LLensr

System Documentation

Architecture &
System Design

A comprehensive guide to Lensr's production infrastructure — from Azure Container Apps and AWS Bedrock to semantic caching and multi-agent orchestration.

Lensr Architecture

Lensr is a modern, AI-powered search engine built to provide high-quality, structured answers instead of just a list of links. The architecture is designed to be fast, scalable, and modular.

🔄 End-to-End Flow

  1. User Request: You enter a search query (or upload an image) in the browser.
  2. Frontend App (TanStack Start): The web app securely passes your request to the backend using Server-Sent Events (SSE) to stream the response back in real-time.
  3. Backend Agent (Python + LangGraph):
    • Router: A fast LLM (Claude 4.5 Haiku) classifies the intent of your search (e.g. Shopping, Trip Planning, General).
    • Cache Check: The system instantly checks if a similar question was recently answered using Semantic Cache (pgvector).
    • Search & Scrape: The backend searches the web (Google SERP) and scrapes the actual content of the top results.
    • Synthesizer: A powerful reasoning LLM (Claude 4.6 Sonnet) reads the scraped content and synthesizes a final, highly accurate markdown answer.
  4. Display: The frontend streams the text live to your screen, along with images, source citations, and interactive UI widgets.

🛠️ Tools & Technologies

  • Frontend Framework: TanStack Start & React 19
  • Styling & UI: Tailwind CSS v4, shadcn/ui, Framer Motion
  • Backend Framework: Python FastAPI & LangGraph (Agentic workflows)
  • AI Models:
    • Reasoning & Vision: Claude 4.6 Sonnet (via AWS Bedrock)
    • Fast Routing: Claude 4.5 Haiku (via AWS Bedrock)
  • Web Search: Google SERP API (Serper.dev)
  • Authentication: Better-Auth (Google OAuth & Email)

☁️ Azure Infrastructure

The entire application runs on Microsoft Azure, provisioned automatically via Bicep Infrastructure-as-Code.

  • Azure Container Apps: Hosts both the Frontend and Backend in a serverless, auto-scaling environment.
  • Azure PostgreSQL Flexible Server: Stores user accounts, saved searches, and powers the Semantic Cache using the pgvector extension.
  • Azure Key Vault: Securely stores all sensitive API keys and database credentials.
  • Azure Blob Storage: Stores user-uploaded images for visual searches.
  • Azure Log Analytics: Aggregates logs and performance metrics across the system.

Last updated: September 2026

Back to search