We evaluate our interface from the perspective of a user.

For the chat page, we've implemented a messaging interface designed to guide users seamlessly through their interactions with the chatbot. To enhance user experience during wait times, a loading overlay appears, signaling users to wait while their response is being prepared. In instances where the chatbot lacks information on a specific query, it transparently communicates its lack of knowledge, maintaining clear and honest communication. Additionally, to aid users in navigating through their chat history, the interface displays each chat with a small excerpt, allowing users to easily distinguish between different conversations and topics.

On the data sources page, users can easily identify the status of each data source, with labels clearly marked as 'Accepted' or 'Pending'. Users can add new data sources via buttons, which are inscribed with messages indicating the option to either add a website or a PDF file. For managing existing data sources, the interface includes straightforward buttons featuring universally recognizable icons: a pen on paper for editing and a bin for deleting. These icons ensure users can easily navigate and modify their data without confusion.

In the development of our platform, we diligently followed all specified requirements to ensure maximum utility for our users. We successfully integrated essential functionalities, including the ability to read text from websites and PDFs, which broadens the platform's accessibility by accommodating various data formats. Additionally, our platform is designed to formulate responses by distilling only the necessary information, thereby enhancing efficiency and preventing information overload. Furthermore, we crafted a user-friendly and intuitive interface, making the chatbot accessible to users with varying levels of technical proficiency and ensuring a seamless interaction.

The choice of Netlify and Azure for deployment services, was their guarantee of uptime and availability. Accordingly, but using Azure App Services, it employs autoscaling capabilities which can adjust system's resources depending on current demand.

Moreover, both deployment services have analytics and monitoring tools which can be used to track the system's performance and identify potential issues. This enables the admins to identify and address performance issues proactively.

A stability concern for our system is its dependency on the OpenAI API, which necessitates a continuous financial commitment to maintain access to the API's capabilities. Consequently, the sustainability of our system is tied to the availability of our budget. In the event that our financial resources are exhausted, our system will encounter operational challenges, disrupting our service offerings.

Our project's performance may experience a decline during extended interactions with the chatbot, particularly when users engage in lengthy conversations. This performance degradation is attributed to our method of preserving context within the chat, wherein we feed all preceding queries from the conversation to the Large Language Model (LLM) to maintain an understanding of the dialogue's context. As the conversation progresses and accumulates more queries, the LLM is required to process an increasingly larger volume of information. This heightened demand on the LLM's processing capabilities can lead to longer response times and a decrease in overall performance as the conversation lengthens.

We deployed our backend using Azure, which is well-suited for a diverse range of applications, ensuring a flexible and scalable foundation for our system. On the front end, we utilised Netlify for deployment, which optimises performance and simplifies the development process, thereby broadening our system's reach to accommodate users from various technological backgrounds. This strategic combination ensures that our platform is accessible and user-friendly, catering to a wide audience with different needs and tech preferences.

Our project has been coded following object-oriented programming (OOP) and software engineering principles, ensuring a robust and scalable structure. By adhering to these principles, we have organized the code into logically distinct files, facilitating modularity and maintainability. This approach not only promotes code reusability but also enhances the readability and navigability of the codebase, allowing for efficient debugging and seamless expansion. Consequently, any new features can be integrated with minimal disruption, and potential bugs can be swiftly identified and rectified, ensuring the project's long-term sustainability and adaptability.

In our group, we divided the workload, ensuring that each member focused on a specific task to maximize efficiency and cohesion. We employed a Gantt chart to track our progress and maintain alignment with our project timeline. Despite these efforts, we encountered challenges with the frontend development due to the unexpectedly high volume of code required, which made us fall behind slightly, causing a deviation from our planned schedule.

In light of this, we have strategised for future projects to allocate one team member to commence frontend development from the onset. This approach will allow for parallel progress, with the remaining team members focusing on backend tasks, which typically demand less time. Once the backend is complete, these members can then assist in frontend development, facilitating a more balanced workload and efficient use of time.

Throughout the project, we maintained open lines of communication with our clients through weekly meetings, ensuring that their feedback and requirements were continuously integrated into our development process. We held additional meetings as needed to address urgent issues and utilised Google Drive to effectively share and manage files amongst the team. This structured approach to project management and client interaction allowed us to navigate challenges and adapt our strategies accordingly.