Show HN: JS Local-only AI Apps starter kit: cost $0 to run and test locally https://ift.tt/3AZQTuX

Show HN: JS Local-only AI Apps starter kit: cost $0 to run and test locally Hi HN! Here's a local-only stack I built over the weekend - hope it can be useful for you! I have been building a lot of AI apps - https://ift.tt/l10i7I6 https://ift.tt/nlerN4W ... And there were definitely times I spent way too much $$ before deploying the app to production. So I was looking for a "local only" stack and found a few tools that worked well together. I used the following set of tools but may add more options later: - Inference: Ollama - VectorDB: Supabase pg-vector - LLM orchestration: langchain - Framework: nextjs My learning from building ai-town was also that many people don't have the luxury of swiping a credit card to test an app with a lot of AI workloads -- not surprisingly, many models on Ollama are already very good - I used llama2-7b for majority of my dev work. Hope this can lower the barrier to entry for building AI apps. :) https://ift.tt/5c4UQu3 October 31, 2023 at 10:12PM

Show HN: An instantly usable To-Do List desktop Web App, no sign-up/log-in https://ift.tt/fkDObxW

Show HN: An instantly usable To-Do List desktop Web App, no sign-up/log-in ... by saving data in the URL, base64 encoded. About a year ago I wanted a fast, simple, throwaway, no sign-up/log-in To-Do List at my fingertips at work, I couldn't find one so I built one myself, in a quick and dirty way. Recently I decided to release to public, so I rewrote, productise it a bit, added markdown support, added dark mode, added extension for chromium browsers etc, and this is it. May you find this small utility website useful if you happen to need something like this as well. Some tech & non-tech bullets - Pure static website - Data goes up to URL on updating, base64 encoded, non-hashed spa style - Rewrote with React & Redux - ChatGPT assistance for markdown parsing, regex and its unit testing - Playwright for testing - CDN with AWS CloudFront Any feedback and comments are appreciated. ---- A note on saving data in the URL, To-Do lists usually wouldn't be too long so saving its data in URL seems a natural fit, plus my initial sketchy research showed ff max URL is 64k and chrome can go MBs. But actually in real world, if you take server and CDNs into consideration, 8k seems to be a de facto limit ( https://ift.tt/i28wJr3 ), still enough for usual todo lists :) Previous discussions on HN about saving data in URL, https://ift.tt/MR9ZLcK , https://ift.tt/bLI5wnm , https://ift.tt/ehqmcSn https://2doli.st/ November 1, 2023 at 03:10AM

Show HN: Undead – scary fast, bewitching web pages on the JVM with no JavaScript https://ift.tt/0OFptgP

Show HN: Undead – scary fast, bewitching web pages on the JVM with no JavaScript Hey HN. I wrote a LiveView implementation for the JVM called Undead. (The name Undead is a play on the "Live" part of LiveViews.) I had been waiting for the new StringTemplates feature of Java 21 which made it scary-easy to get the right data structures to make things work. It's been enchanting to write Java after about ten years in Javascript and Go lands. Would love folks to try it out and lmk what you think. If you aren't too scared that is. ;) https://ift.tt/5qrHuDW October 31, 2023 at 10:46PM

Show HN: Phind Model beats GPT-4 at coding, with GPT-3.5 speed and 16k context https://ift.tt/buILNWO

Show HN: Phind Model beats GPT-4 at coding, with GPT-3.5 speed and 16k context Hi HN, We’re excited to announce that Phind now defaults to our own model that matches and exceeds GPT-4’s coding abilities while running 5x faster. You can now get high quality answers for technical questions in 10 seconds instead of 50. The current 7th-generation Phind Model is built on top of our open-source CodeLlama-34B fine-tunes that were the first models to beat GPT-4’s score on HumanEval and are still the best open source coding models overall by a wide margin: https://ift.tt/Gudb98w... . This new model has been fine-tuned on an additional 70B+ tokens of high quality code and reasoning problems and exhibits a HumanEval score of 74.7%. However, we’ve found that HumanEval is a poor indicator of real-world helpfulness. After deploying previous iterations of the Phind Model on our service, we’ve collected detailed feedback and noticed that our model matches or exceeds GPT-4’s helpfulness most of the time on real-world questions. Many in our Discord community have begun using Phind exclusively with the Phind Model despite also having unlimited access to GPT-4. One of the Phind Model’s key advantages is that it's very fast. We’ve been able to achieve a 5x speedup over GPT-4 by running our model on H100s using the new TensorRT-LLM library from NVIDIA. We can achieve up to 100 tokens per second single-stream while GPT-4 runs around 20 tokens per second at best. Another key advantage of the Phind Model is context – it supports up to 16k tokens. We currently allow inputs of up to 12k tokens on the website and reserve the remaining 4k for web results. There are still some rough edges with the Phind Model and we’ll continue improving it constantly. One area where it still suffers is consistency — on certain challenging questions where it is capable of getting the right answer, the Phind Model might take more generations to get to the right answer than GPT-4. We’d love to hear your feedback. Cheers, The Phind Team https://ift.tt/ORTZvDX October 31, 2023 at 11:10PM

