Raunak Chowdhuri raunakdoesdev

My Home Assistant setup: a Mac mini, two radios, and automations that behave like household software

Last updated July 2026. This write-up was assembled with help from OpenAI Codex from my live Home Assistant device registry and configuration, then reviewed for public sharing. Names, addresses, entity IDs, network details, tokens, and private endpoints are intentionally omitted.

I run Home Assistant because I want the house to respond to context, not just expose a larger collection of remote controls. The interesting part is less "I can turn a bulb on from my phone" and more "the house understands who is home, which alarm matters, whether I just arrived in the Tesla, and whether opening the shower door means entering or exiting."

This is the current architecture, device inventory, automation layer, and deployment workflow.

Runtime architecture

Background agents: autonomous app builder

This is an open-ended systems design and implementation challenge. There's no single correct architecture, and we don't expect you to finish every milestone in one day. We care about how you approach it: how you evaluate tools, how you use AI agents throughout, and the tradeoffs you make. We'll ask you to walk us through your thinking, so optimize for understanding and being able to explain your decisions, not just output.

You're encouraged to use AI coding agents (Claude Code, Cursor, Copilot, whatever you prefer) throughout this exercise. This isn't a test of whether you can write code from memory. We'll evaluate you on your understanding of the system you build and the decisions you made. Keep traces/logs of your agent interactions, we'll review them during the debrief.

	/**
	* Simplified belief verification using a single Opus agent with Exa search tools.
	*
	* Replaces the 8-agent adversarial pipeline with one agent that can:
	* - Search for supporting AND contradicting evidence via Exa
	* - Distinguish first-person vs third-party sources
	* - Render per-field verdicts with citations
	*
	* Usage:
	* EXA_API_KEY=... ANTHROPIC_API_KEY=... npx tsx belief-verifier.ts

	import anthropic

	client = anthropic.Anthropic()

	message = client.messages.create(
	model="claude-3-5-sonnet-20241022",
	max_tokens=1000,
	temperature=0,
	system="Always respond in JSON.",
	messages=[

	{
	"usage": {
	"num_pages": 63,
	"num_fields": 1098
	},
	"result": [
	{
	"companyName": "MetLife",
	"effectiveDate": null,
	"benefitPlans": [

	import requests
	import os.path
	from pdf2image.pdf2image import convert_from_path
	from PIL import ImageDraw


	pdf_path = os.path.expanduser("~/Downloads/research-highlight.pdf")
	upload_form = requests.post("https://v1.api.reducto.ai/upload").json()
	requests.put(upload_form["presigned_url"], data=open(pdf_path, "rb"))
	response = requests.post(

	{
	"$ref": "#/definitions/extension",
	"definitions": {
	"extension": {
	"type": "object",
	"properties": {
	"$schema": {
	"type": "string"
	},
	"name": {

	# Setup OpenAI GPT API

	import openai
	import textwrap

	# Set the GPT-3 API key
	openai.api_key = "YOUR_API_KEY"

	def get_gpt_response(prompt, max_tokens=2048, temperature=0.7, n=1):
	response = openai.Completion.create(

	<?xml version="1.0" encoding="UTF-8" standalone="yes"?>
	<OfficeApp xmlns="http://schemas.microsoft.com/office/appforoffice/1.1"
	xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
	xmlns:bt="http://schemas.microsoft.com/office/officeappbasictypes/1.0"
	xsi:type="ContentApp">
	<Id>03648872-fa1f-4c2d-b9aa-29ce4d0a2bbc</Id>
	<Version>1.0.0.0</Version>
	<ProviderName>Contoso</ProviderName>
	<DefaultLocale>en-US</DefaultLocale>
	<DisplayName DefaultValue="Oloren PPT Visualize"/>

Raunak Chowdhuri raunakdoesdev

My Home Assistant setup: a Mac mini, two radios, and automations that behave like household software

Runtime architecture

Background agents: autonomous app builder

The challenge