For Harvard University’s Arts and Humanities Research Computing unit (DARTH), the increased instability of digital humanities funding threatens both new and existing DH projects, especially those that take the form of dynamic, database-driven research tools. Built for faculty researchers to perform rich queries across complex data models, these projects are expensive to build as well as maintain. Nonetheless, faculty continue to view these sophisticated, customized discovery tools as indispensable to their research. This presentation introduces DARTH’s use of the free, client-side static search library Pagefind to balance search, sustainability, and budget in DH projects. Three case studies—Tsumeb Mine Notebook, the Amendments Project, and the Black Teacher Archive—illustrate how Pagefind can be used to build new projects, sunset legacy databases, and embed preservation planning from the start, all with greatly reduced financial and engineering resources.

In recent years, many DH practitioners have appropriately shifted toward Static Site Generators (SSGs) to address pressing sustainability concerns. This “minimal computing” approach simplifies the overall technological stack, reduces maintenance overhead substantially, and enables project hosting on highly inexpensive or even free services. While static preservation effectively guarantees the long-term availability of project files, this strategy introduces a significant trade-off: it often sacrifices the sophisticated search interfaces necessary for scholarly interpretation and analysis. When a dynamic site is sunsetted to a static form, the underlying database may be archived on specialized platforms, such as Dataverse.

https://dataverse.harvard.edu/.

But such data repositories, as Raffaele Viglianti and Gimena del Rio Riande argue, are built specifically for long-term storage and not for the kind of user-facing interfaces that foster rich engagement (Vilianti and Riande, 2025). Data may persist, but discoverability diminishes when dynamic interfaces disappear, leaving users with basic keyword search or, in some cases, no search at all. Consequently, when dynamic interfaces vanish, discoverability diminishes, sometimes leaving users—including faculty Principal Investigators (PIs)—with only basic keyword search or, in some cases, no search capability whatsoever.

Humanities faculty at Harvard cite sophisticated, customized discovery tools as key to their research. Many faculty-led projects involve meticulously collected and cleaned data, typically organized, added, and enhanced within PostgreSQL databases which then serve dynamic front-end websites.

https://www.postgresql.org/.

DARTH has implemented ElasticSearch in several projects, providing advanced features like faceted interfaces, field weighting, and rapid multi-parameter querying that dramatically enhance users' ability to traverse material structured in complicated data models.

https://www.elastic.co/elasticsearch/.

Faculty PIs rely heavily on these flexible search interfaces because they do more than merely help locate specific items; they enable analytical engagement with the data. These capabilities are crucial for supporting exploratory research questions, surfacing connections between records, and contrasting different records. Thus, search interfaces function as critical instruments that actively encourage researchers to interact with cultural and historical materials as data, setting the crucial stage for deeper analysis.

Declining funding has put further pressure on this functionality gap between the needs of researchers and the limitations of sustainable technology. DARTH is directly addressing this challenge by using Pagefind, a free, open-source, client-side static search library released in 2022.

https://pagefind.app/.

Pagefind offers significant technical and functional advantages for modern DH projects. Its zero-infrastructure, client-side operation means it runs entirely in the user's browser, eliminating the need for and ongoing costs of server components. It also boasts rich features while remaining easy to set up. Out of the box, it provides a comprehensive suite of advanced features, including full-text search, faceted filtering, customizable user interface components, multilingual support, content weighting, and metadata indexing.

Pagefind’s unique indexing process makes it superior to existing static search tools, especially for large datasets or complex research requirements. Typically, Pagefind begins by locating and parsing the HTML files within the static site directory, extracting textual content and organizing it into searchable blocks. Each block is treated as an identifiable and individual segment. During the tokenization process, every token is mapped back to the specific block and page where it appears. Crucially for performance, Pagefind then "chunks" the resulting index into small pieces, which are compressed to facilitate faster retrieval by the browser. When a user initiates a search on the site, the browser is only required to load the necessary pieces of the index. By comparison, libraries like LunrJs and FlexSearch.Js require the user to download the entire search index, slowing down page load speed and decreasing performance.

https://lunrjs.com/; https://github.com/nextapps-de/flexsearch.

To demonstrate Pagefind’s practical utility and its role in meeting the challenges of search, sustainability, and budget, the presentation will offer three distinct case studies, each representing a different stage in a DH project’s lifespan: planning a new project, sunsetting a legacy project, and designing a project for static maintenance. DARTH’s implementation of Pagefind across these projects utilizes the Pagefind JavaScript API via a set of reusable Vue components created internally.

https://vuejs.org/

Although Pagefind's pre-built UI offers excellent initial functionality, these components were developed to simplify the creation of bespoke search interfaces incorporating single- or multi-select filters and text inputs. While this strategy introduces some volatility into the development process, it provides flexibility of form and functionality while significantly reducing the developer workload when integrating Pagefind across diverse projects.

The first case study focuses on a new project, the Tsumeb Mine Notebook.

https://tmn.fas.harvard.edu/.

This digital collection documents minerals from a world-renowned mine in northern Namibia. Built with the Astro SSG framework and hosted on Github Pages, this project was born from an episode of digital loss after a previous online resource went offline.

https://astro.build/; https://docs.github.com/en/pages.

The presentation will outline how Pagefind transforms data exported from a museum database into a rich search experience. A technical demonstration will show core research tasks—such as filtering minerals by combinations of associated minerals, chemical group, and provenance—that directly serve interpretive questions posed by the project team’s domain experts.

The second case study, the Amendments Project, illustrates Pagefind’s role in responsibly solving the pervasive problem of "zombie" projects often left stranded on university servers.

https://amendmentsproject.org/.

This project, originally created for historian Jill Lepore, contained a database of 22,788 attempts to revise the U.S. Constitution, powered by a dynamic Django backend.

https://www.djangoproject.com/.

Upon the expiration of grant funding, DARTH investigated how to transition the database to a sustainable static architecture without sacrificing the powerful search capabilities the project was known for. By manually indexing a CSV export of the database using Pagefind’s APIs, DARTH preserved the project’s core research value in a completely static, archivable format. The presentation will feature a direct comparison of search tasks to prove the static site retains the equivalent analytical power of the original dynamic system.

The final case study involves the Black Teacher Archive, focusing on the planning phase of a new project.

https://bta.gse.harvard.edu/.

The project was initiated due to dissatisfaction with the limited search capabilities of an existing many-purpose library database. To adequately address the PIs' research needs during the active research phase, the project is initially utilizing ElasticSearch. However, the project plan strategically leverages Pagefind within a SvelteKit framework.

https://svelte.dev/

This approach facilitates the creation of a highly interactive user experience that successfully combines faceted filtering with full-text search. Critically, the use of the SvelteKit framework means that the entire site can be easily converted to a static form for long-term preservation simply by switching the deployment adapter. This forward-looking strategy ensures that sustainability is embedded into the project’s core design, making the project "preservation-ready" from its inception.

For DARTH, Pagefind has become a proven means of preserving the analytical richness inherent in DH projects while simultaneously achieving high sustainability with reduced resources. It is not meant as a blanket solution, but rather a new strategy for pursuing a balance across user engagement, longevity, and budget. As the field of digital humanities adapts to new technological and financial circumstances, we hope this presentation will contribute to a broader discussion about the import and viability of custom, public-facing research tools in future humanities scholarship.