Show HN: Unlogged – open-source record and replay for Java https://ift.tt/YPX3hEa

Show HN: Unlogged – open-source record and replay for Java Hello HN! Parth, and Shardul here. We have been building unlogged.io for the last 21 months. We started as a time travel debugger and pivoted to record and replay with assertions, mocking, and code coverage. You can save the replays in the form of a JSON and commit them to your git. Both Parth and I come from an e-commerce/payments background where production bugs meant heavy financial losses. Big billion days/Black Friday sales meant months of code freezes with low productivity. Before committing the code, we wanted to replay production traffic and know the breaking changes right away, like in sub-second. Kind of like unit+integration tests on steroids. So, we built an SDK that adds probes to the code in compile time. The SDK logs code execution, in detail. Git: https://ift.tt/YBAWpUe We also built an IDE plugin that keeps monitoring code changes, hot reloads these changes, replays the relevant methods, and alerts on failing replays. It also lets developers call Java methods directly, mock downstream methods in run time, highlight code coverage in real-time, and show performance numbers for methods with inlay hints. (right above each method) Git: https://ift.tt/QBpaqd6 We are excited to launch the first version of our product that replays with assertions + mocking + code coverage reports right inside the IDE. Link to our IntelliJ plugin: https://ift.tt/nipXFND Record and Replay Demo: https://www.youtube.com/watch?v=muCyE-doEB0 Define Assertions on Replay: https://www.youtube.com/watch?v=YKsi1p634-M Track Code Coverage: https://www.youtube.com/watch?v=NMmp954kfaU Generate JUnit Test Cases: https://www.youtube.com/watch?v=rTUmg5b1Z_Q Mocking when replaying: https://www.youtube.com/watch?v=O_aqU1u-Kmw Documentation: https://ift.tt/RQFIBGx Roadmap: 1. Create a production logger -So that the performance impact is minimal -out of the box masking PII from production logs -creating meaningful input/return value combinations from production traffic to be replayed locally. 2. Creating a CI test runner that can integrate with CI/CD pipelines. 3. Auto-Replaying API endpoints of only the changed code. 4. Real-time alerts for the performance impact of code changes. 5. Creating a dashboard with reports, email/slack alerts. October 30, 2023 at 09:47PM

Show HN: I made a trivia game for fellow movie lovers https://ift.tt/kZPyMxa

Show HN: I made a trivia game for fellow movie lovers This is for those who, like me, are cinephiles. You just need to ask questions and try to guess the movie's name that ChatGPT picked. It's a great game to exercise your brain and memory. Give it a try and let me know what you think (: https://movieguess.com/ October 31, 2023 at 12:23AM

Show HN: Launch a private Ethereum Testnet with all clients and MEV infra https://ift.tt/pSN3yRf

Show HN: Launch a private Ethereum Testnet with all clients and MEV infra We've been working with the Ethereum Foundation & Flashbots to build tooling and infrastructure for developers to test various workflows for Ethereum. As part of that work, anyone can now spin up a local Ethereum dev net with the entire Flashbots mev-boost infra (relayer, builder, boost) using the ethereum-package. This package supports all EL and CL client types, works on Kubernetes for scale testing, and comes with a few bells and whistles like metrics, mock-builders, and beacon chain explorers. This is the de-facto tool for teams modifying and testing the consensus layer. Here's a full tutorial - https://ift.tt/qe4xBFA... https://ift.tt/E3AkQpL October 30, 2023 at 11:22PM