month. Nearly every trader we’ve talked to can recall a time in the past 6 months where a document or process error resulted in the firm losing hundreds of thousands of dollars.

This led us to explore how we could tailor AI, automation, and collaboration tools to automate the handling of shipment workflows for more efficient operations and accurate documentation. Our team has the background to solve this problem—an ex-sugar trader with 8 years of experience and two engineers with a combined 14 years of experience from logistics and software companies.

There are a few layers to solving the pain points of these customers:

(1) Data Preprocessing: we give them a way to clean up and organize their existing data, which is usually a mess. We help them process documents, classify them, extract relevant fields, organize them by shipments, and search for them.

(2) We use AI and automation to leverage OCR, Multi-Modal Transformers, and indexing methods across 150+ different commodity and logistics document types to automate business processes.

(3) Actions and Workflows: Once the data is processed, we enable users to take actions or kickoff workflows with them, such as entering data into ERPs, using AI agents to compose emails, update freight matrices, track their shipments and predict delays, and more. All of these things are things the user does manually currently and are prone to error, so we automate them to save time and money. An ops team currently spends 60% of their day on these tasks. With our initial product, we are already saving them 2 hours a day.

Our customers so far include one of the largest U.S. agricultural traders, who use our platform to process over a million documents and automate internal workflows, like creating vessel nominations and updating shipment statuses. They also use us to find critical documents quickly. Importers and exporters randomly get audited to make sure the commodity they declared matches what was actually delivered (quality, weight, polarization). These audits can happen years after the shipment, and searching for the relevant documents is a needle/haystack situation. Sometimes they have been unable to find them, and you can imagine how well that goes over with the auditors. We make this easy.

We can’t give HN readers direct access to the product because it can only be used if you have commodity data and that data is sensitive. But here’s a video showing how it works: https://www.youtube.com/watch?v=slOGliSR6IM#t=26s

There probably aren’t too many commodity traders checking HN every day, but we figure some of you have worked as traders or know traders from previous work experience, and we’d love to hear whatever you think about this!

Here’s a video showcasing the platform and demonstrating how to easily manage new data and changes using the RAG stack of your choice: https://www.youtube.com/watch?v=-OdWRxMX1iA

If you want to try release.ai out, we’re offering a sandbox account with limited free GPU cycles so you can play around and get a feel for Release.ai: https://release.ai. We suggest playing around with some of the RAG AI templates and adding custom workflows like in the demo video. The sandbox comes with 5 free compute hours on an Amazon g5.2xlarge instance (A10 with 24GB VRAM, 8vCPUs and 32GB). You will also get 16 GB and 4vCPUs for cpu workloads such as web servers. You will be able to run an inference engine plus things like an api server, etc.

After the sandbox expires, you can switch to our free plan, which requires a credit card and associating an AWS/GCP account with Release to manage the compute in your cloud account. The free account provides 100 free managed environment hours a month. If you never go over, you never pay us anything. If you do, our pricing is here: https://release.com/pricing.

For those that like to read more, here’s the deeper background.

It’s clear that open source AI and AI privacy are going to be big. Yes, many developers are going to choose SaaS offerings like OpenAI to build their AI applications, but as open source frameworks and models improve, we’re seeing a shift to open source running on cloud. Security and privacy is a top concern of companies leveraging these SaaS solutions, which forces them to look at running infrastructure themselves. That’s where we hope to come in: we’ve built Release.ai so all your data, models and infrastructure stay in your cloud account and open source frameworks are first class citizens.

Orchestration - Integrating AI applications into a software development workflow and orchestrating their lifecycle is a new and different challenge than traditional web application development. Release also makes it possible to manage and integrate your web and AI apps using a single application and methodology.

To make orchestrating AI applications easier, we built a workflow engine that can create the complex workflows that AI applications require. For example, you can automate the redeployment of an AI inference server easily when underlying data changes using webhooks and our workflow engine.

Cost and expertise - Managing and scaling the hardware required to run AI workloads is hard and can be incredibly expensive. Release.ai lets you manage GPU compute resources across multiple clouds with different instance/node groups for various jobs within a single admin interface. We use K8s under the covers to pull this off. With over 5 years of building and running K8s infrastructure our customers have told us this is how it should be done.

Getting started with AI frameworks is time consuming and requires some pretty in-depth expertise. We built out a library of AI templates (https://docs.release.com/release.ai/release.ai-templates) using our Application Template format (which is kind of a super docker-compose: https://docs.release.com/reference-documentation/application...) for common open source frameworks to make it easy to get started developing AI applications. Setting up and getting these frameworks running is a hassle, so we made it one click to launch and deploy.

We currently have over 20 templates including temples for RAG applications, fine tuning and useful tools like Juypter notebooks, Promptfoo, etc. We worked closely with Docker and Nvidia to support their frameworks: GenAI and Nvidia NEMO/Nims. We plan to launch community templates soon after launch. If you have suggestions for more templates we should support, please let us know in the comments.

