TechEditor 5: How Artificial Intelligence Helps Engineers with Technical Documentation

Artificial intelligence has burst into our lives, promising a revolution in every field. But for an engineer, where the cost of an error can be critical, a crucial question arises: can we trust calculations performed by ChatGPT, Gemini, or Claude? At Dystlab, we've wrestled with this question for a long time.

On one hand, ignoring the power of LLMs means falling behind. On the other, blindly trusting them with critical tasks is unacceptable. That’s why in TechEditor 5, we’ve implemented a unique approach that lets you get the best of both worlds: the speed of AI-driven idea generation and the reliability of a proven engineering engine. This isn't just another feature; it's a whole new way of working with technical documentation. Let's dive into how it works.

The TechEditor environment is designed for developing text documents and engineering calculations. You can create engineering reports and other documents from scratch in all versions, but starting with version 5.0, this process can be significantly accelerated. How? Through integration with artificial intelligence services. It's quite simple: you write a prompt, choose a language model, and get automatically generated data right in your report.

Seems too simple? Let's find out.

How AI Integration Works in TechEditor: The Core Principle

It's important to state right away that TechEditor has no native neural networks of its own, so the program itself does not generate anything. It still operates on your direct commands, much like AutoCAD or Excel. You can see examples of technical documentation created in TechEditor in our catalog.

So where does the AI magic come from? It all works through an API (Application Programming Interface). In simple terms, TechEditor acts as a reliable intermediary: it takes your prompt, securely sends it to the servers of your chosen Large Language Model (LLM), receives the response, and inserts it into your document.

Currently, TechEditor supports integration with eight of the world's most renowned brands:

• OpenAI (ChatGPT)
• Gemini
• Grok
• Claude
• Perplexity
• Ollama
• Mistral
• DeepSeek

This list will likely expand over time.

Step-by-Step Guide: How to Start Generating Reports with AI

1. Get an API Key from a Neural Network Provider

All the listed brands offer access to their services via API, so your first step is to register with at least one of them and generate an API key. Here is what the key management page on Claude.ai looks like:

https://console.anthropic.com/settings/keys

Save the generated API key in a secure place — for example, in a text file or spreadsheet.

ATTENTION! For security reasons, you can only copy the key once, at the moment of creation. If you lose it, you will need to generate a new one.

As an example, the DeepSeek service explicitly warns its users about this:

2. Connect to the Neural Network in TechEditor

Once you have your key, you can connect to the LLM. Go to the AI tab and click "Connect to LLM...":

In the LLM list, choose your provider. Paste the access key into the "API Key" field and click "Connect to LLM | Get models." Once TechEditor successfully connects, a list of available models will appear in the "Model" dropdown. Each brand offers its own set of unique builds and versions; we'll cover their specifics in future blog posts. Click OK.

3. Start Working with AI in Your Documents

Now, you can enter a query into the "Prompt" field. To run it, click "Execute Prompt":

The AI's Response: Where the Magic Begins

As an example, we asked the AI "how to calculate the area of a rectangle." It's a basic prompt, but it clearly illustrates the workflow between the LLM and TechEditor.

After a few seconds, the AI provided a response that was immediately placed into the report. And this is where things get interesting.

Full content control

Once you have AI-generated content in your report, you can work with it just like any other text: edit, copy, move, delete, and format it. Essentially, it's just regular text — do with it what you see fit. TechEditor has an advanced word processor and a wide range of formatting tools to help you.

Static Formulas in LaTeX

Formulas that are purely explanatory are inserted into the report in LaTeX format:

But this is not a static image! Every formula can be modified in the TechEditor formula editor. You can also change its font, size, and other attributes as needed.

A New Paradigm: Generate with AI, Verify with TechEditor

From a scientific and engineering perspective, the most interesting part of AI-generated content is the math objects.

If the LLM's response includes calculations, they are inserted into the report as Math Objects — special mathematical entities in TechEditor:

The language model automatically fills in the "Input" or "Output" section as needed. In our example, the inputs are the length and width of the rectangle. The LLM created two math objects with variables "a" and "b" and assigned them the appropriate values. The final object simply calculates the area, so its "Input" field is empty, while the "Output" section is filled.

Notice that the neural network handled not only the numbers but also the units of measurement, all in a format that TechEditor understands perfectly.

But here’s the most important part: the final calculation is performed by TechEditor, not the LLM. Why? Because AI models make mistakes.

Why are calculations in TechEditor safer than with ChatGPT alone?

Neural network providers are transparent about the fact that their models can be wrong. This isn't surprising, given their probabilistic nature, which means the same prompt can yield different — and sometimes incorrect — answers.

This is particularly true for calculations and handling units of measurement. These issues often complicate, and can even make impossible, the reliable use of AI in engineering tasks.

So what's the solution?

To solve this, Dystlab developed an approach where all mathematical calculations are performed on the TechEditor side, not within the neural network. Here's how it works:

1. AI Generates the Draft: You give a prompt to the neural network (e.g., "calculate the area of a I-beam"). The LLM generates the response structure and formulas as a set of instructions.
2. TechEditor Validates and Computes: Our native physics and mathematics engine, MathSIV, then takes these instructions and performs all the actual mathematical operations itself.

What does this mean for you? If ChatGPT makes a mistake with units or suggests an invalid operation (like adding meters to kilograms), TechEditor will catch the error, stop the calculation, and notify you. In this workflow, the AI acts as a fast assistant, while TechEditor serves as the expert engineer who verifies every detail.

Other Potential Issues

Interacting with AI won't always be seamless. Other potential issues include:

• Network connection errors
• Insufficient funds or token limits on your AI service account
• The general availability of the LLM
• Incorrect formatting in the LLM's response

In such cases, you will typically receive an error message. Here is an example of a notification for an exceeded quota (HTTP error code 429):

The format of these messages and the overall interface for working with LLMs will be improved in future versions of TechEditor.

What's Next?

We have just crossed a certain "event horizon", marking the arrival of the fifth generation of TechEditor. With it, we've begun open testing of our AI features with our wider user community. We are actively gathering feedback, conducting research, and forming a list of improvements.

In the near future, we plan to implement:

• The ability to include files in prompts;
• AI that can work with images;
• LLM support in other Dystlab digital products, and more.

Give It a Try and Join the Discussion!

Integrating AI into engineering is new and exciting territory that we are exploring together with you. The best way to appreciate the benefits is to try it for yourself. Update your TechEditor to version 5.0 and start experimenting!

We invite you to share your impressions, your best prompts, and real-life case studies in our Discord server, "TechEditor & Engineering Solutions." That’s where our community connects and where user support happens.

• How are you already using AI in your work?
• What challenges or successes have you encountered?
• Let's share our best prompts for engineering tasks!

Join the discussion on Discord

Vitalii Artomov

"I am working to make «Made in Ukraine» a global symbol of quality and style"

CEO, co-founder of Dystlab, developer of TechEditor. Engineer, scientist, Ph.D. with over 20 years of experience in structural analysis and automation of engineering calculations. I advise engineering companies in Ukraine, Europe, and North America.

Discuss business solutions: This email address is being protected from spambots. You need JavaScript enabled to view it. | +380504576819 (WhatsApp)