What’s on our mind: Exploring AI in open source library software
Last updated on: 12th May 2025| 7th May 2025 | Sam Goldsmith | Koha
Balancing innovation and openness in Koha
Introduction
The principles upon which open source software (OSS) is built are community collaboration, transparency, and universal access. For those committed to the OSS principles, all enhancements must align with these values, ensuring new features remain freely available and do not introduce hidden costs or proprietary dependencies. What does this mean for the role of artificial intelligence (AI), particularly large language models (LLMs) such as ChatGPT, which has captured the attention of the library technology community and much of the online world at large?
Measured approach to AI integration
The huge potential of LLMs is exciting, with innovations, workflow efficiencies and transformation of data analysis already widely observed from this relatively nascent technology. However, there are ethical considerations from the training, running and outputs of these models against which this potential must be balanced. For example, energy consumption (the potential of AI may be unlimited, but the earth’s natural resources are not), information accuracy, biases, and intellectual property rights (primary considerations for information professionals), and transparency and accountability (key concerns for open source software).
With all of this in mind, it may be that open source library software communities take a measured and cautious approach in their exploration of AI; harnessing its benefits whilst remaining mindful of its impacts and to not compromise the core values of open source software and libraries.
Koha library management system and AI
Koha is an open source, fully-featured and scalable library management system used in over 18,000 libraries worldwide. The Koha development community is steadfast in its commitment that solutions and enhancements must not require payment to function. Developments by the community are shared openly with no bespoke, closed-off versions. This ensures that every library, regardless of size or budget, benefits equally. The ethos is clear: innovation should not come at the expense of openness or accessibility. Outside of the source code and architecture, there are fantastic opportunities available to users to integrate AI with Koha, with an example overleaf.
Other examples are also available of uses of AI within open source library software, for example the integration of AI chatbots on library websites, or Health Librarian Chris Rowlands’ use of AI in Metabase which is described in the blog ‘Using an AI tool to assist with my role in health libraries’.
Looking ahead: AI as a community tool
The integration of AI into Koha is not just about technical innovation, it’s about empowering the community whilst upholding its principles. By ensuring that all developments are freely shared and do not introduce costs or proprietary barriers, Koha remains true to its mission of serving libraries worldwide. As AI tools become more sophisticated, the focus will remain on ethical, open, and collaborative adoption where appropriate, ensuring that libraries can continue to innovate without compromising their values.
A practical example: ChatGPT for Koha reports
A notable recent development comes from Koha community developer Arthur Suzuki and released by BibLibre, which created a version of Open AI’s ChatGPT trained on the Koha community reports library to create ‘Koha Reports Assistant’. Due to this training based on the huge collection of SQL reports contributed by the global Koha community, it can offer broad coverage of reporting needs. This presents a valuable resource for library staff who want to generate complex SQL reports using plain, everyday language (with some caveats).
Key features of this tool include:
- Natural language querying – Simply describe the report you need, and the assistant will help you construct the appropriate SQL query, no advanced technical skills required.
- Step-by-step refinement – The tool suggests filters and categories, allowing users to build and refine their queries incrementally.
- Terminology flexibility – Is your library visited by patrons, users, borrowers or staff? Regardless of terminology, the tool is able to deliver relevant results.
- Error handling – If a generated report fails (for example, due to legacy code in the reports library), users can paste the error back into ChatGPT for a revised solution. As with all LLMs, outputs should be reviewed by an experienced user to assess accuracy.
- Simple explanations of SQL. Would you like assistance in understanding what an existing query, written by someone else, does? You can copy the SQL into the Koha Reports Assistant chat and ask for the tool to explain it to receive a simple explanation of a complex query.
If you would like to use it, the Koha Report Assistant is publicly available to try using a free version of ChatGPT, with no usage fees or paywalls. The assistant does not access or process your local library data, mitigating security or data protection risks.
The practical uses of this tool are already being felt by Koha users. Sharon Dugdale, of Cumberland and Westmorland & Furness Council Shared Service, shared:
“I recently had the opportunity to try it out for what (for me) was quite a complex SQL report. After a few adjustments to how I phrased my query, I was able to get exactly the result I needed. I just wanted to share how straightforward and user-friendly I found the experience — it made the process so much easier.”
Kerry-Ann Steeden, Digital and LRC Manager at Blackpool & the Fylde College, shared:
“The Koha Report Assistant in GPT has helped our team refine SQL queries, troubleshoot complex reports, and gain clearer insights into library operations. While we primarily use Power BI with support from Open Fifth, we often need to adapt data for specific stakeholder needs. This tool has boosted our efficiency and confidence by not only generating useful SQL but also explaining it clearly, even when not tailored to our institution’s database.”
How to access Koha Report Assistant
The Koha Report Assistant – ChatGPT can be accessed via your browser.
Watch Open Fifth’s Koha LMS: AI Report Assistant ChatGPT video on YouTube.
Getting started:
-
- Start a chat and describe the report you want in natural language. For example, “Show me all new patrons who have joined since January 2023”.
- The assistant suggests example reports and guides you step-by-step, allowing you to refine your request or add filters of what to show in the report as needed. For example “to the above report show only patron category “student” and list by home library in surname alphabetical order”.
- When you’re happy with the SQL generated, simply copy and paste it into Koha’s reports module to run it.
- If you encounter errors, just paste the error message back into the assistant, and it will suggest corrections.
Once you’ve generated a great reporting SQL, you could share it with your Koha community such as via the Koha Community Mattermost platform, reducing the need for duplication of efforts and energy load.
Important usage notes:
The Koha Report Assistant is a publicly available tool and does not use your local data, however, do not enter any personal or sensitive information in your queries, doing so may breach your institution’s data-handling policies.
Your ChatGPT conversations may be used to train the model, as per OpenAI’s terms.
As with all LLMs, it is necessary to review the outputs produced for accuracy before use.
Outputs should not be used in official decision-making or where they may impact people.
It is recommended that queries are run in test environments first.
Server capacity may be impacted by large queries.
ChatGPT is not open source software.
Please note, this blog is not intended to promote or endorse any specific AI / LLM tools or their utilisation, but to share with readers considerations and developments in this area. Caution must be exercised when trying any new tools, as well as observance of institutional policies, ethical guidelines, licence agreements, and terms and conditions.
