Show HN: Memories, FOSS Google Photos alternative built for high performance https://ift.tt/34O7xAW

Show HN: Memories, FOSS Google Photos alternative built for high performance Memories is a FOSS Google Photos alternative that you can self-host (it runs as a Nextcloud plugin). Website: https://ift.tt/tCRedsj GitHub: https://ift.tt/S4skW8f Demo Server: https://ift.tt/GUjIY7A (demo runs in San Francisco on a free-tier cloud vm) Memories has been built ground-up for high performance and is extremely fast when configured correctly. In our testing environment, it can load a timeline view with 100k photos in under 500ms, including query and rendering time! Some features to highlight: * A timeline similar to Google Photos where you can skip to any time in history instantly. * AI-based tagging that runs locally on your server, identifying and tagging people and objects. * Albums and external sharing. * Metadata editing support * A world map of your photos, supported both on mobile and the web * Did I mention it's extremely fast? Would love to hear feedback from the HN community! :) https://ift.tt/tCRedsj March 22, 2024 at 12:55AM

Show HN: DaLMatian – Text2sql that works https://ift.tt/Cn5VEc4

Show HN: DaLMatian – Text2sql that works Hey HN, we've built DaLMatian, a text2sql product that meets the needs of data analysts working with enterprise data. We built this app because as a data analyst at an enterprise I could not find a text2sql product that was (1) actually useful for my day-to-day and (2) easy to set up on my computer. Existing products either fall apart when tested on gnarly enterprise data/queries or require going through a sales/integration process that I wasn't in a position to push for - I just wanted something that I could quickly set up to help make my job easier. Our goal is to make this a reality for any data analyst that feels the same. There are many constraints that make this reality difficult to achieve. The product needs to scale to databases with millions of columns and extract business logic from very complex queries. It also needs to be fast, at least faster than an analyst would take to write the query. On top of all this, an analyst needs to be allowed to use it from a security standpoint. Our app meets all the key requirements of an enterprise data analyst while also being lightweight enough to run locally on a typical laptop. Here's how it works. To get started, you simply need to open a file of past queries in our IDE (try it here: https://ift.tt/x0YS14y ) and add a file with your database schema (instructions here: https://ift.tt/bZ2gm5o ). There is also an option to connect a database to auto pull your schema (no actual data is seen by the LLM). We do not see anything you input since the app is local and the only external connection is with OpenAI. It's just like asking ChatGPT for help with queries, but in a streamlined way. If you'd download our free IDE and try to break it, we'd love to hear what you come up with! https://ift.tt/x0YS14y March 21, 2024 at 10:41PM

Show HN: Personal Knowledge Base Visualization https://ift.tt/3ltrpC5

Show HN: Personal Knowledge Base Visualization My personal knowledge base is hosted on GitHub at https://ift.tt/0VmrORy . It scans the documents I like every day using GitHub Action, Zotero, HackerNews upvote and Github Likes. It's not yet optimized for smartphones. It cost me $5 to host it for a year. https://ift.tt/PSNKbei March 21, 2024 at 03:28AM

Show HN: GritQL, a Rust CLI for rewriting source code https://ift.tt/8rL7V1y

Show HN: GritQL, a Rust CLI for rewriting source code Hi everyone! I’m excited to open source GritQL, a Rust CLI for searching and transforming source code. GritQL comes from my experiences with conducting large scale refactors and migrations. Usually, I would start exploring a codebase with grep. This is easy to start with, but most migrations end up accumulating additional requirements like ensuring the right packages are imported and excluding cases which don’t have a viable migration path. Eventually, to build a complex migration, I usually ended up having to write a full codemod program with a tool like jscodeshift. This comes with its own problems: - Most of the exploratory work has to be abandoned as you figure out how to represent your original regex search as an AST. - Reading/writing a codemod requires mentally translating from AST names back to what source code actually looks like. - Performance is often an afterthought, so iterating on a large codemod can be painfully slow. - Codemod frameworks are language-specific, so if you’re hopping between multiple languages—or trying to migrate a shared API—you have to learn different tools. GritQL is an attempt to develop a powerful middle ground: - Exploratory analysis is easy: just put a code snippet in backticks and use $metavariables for placeholders. - Incrementally add complexity by introducing side conditions with where clauses. - Reuse named patterns to avoid rebuilding queries, and use shared patterns from our standard library for common tasks like ensuring modules are imported. - Iterate on large codebases quickly: we use Rust for maximum performance GritQL has already been used on thousands of repositories for complex migrations[1] but we're excited to collaborate more with the open source community. [1] Ex. https://ift.tt/iVSYHk5 https://ift.tt/utJmezR March 21, 2024 at 12:53AM