We’re thrilled to share Release.ai with you and would love to get your feedback. We hope you’ll try it out, and please let us know what you think!

fed up with cringy stock ringtones on my new phones, a few months ago I decided to "compose" my own ringtone for calls and notifications. It had to be minimal, no bells and whistles, low tones.

The result was nothing fancy, but that's exactly what I was looking for. I have been using it ever since and I'm pretty happy with it.

So after some time I decided to make a few more variants and share them online for other people who hate ringtones that sound like overblown symphonies, cheap disco hits or overly catchy jingles.

Would love to know if anyone is interested in this or if you're still stuck that default Samsung ringtone (https://www.youtube.com/watch?v=mjzdDqpGt6A) ;)

I've been using 102, btw.

FAQ

Q: What the project does?

A: A lot. You can build your own chatbot, interact with as many as 13 tools (search, wolfram, dall-e, blip, clip, ocr, pinecone, whisper, selenium, local files, etc.), pretty much most of the things you've already seen hyped or cool on the social media.

Q: Sounds close to… LangChain…?

A: Briefly, I think LangChain grew too fast and became the jack of all trades but the master of none. I'm sure they had their reasons for approaching things the way they did, and I don't want to make this post about them more than I already have. Others have had more thorough investigations of this topic and better rants than I would.

Q: Ok, then tell me why would I want to be part of it?

A: We're x2 core developers. Sometimes less is more, giving us time to think more deeply about designing the framework and making it accessible to others. Some principles:

- Ease of use and flexibility: we were heavily inspired by PyTorch, and we aimed to follow the same code structure one uses with torch. Our original intuition was that when you're introducing something new, tying it with something people are familiar with will make it more accessible (in terms of read/write). Not only this, but the initial recipe proved quite successful and replacing it with something else without concrete reasons is not worth doing IMHO. Moreover, one of our long-term visions is to have smooth integration with torch. We aim to grant SymbolicAI differentiable features. Imagine your chatbot learning to better use its memory (e.g. how to update its memory with relevant information).

- As in torch everything is a tensor, in our framework everything is a Symbol. A Symbol once defined gets accessed to some primitives (as an analogy think of PrimTorch) which would easily allow you to compose complex expressions or manipulate Symbol variables. This unlock very fast manipulations (i.e. dot notation <|object|><|dot|><|method|>).

- The hard work is done by decorators. We use them for the following reasons: (1) modularity, (2) composition, (3) flexibility, and (4) readability.

- We want to make a cohesive dev environment. I'm a script kiddo and I don't like to leave my terminal. I dislike web interfaces. I want to use my local env with my own setup. We have an experimental feature that is built on top of git and would enable package management. It's similar to pip, but for extensions built with our framework. Another long-term vision is to make accessible to anyone using our framework a quick share with the community. See https://github.com/ExtensityAI/symscribe for a showcase of how to do transcription and create youtube chapters with Whisper with our package manager.

There's much more to say, but I will stop here. Please check our GitHub README (https://github.com/Xpitfire/symbolicai) for a more deep dive or our latest tutorial video that highlights some relevant use-cases from a more high-level POV (https://www.youtube.com/watch?v=0AqB6SEvRqo).

I really do hope that at least some of you reading will get interested. We have so many goals we want to reach, so many ideas we want to test, and probably just as many bugs (we call them maggots just for fun) we need to fix.

We need you.

One of the most common uses of LLMs is to go beyond what traditional RPA or IDP can do when it comes to structuring unstructured documents. However, there are a lot of challenges in getting this done right from extraction of text data from PDFs, scanned images or other formats, prompt engineering, evaluation and integration with existing systems. This very specific use case is where Unstract can help teams move really fast, leveraging LLMs. By doing the heavy-lifting in this fast-changing ecosystem it lets engineers concentrate on implementing core business workflow automations.

Here's what's special about Unstract:

AGPL-licensed:
Unstract is available under the AGPL, which is friendly for personal or commercial use. Getting started is simply a matter of running a script, which should take care of everything leveraging Docker Compose. We've also made is very easy to update the container images should you wish to. If you have a Mac or a Linux laptop, you can get started in minutes.

LLMWhisperer:
Results from LLMs are as good as the input they receive. LLMWhisperer is a cloud service that given PDFs (native text, scanned) returns text data in a way LLMs are best able to "understand". LLMWhisperer is free for up to 100 pages a day. While Unstract supports LLMWhisperer, we're also adding support for unstructured.io and LlamaParse as well (they're both great services!). For more details: https://llmwhisperer.unstract.com

Strong, improving ecosystem support:
Unstract supports a variety of providers for LLMs, Vector Databases, Embeddings, Cloud File Storage systems and databases/data warehouses. A full list is available on our Github page (https://github.com/Zipstack/unstract).

