sauced.ai

Inspiration

As students who try to meet certain macro goals, we find it hard to meal prep when most of the recipes out there don't include nutritional info, require tools we don't have in our limited kitchen setups, or exceed our budget. Even once we find a recipe that suits our needs, the process of scaling the ingredient quantities for larger batches and then hunting down those ingredients in the grocery store is unnecessarily tedious. These barriers make it harder and harder to choose meeting our goals over eating out. We needed a tool that does the heavy lifting to make the healthy choices easier.

What it does

sauced.ai is your personal meal prep assistant. Just paste it the link to a Substack recipe newsletter and tell it about your nutritional goals, taste preferences, budget, or anything else related to the meal prep process. You can talk to it like a human instead of worrying about hitting certain keywords like with search engines. sauced.ai with then scrape the newsletter for recipes, generate nutritional info for each of them, and then rank them based on your criteria. In the final step, it'll scale the top selected recipes based on your specified meal prep batch size and create a grocery list. You even get the option to add the entire list to your Instacart cart with just the click of a button. With the help of sauced.ai, all that's left for you to do is prepare the meals and enjoy your healthy choices.

How we built it

We implemented a skeletal large scope frontend and backend to handle features and setup API calls to Browser Use and Gemini. We successfully configured the Gemini API to receive calls and perform baseline logic actions related to sauce.ai's goal.

Challenges we ran into

We experienced a bit of a learning curve building our agents since we've never used the Browser Use or Gemini API before but were ultimately able to learn a lot. We initially also wanted to use the fetch.ai agent framework but later had to scrap that goal because it was a lot to take on in addition to Browser Use and Gemini. We also faced some constraints when it came to our credits that stalled development and testing at the end of the first day.

Accomplishments that we're proud of and what we learned

We're proud of our sleek and untuitive UI as well as our lighthearted project name. But most of all, we're happy that we were able to implement all the core features we initially sought out to create. Along the way, we learned that the best tool is one that's simple and that doesn't do or solve more than it needs to. Meal prepping should be straightforward, and so is sauced.ai.

What's next for sauced.ai

Given more time and resources, we'd love to expand sauced.ai's scraping scope from individual, manually selected newsletters to all Substack newsletters. It would monitor Substack constantly and on its own to constantly feed users new recipes they wouldn't otherwise encounter. That way, a bigger part of the recipe search process is taken care of. We'd also love to further refine our agent prompting protocol to suit users' needs through extensive testing.

Built With

• browser-use
• css
• gemini
• html
• javascript
• python