Building with GenAI

Challenges, Learnings & a Bit of News from the Builders of Quivr.app

New features and learnings from the team at Quivr and news from across the exciting world of Generative AI.

Last week at Quivr

• Last week in Numbers

  • GitHub Trending n*1 for 4 days 🙌

  • 8k stars in a week

  • 1000 active users per day

  • 3 releases

  • 10 contributors

• New Features and Improvements

  • You can now share brains with other users!

This is a huge feature, and thank you to the team for getting it over the line 😍

Demo: Quivr.app - Source Code: GitHub

New Open Source Announcement 🎉

Code Review GPT - AI Code Reviewer

Improve code quality and catch bugs before you break production 🚀

• It saves us time in code reviews by spotting bugs

• Improves code quality by suggesting style changes

• Add the tool to your repositories!

• npm i code-review-gpt && npx code-review-gpt 

Give the repo a star ⭐️ if you like it 🙌

Learnings from the week

• Stan - Sensei: DB migrations are tricky. We made a migration that made our database explode 🤯 and had to delete some vectors in our demo DB [tweet]. Now I’m a Supabase contributor because I had to use a feature that should never be used 🤷‍♂️

• Zineb - Chief Dev: Releasing to prod 4 times a week is a big challenge. Embrace LEAN dev practises 🏃🏼‍♀️. Slowly but surely getting closer to our release goal with a 0 bug policy.

• Mamadou - Frontend Wizard: Custom hooks are reducing overhead in our development 🪝 They have helped us separate pure UI code from functions and have made our frontend easier to test 🧪

• Matt - Backend Dude: Building out private LLMs has a bunch of unique challenges. We are essentially building a scaled-down private cloud for our users 🫣 Setting the right domains is going to be key to ensuring our success in building a multi-tenancy 👀

• Chloe - Product Maestro: It’s essential to convey Quivr’s value clearly! We need to work on both features and the presentation of those features ⚖️ One cannot succeed without the other! We're also working on improving our efficiency in meetings and staying focused on providing value 🚀

• Brian - Methodology King: You miss what you have when it’s gone! High deployment velocity is important to maintain a healthy product and we broke that this week with the database issues 😬 This is the feature flow from idea conception, tickets for dev, to validation and releases to production ⏳

• Daniel - Feedback Warrior: Google blocked my email account, due to sending 500 daily emails for a Quivr survey 😔 The incident highlighted the need for a dedicated CRM, strict user consent, and adherence to platform policies for efficient community outreach 🤝

News from the World of Gen AI

• 🎉 Huge news that Meta just released Llama 2! Their next generation of open source large language model. Llama 2 is free for research and commercial use.

• 🧐 Worryingly GPT-4 is getting worse not better over time. Here is the link to the research paper which explains it in detail.

• 🧪 Langchain just announced LangSmith: a unified platform for debugging, testing, evaluating, and monitoring your LLM applications. We use it on the Quivr demo and it is really bridging the gap to help create production apps.

• 🛠️ Last but not least, go check out our new tool: Code-Review-GPT. Super useful for building fast and following Lean software principles.