No-code Prompt Studio:
You need a way to figure out how well your prompts are working for a set of documents to make sure the prompts you're writing are as generic as possible. This is not easy to do with a regular IDE and a file manager on your laptop. Prompt Studio a purpose-built, no-code environment that makes it easy for you to develop, run and debug prompts while referring to sample documents side-by-side. The whole project is stored as a YAML (export/import options coming soon).

Easy integrations:
Unstract supports the launching of APIs and ETL Pipelines that take in unstructured documents and produce JSON data or sync data with various data sources like Snowflake, Bigquery, Redshift, PostgreSQL, etc. Prompt Studio no-code projects can be launched as APIs or ETL Pipelines with ease in just minutes letting you leverage LLMs to solve business automation problems. Unstructured data APIs are useful when an existing application needs to automate the processing of unstructured docs. Unstructured data ETL Pipelines are useful when unstructured docs in cloud file or object stores need to be structured and pushed into databases or data warehouses.

Quick start:
See our documentation (https://docs.unstract.com) to walk through a quick start project structuring standard credit card statements from a couple of different banks. You can also view a 3-min video of Unstract here: https://www.youtube.com/watch?v=8aZh-pRwZh8

Enterprise-only features:
Unstract has a couple of enterprise/cloud/managed features that are not part of the open source offering: LLMEval uses more than one LLM to arrive at a consensus on extracted fields. If there is no consensus, the field is set to null (a null value is always better than a wrong value), after which a human can review the extraction results. There are a couple of advanced Prompt Studio features that can cut token usage (and thus costs) by up to 6x (Single Pass and Summary-based Extraction). Otherwise, all other features are available in our open source version.

I work as an Android developer at a large retailer, on the app that the in-store employees use to
see their schedules, view product details, get support, etc. Part of our engineering on-call
rotation involves helping these in-store employees with any app issues they may encounter. There are
always a few times every rotation where I know solving someone's problem would be way easier if I
could just see their screen as if I were in-store with them.

However, many of these employees are non-technical and wouldn't want to go through the onboarding of
TeamViewer or Zoom, are under time constraints in a busy store, or (understandably) just wouldn't
know how to get set up.

As a side project over the past few weekends, I've been building an Android screen sharing tool
called Zipshare (https://www.getzipshare.com, demo video: https://www.youtube.com/watch?v=Iuc3HBIuMZE)
that aims to be absolutely as easy as possible to use for the person sharing their screen. All the
sharer needs to do is receive an email (or eventually a text) on their Android device and click a link.
They will be sent to the Play Store, and upon launching the app, will automatically be sent to their
session. Users don't need to create an account or enter a meeting code - there's literally one screen
in the app ("xyz@abc.com invited you to share your screen, would you like to?") In fact, I originally
wanted to use Play Store Instant Apps to eliminate the installation step entirely, but you can't
open a raw socket from an Instant App, which I currently depend on... [1]

You can try the live demo by entering your email address and waiting on the next page, while clicking the link
you receive on an Android device. Allow the screen share and you'll see it appear in your browser
on the first device. When this is more built out, support agents will be able to sign in to their
account and click a button to generate a new link/send it out. I also want to integrate with
platforms like ServiceNow, PagerDuty, and Zendesk to allow generating screen sharing sessions
directly from tickets.

I focused on Android for this initial version because the wide variety of devices makes giving
support more difficult than on iOS, and because I have way more experience with Android :)

Other features I want to include are automatic notification suppression while sharing (for privacy),
an "instant connectivity check" that reports basic system state to the agent like cell/Wi-Fi signal
strength and whether a VPN is enabled, and an annotation feature so that agents can draw attention
to certain areas of the screen.

My goal with Zipshare was to do one thing and make the experience as streamlined as possible. For
example, there's no voice or video calling because chances are the two parties will already be on
some other kind of call.

Technical details:

* Screen sharing: WebRTC with signaling provided by a self-hosted LiveKit server instance

* Android side: Kotlin, Jetpack Compose, LiveKit SDK, Play Store Referrer API to route the session
ID through to the app install, "classic" Android 3rd party libraries (OkHttp, Retrofit, KotlinX
serialization, ...)

* Website and API: Node.js, Express, two Pug templates, one CSS file, 50ish lines of JS excluding
LiveKit, SQLite DB with one table.

Design details:

* Icon: Fiverr

* Colors: ChatGPT

* Landing page copy: The best I could come up with

Please give it a try and let me know what you think! I am eager to test this on as many
devices as possible, especially older or low-end devices. I built this because I wanted it, but
would it be useful for you or your team? Any feature requests?

[1] https://issuetracker.google.com/issues/191070661, naturally "Won't Fix", definitely something
to hack around in the future

(note: I was given the OK to repost this for a second chance)