Thanks for checking it out and the record mode is a great idea. I've been playing around with ways to get the terminal outputs but so far I haven't loved the UX of my solutions. Your co-pilot approach that can explain the commands and iterate is really valuable.
If you're open to joining, I have a small AI engineer/ open source dev Slack community in Boston. Id love to have you (https://smaht.ai)
I really just wanted the feeling of tab-based auto-complete to just work in the terminal.
It turns out that getting the LLM responses to 'play nice' with the expected format for bash_completion was a bit of a challenge, but once that worked, I could wrap all the LLMS (OpenAI, grok, Claude, local ones like Ollama)
I also put some additional info in the context window to make it smarter: a password-sanitized recent history, which environmental variables are set, and data from `--help` of relevant commands.
I've just started to promote it around the Boston area and people seem to enjoy it.
I hate trying to build something quickly and forgetting the syntax, "How do I resize a video in ffmpeg?" or "create an postgres docker container with all right environment variables and ports set".
So I built autocomplete.sh. It supports OpenAI, Grok, Claude and Local models like Ollama. Just install, select the model, enter API key and then when you type <tab><tab> it will return 3-5 suggestions on ways to do the task.
It puts into the context window things like recent command history, which environment variables are set (not the values!) and what type of shell it is so that the results are tailored. I do also try to remove things that look like passwords to keep the prompt free of PII. I also like that you can run locally using ollama or run things extremely fast (and free!) using Grok.
The script is GPT wrapper written entirely in bash and so it can be extremely portable.
It was fun to build and I recently showed it around the Boston Startup scene and people liked it. So I wanted to share it here before the weekend!
Professional Investors cannot run their strategies on auto-pilot; we provide forecasts, alerts and analysis to identify when models start-misbehaving. We put experts back into the driver-seat at the key moments to improve investment returns, reduce losses and reduce the time to insight.
As a former portfolio manager who also runs a "machine learning for economists company", I have seen many examples where AI and better data can help economists quantify their intuitions and more rigorously test their models. My product is even called the "AI Economist"
Let's face it, humans are great at organizing problems and hypothesing about the future, but we are terrible at quickly reasoning about complex interactions and finding patterns in highly noisy data.
One area I'm pretty excited about here is using better machine learning algorithms to organize global production, employment and trade data to improve out definitions of business cycles. We have relied on 'high-level stylized facts' to describe the complex and fascinating global economy for too long.
Now that I have some Mac iOS dev work to do I'll probably build and test it