AI Research2026

FormVerse

AI research platform that turns conversational input into structured, source-backed reports. Built on the Mastra agent framework with GPT-4 Turbo, Serper web search, and an explicit agentic loop.

Built by Abdullah Nadeem, AI Engineer

View Live Project
FormVerse interface screenshot

Overview

FormVerse is an AI research platform built with Next.js 15, TypeScript, Tailwind CSS, and the Mastra agent framework. It turns conversational user input into structured, source-backed research reports.

The system is built around an explicit state machine with two specialized Mastra agents: one generates adaptive forms, the other orchestrates research.

What I Built

  • GPT-4 Turbo conversational interviewer that dynamically generates adaptive forms with conditional field logic
  • Multi-tool research agent running an explicit agentic loop (Plan, Execute, Reflect, Terminate) across Serper web search and IP-based geolocation for regional context
  • Persistent agent memory via Upstash Redis, client state with Zustand, and animations with Framer Motion
  • Rich report rendering with react-markdown, GFM, KaTeX, and syntax highlighting, plus PDF export via jsPDF
  • Session management with chat history, safety guardrails against inappropriate prompts, and production-ready error handling

Tech Stack

  • Next.js 15
  • TypeScript
  • Mastra
  • GPT-4 Turbo
  • Upstash Redis