After a tough loss with our previous quail farm, due to Gumboro disease (a highly contagious viral infection), my parents and younger brother are back at it.

The R&D Project: 500 Sonali chickens.

While I sit at my desk, they are doing the real heavy lifting.

The physical reality:

• Smelling the air to detect ammonia levels 👃

• Manually checking rice husk humidity

• Reading chick "clustering" patterns to see if they’re cold

• Inspecting poop for gut health

• Mixing medicine and warm water manually for each refill

• Treating rice husk bedding for hygiene maintenance

I can’t help with the physical labor from a distance.

But as a Software Engineer, I wanted to remove one worry from their workload: Environmental monitoring.

The Constraint: $0 Budget

I immediately went to spin up a project on Google Cloud Platform (GCP). I hit a wall instantly: a mandatory $50 prepayment just to verify the account.

For a fragile family business in recovery mode? That was a no-go.

I needed an enterprise-grade logic on a hobbyist budget. I needed to build a Full Data Lifecycle (Ingestion → Storage → Visualization) for 0 BDT ($0).

The Architecture: The "Broke Engineer" Stack

I mapped the standard Data Engineering Lifecycle to free-tier tools:

1. Generation: Python Scripts (Mocking DHT22/MQ135 sensors).

2. Transport: HiveMQ (Free MQTT Broker).

3. Ingestion: Render (Hosting the Python subscriber script).

4. Storage: Supabase (PostgreSQL for time-series data).

5. Serving: Streamlit (Data visualization).

Step 1: Data Generation (The Sensors)

Since the physical sensors (DHT22 and MQ135) are currently shipping, I wrote a Python script to simulate the edge devices.

{
    'device_id': 'esp32-mock-01', 
    'timestamp': 1771171288.1688237, 
    'temperature': 28.66, 
    'humidity': 62.92, 
    'ammonia': 1.09, 
    'status': 'active'
}

Step 2: The Storage Layer (Supabase)

I chose Supabase because it gives me a full PostgreSQL database on the free tier. I created a simple table to store time-series data:

create table if not exists brooder_telemetry (
  id bigint generated always as identity primary key,
  device_id text not null,
  timestamp timestamptz not null default now(),
  temperature float,
  humidity float,
  ammonia_ppm float,

  -- Store raw JSON payload for debugging or schema evolution transparency
  raw_data jsonb,

  -- Metadata for auditing
  created_at timestamptz default now()
);

Step 3: The Challenge (Render & The "Sleep" Problem)

The biggest engineering hurdle wasn't the code—it was the infrastructure limitations.

I deployed my MQTT subscriber script to Render.com. It worked perfectly... for 15 minutes.
Render's free tier spins down web services if it detects no inbound HTTP traffic for 15 minutes. Since my script was a background worker listening to MQTT (not HTTP), Render thought it was idle and killed it.

(Docs: https://render.com/docs/free#spinning-down-on-idle )

The Zero-Cost Solution: Cron-job.org
I didn't want to upgrade to a paid plan yet. So, I added a dummy HTTP endpoint to my Python script and set up Cron-job.org to ping that endpoint every 10 minutes.

It tricks the server into staying awake—keeping the pipeline alive 24/7 without a credit card.

Step 4: Visualization (Streamlit)

For the interface, I used Streamlit. It connects directly to Supabase via SQL and renders the charts. It’s lightweight, Python-native, and free to host.

🚀 The Result & Roadmap

We now have a live V0 dashboard monitoring the simulated environment.

Project Roadmap:

• V1 (Automation): The next step is connecting the relay switches to control exhaust fans based on the temperature data.

• V2 (Edge AI): I plan to use OpenCV to analyze the "clustering patterns" of the chicks. If they huddle together, it means they are cold—a visual data point that sensors might miss. For now, it’s a simulation.

Conclusion

This isn't Smart Farming yet. My code doesn't mix feed or change the bedding. But it’s a solid V0, and I'm hoping my code becomes half as resilient as my parents.

But you can see the live V0 dashboard here:

👉 https://agri-stack-dashboard.streamlit.app/

Time to move from simulated data to real Edge IoT. 🚀

#IoT #AgroTech #SoftwareEngineering #BuildInPublic #CloudComputing #EdgeComputing #Supabase #Streamlit