Automation - parismolver.com

Automated Pi Audit and Report

pmolver — Thu, 22 Jan 2026 09:20:32 +0000

Raspberry Pi Auditing and Report Generation

The Problem

Managing a network of remote devices individually is a massive bottleneck. Logging into every Raspberry Pi manually to run updates or check statuses is time-consuming, prone to human error, and makes it nearly impossible to get a “big picture” view of the network’s health in real-time. Additionally, it becomes hard to double check someones work if mistakes happen.

Considerations

– Software maintainability and modularity. I needed something that was easy to adjust and modify and allow for future expansion as the software could be useful outside of single script executions.

– Proposed actions, expected responses, and reporting. I needed to be able to propose a list of command executions, discuss with others what the expected results and outcomes were, and then report back to them when the script completed or if issues arose.

– Ease of access outside of code, as devices could be taken offline or new ones commissioned, I needed a database external to the software that could me maintained and modified by people with no knowledge of how the software itself worked.

The Proposed Solution

I built a centralized Python script runner that had a couple jobs.

Access google sheets to gather device information
Connect to remote devices using VPN and SSH
Execute a script module
Select a single target or set of targets
Package up the results of the script module into an HTML format

Each script module was designed for one task, such as checking the python version, or executing a task with expected response. Each was designed as a single device execution and was capable of retrying. This is so that the execute module could perform it’s task and create the needed data for the report with minimal overhead.

In Action

When used, this eliminated almost all user error from the process. The reports allowed for me to flag issues and reassure the lead engineers of the state of the network. Additionally, whilst running this script I could essentially be doing two work loads at once as I could perform other tasks whilst I waited for the code to execute.

The post Automated Pi Audit and Report first appeared on parismolver.com.

Web Agents

pmolver — Thu, 22 Jan 2026 09:13:23 +0000

Web Agent Automation

The Problem

Working at OnFarm Data, at one point we had a new web service with some a minority of users on the service. We also had one full time employee doing both development and maintenance of this service.

Resources were stretched thin. So when I needed to perform a task such at sending commands to all the devices on a network, or capturing the status of all devices on the network, I was limited to the tools available on the web platform.

Thinking outside of ‘the box’

Using Selenium, a web driver, I was able to use python and google sheets to create either a list of commands, or a list of information that needed to be extracted. This was a way of finding a solution to a problem without creating additional work for the sole developer.

The software was loaded with a list of URLs, and at the destination website, the web driver could extract information about the device. One of the activities was a simple auditor, it would send a message and wait for a web element to change, which was the result of the physical device reacting to the command.

All of this was recorded and repeatable with minimal time required outside of developing the web agent.

Saving time, and saving face.

Working in such a small company means you don’t have the same resources available. Whilst we all know what something should look like, or be capable of doing, this isn’t always feasible.

Using my web agents, I was able to offer services to the customers I was dealing with, without having to create work load for the product engineers, or web developers.

The best example of saving time and saving face, was when a customer needed to switch their irrigation schedule to align with cheaper power. Traditionally this would be done by sending somebody on site. Another way is manually send commands over the network, confirm changes had occurred, and deal with devices which were harder to communicate with. With my scripts I could create an action table, run the software, and resume my other tasks.

The post Web Agents first appeared on parismolver.com.

Embodied Carbon Analysis

pmolver — Thu, 22 Jan 2026 07:40:08 +0000

Embodied Carbon Matrix

The Problem

During my internship at Energyline (located Woolston Christchurch), the marketing team required information about the embodied carbon in the products, this is required for architects trying to perform lifecycle analysis on their buildings.

Typically this is a pretty straightforward process, you consider the materials used in your product, you use the tables and methods provided by the CIBSE TM65 to translate to embodied carbon, and you have your figure. What made this tricky is that Energyline makes lights of different profiles, with different dimensions and features, so there is no one size fits all figure.

Creating the matrix

Using Excel, I was able to create a table that would capture the embodied carbon of their smallest possible light, and then apply a linear increase of carbon as the light length increases. At certain points, another LED panel is required, at which point, a new linear segment would begin. This repeats and continues as additional components are required such as additional power supplies.

Some components such as end caps, power cables, and mounting hardware could actually be calculated separately, so for this case, the embodied carbon of the total product becomes the customisable product in addition to these optional parts.

To complete the carbon analysis, I would have to duplicate this table and adjust the parameters for the different profiles. Using this method, I was able to solve for the embodied carbon values of all the possible products.

The Results

Working through this carbon analysis found that the use of recycled aluminum smelted here in NZ by McKechnie made a considerable impact. Because aluminium is typically a material with a lot of embodied carbon, having a local smelter using recycled aluminum meant that the Energyline embodied carbon result was significantly lower than products made with virgin material or made in countries without clean power. Additionally, the long life span of these products meant that Architects using this product in their designs would be able to get lower carbon figures when accounting for the building life span.

This project could be taken a step further as the carbon figures could be quickly generated with a few inputs if needed. This would be good for a dynamic website where a user could be adjusting lighting length with a slider to find the values of lowest carbon per light output.

The post Embodied Carbon Analysis first appeared on parismolver.com.

Job Search Software

pmolver — Thu, 22 Jan 2026 01:49:04 +0000

Job Search and Application tool

The Problem

Looking for jobs can be tedious. Unless you work under well defined terms such as ‘mechanic’, you often have to perform multiple searches. Additionally, there are new jobs posted every day so having to perform a series of searches, saving the interesting ones, and creating a custom cover letter is a frequent task. Some of the issues I had to deal with are

– Performing searches for multiple key words and doing so frequently.

– Dealing with multiple user interfaces, bouncing between searching, cover letter creation, and back to the application page to post.

– Creating bespoke cover letters based on the application when I already had several ‘close’ but not exact cover letters stored from previous submissions.

Specifications

As somebody who likes to automate even the most menial tasks, I thought creating a job search software would be a great candidate. The following are the requirements I needed to meet to create a software that suited my needs;

-Gathering current listings off of a website using a web driver such as Selenium which was capable of passing cloud flare bot detection and running in headless mode.

– Storing potential applications in a database including details about the URL, and current status of the job.

– Preventing duplicates, this also helped to know when there were no new listings so the search could end prematurely.

– Separate tabs for sorting new listings, listings to be reviewed, listing that have been actioned, and settings.

– A cover letter generation wizard which acts in two parts, a strategist helps form a set of details, and a generator combines knowledge base, template, and listing to create a final cover letter.

– Google documents integration so that the generated cover letter can be edited, saved, and exported as PDF.

– A tracking tab to help identify trends and motivate me.

The Result

The result is a software that helps me keep track of new listings, and helps ease the workload required to custom craft a new application for each job listing. Most valuable however is the better user experience offered by providing a single user interface with a linear step through process instead of the typical job search process which can leave you bouncing between pages.

Additionally, this is a good example of Python integration with the web, where we have google docs being used as a cover letter destination, and Gemini being used twice to form the strategy and the final output. As AI becomes more competent, I can change my models to find either better value in terms of API offerings or output quality.

MVP Potential.

As a potential product, this software could be used as an MVP to test if others are looking for a similar service.

The post Job Search Software first appeared on parismolver.com.