Show HN: Automated Software Documentation for GitHub Codebases https://ift.tt/oicvE3f

Show HN: Automated Software Documentation for GitHub Codebases Hey Hackers, My team and I have been working on an automated software documentation and impact analysis platform for the last 3 years. Our long-term goal is to enter safety/mission-critical applications, where improper documentation can lead to disastrous outcomes, e.g., costly reworks/overruns or endangering human lives. But, in an effort to recognize revenue in the near term with our existing functionality, we have found initial traction with use cases focused on reverse engineering legacy systems. Where getting up to speed with an existing system requires a team of engineers to manually review large amounts of code, taking weeks or months to come to grips with. ______________________________________________ Our Self-Service release is a no-frills offering to leverage a subset of our document generation capabilities. Using only the code, SAFA is able to: -Summarize Code Files -Generate an overall project summary -Generate Upstream Documentation, like Features and Functional Requirements -Map relationships between all code and generated documentation with explanations Our approach leverages our own LLM pipeline, which applies a variety of clustering/refinement techniques, embedding models, and LLMs to keep your entire system within context when generating documentation, change summaries, api flow, and more. We do not use customer data to train or refine our models. We currently only support Github integrations for self-service but will implement flat-file support in the near term. When using self-service, you will receive Code Summaries and a Project Overview for free, but we charge for generating documentation and relationships: 20 cents per code file and generated document (100 File Codebase = $35). Currently, self-service has a 1000 code file limit. ________________________________________________ If you want to see the quality of the documents SAFA generates before trying it with your code, feel free to check out our public codebases page ( https://ift.tt/a43fD6s ). We have serious ones like Autoware's AV Control Module, and more fun ones, like Super Mario 64. Otherwise, our app is directly accessible via https://app.safa.ai (apologies, we do require an account to be made). I very much look forward to your feedback and insights. Feel free to email me directly at aarik@safa.ai. https://www.safa.ai March 21, 2024 at 12:24AM

Show HN: Cloud-native Stack for Ollama - Build locally and push to deploy https://ift.tt/cXvJZRB

Show HN: Cloud-native Stack for Ollama - Build locally and push to deploy https://ift.tt/Vg7mtSW March 19, 2024 at 11:36PM

Show HN: Real-time voice chat with AI, no transcription https://ift.tt/8YmDait

Show HN: Real-time voice chat with AI, no transcription Hi HN -- voice chat with AI is very popular these days, especially with YC startups ( https://twitter.com/k7agar/status/1769078697661804795 ). The current approaches all do a cascaded approach, with audio -> transcription -> language model -> text synthesis. This approach is easy to get started with, but requires lots of complexity and has a few glaring limitations. Most notably, transcription is slow, is lossy and any error propagates to the rest of the system, cannot capture emotional affect, is often not robust to code-switching/accents, and more. Instead, what if we fed audio directly to the LLM - LLM's are really smart, can they figure it out? This approach is faster (we skip transcription decoding) and less lossy/more robust because the big language model should be smarter than a smaller transcription decoder. I've trained a model in just that fashion. For more architectural information and some training details, see this first post: https://tincans.ai/slm . For details about this model and some ideas for how to prompt it, see this post: https://tincans.ai/slm3 . We trained this on a very limited budget but the model is able to do some things that even GPT-4, Gemini, and Claude cannot, eg reasoning about long-context audio directly, without transcription. We also believe that this is the first model in the world to conduct adversarial attacks and apply preference modeling in the speech domain. The demo is unoptimized (unquantized bf16 weights, default Huggingface inference, serverless speed bumps) but achieves 120ms time to first token with audio. You can basically think of it as Mistral 7B, so it'll be very fast and can also run basically anywhere. I am especially optimistic about embedded usage -- not needing the transcription step means that the resulting model is smaller and cheaper to use on the edge. Would love to hear your thoughts and how you would use it! Weights are Apache-2 and on Hugging Face: https://ift.tt/Q4fCmPG... https://ift.tt/J0KwjMa March 20, 2024 at 12:37AM