Skip to content
Next.js · TypeScript · FastAPI · MongoDB Atlas · Vector Search

Insightron

All your numbers, one cockpit

The Problem

A founder's numbers live in ten dashboards — analytics, search console, ads, payments, email, CRM. Checking them all takes hours a week, and the insight that matters is buried in the tab you didn't open.

My Approach

Built an intelligence platform where 42 scheduled jobs snapshot every source into one store, daily narratives get embedded for retrieval, and an AI copilot with 72 tools answers questions against live data over streaming responses. The CRM populates itself from visitor and payment signals instead of manual entry.

Technical Challenges
01

Normalizing 10+ third-party APIs with different auth models, rate limits, and data shapes into one coherent store

02

Retrieval over daily intelligence narratives — embedding and vector search so the copilot cites what actually happened, not what it guesses

03

Streaming tool-use UI — 72 tools calling live APIs while responses render token by token

04

Cost discipline at solo-founder scale — TTL caches, snapshot retention windows, and per-endpoint budgets

Results
22
Live Tabs
72
Copilot Tools
42
Scheduled Jobs
213
Tests
Tech Stack
Next.jsTypeScriptFastAPIMongoDB AtlasVector Search

Like what you see? I build websites and AI integrations through my agency.

Book a Free Consultation →