Foretaste

Inspiration

Small restaurants often make menu decisions based on intuition. We wanted to help owners combine sales, costs, customer feedback, and market trends to make faster, data-driven decisions.

What it does

Foretaste is an AI-powered menu intelligence platform that helps restaurants decide what to launch, improve, reprice, or discontinue.

It combines menu performance analytics, historical sales, demand forecasting, customer feedback, and TikTok trends. Gemini-powered agents analyze these signals and turn them into actionable recommendations.

How we built it

We built the backend with Go, using PostgreSQL + pgvector for data and semantic search and Redis for caching. The frontend uses Next.js and React.

Gemini powers trend analysis, financial reasoning, recommendation synthesis, and embeddings for TikTok trend retrieval. Our backend handles deterministic calculations such as revenue, COGS, margins, and forecasting baselines.

Challenges we ran into

Our biggest challenge was combining very different data sources: structured sales data, unstructured reviews, and noisy social-media trends.

We also had to determine where AI should be used. We kept financial calculations deterministic and used Gemini for contextual analysis and reasoning, making recommendations more grounded and trustworthy.

Accomplishments that we're proud of

We built an end-to-end workflow from restaurant data to actionable menu recommendations.

We're especially proud of combining real business metrics with semantic trend retrieval and Gemini-powered analysis instead of building just another AI chatbot.

What we learned

We learned that useful AI depends on reliable data, retrieval, and system architecture—not just the LLM.

We also learned how to combine Go backend engineering, PostgreSQL, vector search, forecasting, and Gemini into one practical AI product.

What's next for Foretaste

We plan to expand POS and review integrations, improve forecasting, continuously refresh trend data, and track whether AI recommendations actually improve menu performance.

Our goal is to make Foretaste a continuous AI product manager for small restaurants.