I've been considering good ways to use a task queue for this, and might just settle for a rudimentary one in a Postgres table.
The upside is that agent subtasks can be load balanced among servers, tasks won't be dropped if the process is killed, and better observability comes along with it.
The downside is definitely complexity. I'm having a hard time planning out an architecture that doesn't significantly increase the complexity of my agent code.
Others have given some decent advice based on your comment, but would you be interested in a ~30 minute (video) call to dive a bit deeper so I can give more tailored suggestions?
Location: Wisconsin, US
Remote: Yes
Willing to relocate: No
Technologies: TypeScript, Node.js, React, Express, ESLint, PostgreSQL, AI Prompting / Agents, Solidity, HTML, CSS (Sass, Panda), etc.
Resume: https://drive.google.com/file/d/1AMqjgI_QkUnq8hbydyUC87D2-Duvlvif/view?usp=drivesdk
Email: trevinhofmann@gmail.com
I'm a full-stack software engineer with ~10 years of experience and a B.S. in Software Engineering, who has primarily worked in fintech (PayPal, American Express, BGC Partners, Stripe). I'm closing a 2-year stint of solo freelancing for a wide variety of clients, along with launching my own AI software-development SaaS (https://mysticode.ai).
My focus has been full-stack web development for complex apps, but I love learning and would be open to considering other Engineering areas for the right company.
I did take a look at that, which is probably more than most people would have done, and by take a look I mean a skimmed for images because I'm not reading 2000 words of text for an app I don't even use yet. The only images I found showed how drag and drop works.
I know this is common with projects that think Github is a replacement for a website, but I genuinely wonder how does it get so bad that a 5 year project with 9000 commits and 60 contributors doesn't have a single screenshot. Nothing personal or particular about this project specifically, just... the whole open source culture of dropping something on Github and not even doing the bare minimum to have other people get to know the project.
It feels like such a waste. It could be an amazing project but who is going to bother with it if they can't see what it looks like?
I recently launched Early Access for my AI-powered service for automated bug reporting, PR review, fearure implementation, a difficult test/documentation/refactor suggestion PRs.
It was in the works for about a year, and I'm now trying to find ways to make it more marketable and useful.
The three main things I'm working on are:
1. GitLab support (GitHub only at the moment)
2. A demo on the landing page that doesn't require any sign-up.
3. Better "vibe coding" experience through the Chat interface for those who want it.
I built it with TypeScript on the front- and back-end, React, Node.js, and PostgreSQL. And I over-engineeres it, as one does, with a Redis cache and websockets to push the latest data to web clients so the latest info is always shown without needing to refresh. I'm using the OpenAI API right now, but I want to switch to local models when I can invest in the hardware for it.
I'm also having a difficult time getting the first users for my AI software development service. I decided to start with a high-touch Early Access model to understand the users and their needs better, and I'm up to 8 users so far.
For Early Access, I started with a LinkedIn post, reaching out to close colleagues, commenting in relevant Slack groups/channels, and introducing it to some of my better freelance clients. I'll be signing up 2 more of the freelance clients shortly, with one key feature request from each first. It's important not to get stuck in the trap of developing one-off features to get each new user, but these are features that would be broadly beneficial.
Soon, I'm planning to add a super-simple demo at the top of the landing page to increase conversions, and then I'll start promoting it more. Likely also some relevant informational content marketing on a blog.
This is not very concrete, but I've been thinking about the parallels between software development and Factorio and how to tie the two together to be (personally) beneficial. This includes finding ways to make software development as enjoyable as the game* and also pondering ways to make programming more visual. It would be cool to "see" a software project functioning similar to the moving parts of a factory.
* setting small, clear objectives
* spending enough time on refactoring
* giving myself a small reward for completing objectives
I personally recommend joining a local Toastmasters club. They can help you with speed, tone, volume, etc. It can benefit your conversational speech as well, not only public speaking.
Or a local speech course. I took an eight-week course in public speaking in college, and it was probably the most useful course I took there. I can do talks in front of tech groups or hobby groups... not without some sweating, but still more effectively than most people.
I used pixelmatch + puppeteer at a previous job to automate this. The service would screenshot each page on the master branch and the PR branch, then comment on the PR with a link showing the Before, Diff, and After of each page that changed. I highly recommend this type of setup.
Use whatever language works well for you and the task at hand, but many enjoy fullstack JS/TS.