Introducing ReMarkšŸ’¬ ā€“ Robocorp's Generative AI Assistant for Automation Developers

Robocorp has always been at the forefront of advancing software-defined automation, and our latest innovation continues that tradition. We are thrilled to introduce ReMarkšŸ’¬, a chatbot assistant designed specifically for automation developers.

June 12, 2023 ā€“ Tommi Holmgren

Harnessing the power of large language models with code-generation, ReMarkšŸ’¬ helps developers leverage the full potential of the Robocorp platform, making development faster and more efficient. It's available starting today!

Battling the Blank Canvas Syndrome

Software development often begins with a blank canvas ā€“ an intimidating sight for even the most experienced developers. That's where ReMarkšŸ’¬ steps in. However, the features are not limited to getting you started with the development. Chat interface is great at iterating solutions, spotting errors and helping developers to discover the best keywords to use from the libraries spanning thousands of options.

ReMarkšŸ’¬ is an excellent example of how code-generation is revolutionizing the automation landscape that previously was mostly filled with vendor specific low-code tools. By generating robust, functional code snippets, it serves as an invaluable companion in your automation journey - for beginners and seasoned developers alike.

Automation Development with a Twist šŸ

Unlike generic chat assistants like ChatGPT, ReMarkšŸ’¬ has a unique focus - automation development on the Robocorp platform. It's not just a chatbot, it's like a knowledgeable senior developer, familiar with Robocorp's tooling and automation libraries. It's been trained on thousands lines of real automation examples and documentation, ensuring that ReMarkšŸ’¬ provides relevant assistance to developers in their day-to-day tasks.

Being Python-based, Robocorp provides a significant advantage when it comes to code generation. Unlike proprietary RPA tools that operate within a closed system, our platform opens the gate to the vast universe of Python automation libraries.

For example, demonstrating how to use Playwright for high-performance browser automation and rpaframework for connecting with Java-based ERP apps, ReMarkšŸ’¬ makes it super easy to get started with automation projects.

ReMark answer example for Java Access Bridge and Oracle EBS.

The Future Will Bring Integrations

Starting today ReMarkšŸ’¬ is available as a standalone chat. This is just the beginning, as our plan is to seamlessly integrate it into our developer tools such as VS Code. In our vibrant Slack community of 15k automation developers, we have already been experimenting chat based support with a company called Kapa.ai. Itā€™s been a great success with hundreds of thumbs-up answers in just a short time, validating the technology is ready for the prime time. We plan to take this even further.

When it comes to ReMarkšŸ’¬ā€™s ā€œknowledgeā€, there are several avenues to take. We foresee it becoming a great help with broader solutions, integrations or recommending the best 3rd party Python libraries from the open-source community.

Learning Curve

Like any assistant ā€“ robotic or human ā€“ ReMarkšŸ’¬ may not always have the perfect answer. The days of code generation and chat assistants are still early, and we're committed to continuous iteration to achieve the highest standard of results. We encourage you to be critical of whatever a chat tells you and share your experiences with us!

The Tech Behind ReMarkšŸ’¬

ReMarkšŸ’¬ is based on a two-step approach loosely modeled after LangChainā€™s ā€œChat Over Documents with Chat Historyā€ example. We leverage a locally hosted vector database (Chroma) and Open AI API for LLMs.

The first step involves searching for the relevant information needed to answer a question from various stores and currently uses gpt-3.5-turbo. The second step uses that information to generate a comprehensive response and uses gpt-4.

ReMark's LLM architecture overview

This architecture also enabled a fancy but handy feature: generating the ā€œnext best questionsā€ for you. At the end of ReMarkšŸ’¬ā€™s replies, you can always see questions that might be the most relevant to ask next. Saves the trouble of writing!

ā€”

ReMarkšŸ’¬ is a groundbreaking step in our ongoing mission to make automation development as efficient and user-friendly as possible - helping our customers achieve the highest possible ROI.

Happy automation coding together with ReMarkšŸ’¬!

Share this article