Skip to content

Use Cases

If you are asking yourself "what is this even used for?", this page is for you.

Large language models (LLMs) have inspired many creative applications in the recent years, and our aim is to channel these applications into a framework that offers more robust and scalable solutions for the biomedical domain. We see BioChatter as a developer toolkit that can be used to create a range of custom applications for specific user bases. As such, our main target audience are developers in the biomedical domain.

Chatbots

Naturally, the most attractive general use of LLMs is to have a language interface to some functionality that previously was not accessible through such simple means. Depending on the requirements of the application, this can be either trivial ("Create a paragraph from these bullet points.") or extremely complex ("What is the driver of my patient's cancer?"). Biomedical use cases tend to be biased towards the latter, often necessitating the integration of various sources of knowledge and quality assurance of the conversational system. As a result, BioChatter is not designed to function as a sole chatbot without integration of other data or knowledge; this functionality is provided by numerous companies.

Instead, we at least include specific prompts that tune the LLM towards the desired application and user base, which can be configured by the developer implementing the application. More powerful applications are the focus however, including tool use (external software parameterisation), knowledge graph and vector database integration, multimodal inputs, and modular agentic workflows.

Graphical user interfaces

Increasing accessibility to previously inaccessible functionality naturally implies that the developer needs to think about how to make the functionality available to the (by definition non-technical) user. As a result, we include suggestions for diverse graphical user interface (GUI) frameworks in the ecosystem. Specifically, we provide instructive examples for relatively simple prototyping frameworks based on Streamlit, which we brand as BioChatter Light and which can be used for rapid development of a range of user interfaces:

For more complex applications, we also provide a modular system based on current web technologies (FastAPI and Next.js), which we brand as BioChatter Next. The web application is driven by a RESTful API, which we implement in BioChatter Server. An application example can be seen in the real-world use case.