At Softlandia, we focus on developing practical solutions that drive business and technological advances forward. Some of the core pieces of technology inside these solutions are open-source software. We understand the value of open-source software and its widespread impact on technology and society.

Our team members have contributed to open-source projects in AI, computer vision, data management, and others. The contributions enhance, for example, data handling, and security and simplify the integration of advanced technologies like large language models and computer vision into everyday applications. 

By contributing these developments to the open-source community, we foster innovation and offer businesses the tools to streamline operations and accelerate growth. This post outlines some key open-source projects our team has contributed to and their direct benefits, showcasing how they can support your business.

Democratizing LLMs

We are among the early adopters and contributors to the Llama-Index project. This tool makes it easy to ingest data in a format that Large Language Models can understand and allows users to connect LLMs to their data using various data storages (not only vector databases!) and algorithms. With Llama-index, you can feed PDF documents, websites, and code to language models easily, creating a personalized assistant in the blink of an eye.

Generative AI is one of the core pieces of applied AI engineering. As community experts, we’ve helped many to get started with their own LLM applications. Our contributions focused on integrating Llama-Index with the powerful Qdrant vector database, enabling fast RAG solutions.

Verifying LLM outputs

While LLMs are very powerful tools for understanding and creating texts, their nature is a tiny bit unpredictable. Guardrails AI is an open-source package that addresses this feature, allowing developers to build applications with verified outputs from LLMs. For example, Guardrails AI can make sure the output has a structure like a list or does not contain unwanted data such as names.

When the chat-based APIs to LLMs became mainstream, our team took on the job of updating Guardrails AI to support the new format with system and user messages. The contribution made it possible to instruct language models more accurately with Guardrails AI, making the tool even more powerful.

Guardrailing LLMs is important especially when it comes to developing public LLM-based solutions. Without proper guardrails, the LLMs can answer almost any question a user has asked them. This might put your company at risk, since the answers may cause harm to your business or brand. Therefore, you should use a properly guard-railed AI engine for your products and internal processes.

Making Computer Vision More Accessible

Our contributions extend to the popular computer vision library, OpenCV. Softlandia's founder Olli-Pekka is the original creator and maintainer of the popular pre-built OpenCV Python packages. He maintained the project for about 5 years before the maintainership was transferred to the OpenCV organization, making the packages official. These packages are one of the few that have been designated as critical in the Python package index.

Opencv-python packages enable developers to easily implement computer vision applications, from simple image processing tasks to advanced computer vision applications. Before opencv-python packages existed, OpenCV was very cumbersome to install, especially with its Python bindings.

Pre-built packages quickly became very popular, having hundreds of millions of annual downloads. Therefore the impact on various industries is immeasurably large: millions of developer hours are saved annually in installing, configuring, and building computer vision applications.

Countless private and open-source projects depend on the opencv-python package. This work not only boosts the accessibility of computer vision but also fuels innovative projects across various sectors, such as the space industry: the package was used by NASA to fly the Ingenuity Helicopter on Mars!

Enhancing Data Management and Search Capabilities

One of our latest contributions was to the Qdrant vector database project. We enhanced Qdrant Python client capabilities by adding a more flexible authentication feature that enables users to use external authentication token providers.

We wrote a FastAPI singleton client example to demonstrate the feature. Our detailed guide on this can be found in our blog post on deploying Qdrant with gRPC auth on Azure. 

This work supports businesses and developers in implementing more secure and straightforward semantic search and retrieval solutions by providing simple and flexible authentication integration. Less code to write, faster time-to-market!

Streamlining Real-Time Data Processing

We've also written about real-time data processing by evaluating various Python technologies and showing them in action. You can build a Slack bot with RAG capabilities as well!

By contributing to this field, we provide the community with insights and tools that improve the efficiency and responsiveness of real-time data systems. This evaluation aids developers in choosing the right tools for their data processing needs, fostering innovation in real-time applications. For more information, see our evaluation of real-time data processing technologies.

Developing Algorithms for Real-Time Graphics

We have extensive knowledge of algorithm development and signal processing. One of the open-source contributions is related to these fields.

One of our founders, Henkka, has implemented the performant OpenGL-based multi-channel signed distance-field (MSDF) algorithm. The Multi-Channel Signed Distance Field algorithm is an advanced method used in computer graphics to render sharp, high-quality text and vector shapes at various scales.

It is useful for developers aiming to render high-quality text and vector shapes in real-time applications, such as video games or interactive graphics software, where performance and visual quality are critical.

Reporting bugs and fixing issues

The previous contributions are hand-picked examples of our impact on the open-source communities. However, we do constantly also small less visible contributions. 

Often, we work on cutting-edge projects that use the latest versions of many open-source software libraries. In these cases, it is common that we encounter one or multiple bugs in some of the libraries. We report these bugs to the projects. If the bug is critical, we also fix it or at least take part in debugging it further.

This is a very low-effort and practical way of improving the libraries. If you plan to do the same, check that your bug reports are good and contain reproducible examples for the maintainers. This way you can be sure that your report is prioritized and fixed sooner or later. Poor issue reports do not produce value and are often dismissed as spam because the maintainers do not have time to dig deeper without good background information.

The Broader Impact

At Softlandia, our engagement with open-source software significantly enhances its quality and reliability. This collaboration leads to robust, adaptable solutions that support industries and encourage the development of new technologies, such as state-of-the-art agentic CSV, Excel, CRM, and ERP table data processing. By advocating for open standards, we help ensure that technology remains flexible and prepared for future challenges, directly impacting business efficiency and technological advancement.

These contributions serve as well as public references to our capability to deliver outstanding value in software, AI, algorithms, and signal processing. Open-source software is used in almost every product worldwide: most companies already depend on the code we have written without even knowing about it! 

Imagine what we could achieve together by combining our world-class expertise with your business domain. When you consider getting external help for your projects, remember Softlandia.