Inspiration
In the fast-paced world of software development, it's no secret that developers face a multitude of challenges and problems on a daily basis. However, the real challenge lies in effectively capturing, tracking, and discussing these issues within the context of a sprint. Traditional manual methods of sprint analysis, often conducted by managers and stakeholders, tend to fall short in this regard. They can struggle to encompass the full spectrum of issues that developers encounter, and this process can be painstakingly time-consuming. Moreover, non-technical stakeholders might inadvertently overlook crucial insights, as they may not possess the deep technical understanding required to ask the right questions. That's where SprintReflect-AI comes in.
What is SprintReflect-AI?
SprintReflect-AI is a revolutionary tool designed to bridge the gap between the complex world of software development and the need for effective sprint analysis. This cutting-edge solution leverages the power of artificial intelligence to address the challenges faced by both developers and stakeholders. It automates the process of issue tracking and analysis, enabling developers to provide detailed insights through well-structured surveys tailored to the specific problems they encountered during the sprint. This data is then transformed into comprehensive sprint retrospectives, allowing managers and stakeholders to gain a deeper understanding of the issues at hand, identify areas for improvement, and make informed decisions. With SprintReflect-AI, technical and non-technical stakeholders alike can collaborate seamlessly, ensuring that no crucial aspect goes unnoticed and that every sprint becomes an opportunity for growth and refinement in the world of software development.
How we built it
crafted using the powerful Atlassian Forge platform in combination with OpenAI's ChatGPT LLM.
Forge
On the Forge side, SprintReflect leverages the full potential of Forge's serverless platform, employing various key components for a seamless experience:
Storage API: We harnessed the Storage API to securely save survey data, extract critical insights from sprint and issue data, and generate detailed reports with the help of ChatGPT.
AsyncEvents API: For managing time-consuming tasks that require asynchronous execution, we incorporated the AsyncEvents API into our business logic. This ensures that certain operations can run efficiently without impacting overall system responsiveness.
Webtrigger: To accommodate ChatGPT's response time, we implemented a proxy server that operates independently, eliminating any timeout constraints. Forge initiates the call to the proxy, which in turn communicates with ChatGPT. Upon completion, the proxy signals Forge via the Webtrigger mechanism.
Custom UI & UI Kit: We judiciously employed both Custom UI and UI Kit components. Custom UI was utilized for the report menu, which includes intricate elements like charts and tree tables. UI Kit, on the other hand, was chosen for the survey module, which primarily involves form-based interactions.
Jira Product Events: We harnessed Jira Product Events to actively track various data points and seamlessly integrate them into the ChatGPT prompts.
ChatGPT
Regarding ChatGPT, we employed OpenAI's ChatGPT-3 model, optimizing it with specific instructions to deliver efficient and contextually relevant output for sprint analysis. This synergy of Atlassian Forge and ChatGPT forms the foundation of SprintReflect-AI, empowering teams to enhance their sprint analysis process.
Challenges we ran into
During the development of SprintReflect-AI, we encountered a significant challenge due to the serverless nature of the Forge platform, which imposed a strict 25-second execution time limit. This constraint became particularly problematic when making calls to the ChatGPT API, as responses sometimes exceeded this time limit, leading to frequent timeouts. To address this issue, we devised a robust solution. We leveraged the Async Events API for managing time-consuming tasks asynchronously, implemented a proxy server to handle ChatGPT communications free from time constraints, and employed the Web Trigger mechanism to seamlessly signal Forge once ChatGPT responses were resolved. This comprehensive approach successfully resolved the challenge and ensured the smooth operation of SprintReflect-AI.
What's next for SprintReflect-AI:
Enhanced Data Points: We will be dedicated to expanding the range of data points to gather more precise and detailed feedback, ensuring that the insights generated will be richer and more insightful.
Vision Capability: To further enrich the user experience, we will work on integrating vision capabilities that will enable SprintReflect-AI to understand and process images added within issues, providing a holistic view of the development context.
Cross-Platform Integration: We will actively explore the integration of various code versioning platforms to parse and incorporate data from Pull Requests. This will offer a more comprehensive analysis by including data points from multiple sources.
Advanced Security Measures: Our commitment to responsible AI will extend to security. We will implement advanced filters, utilizing technologies like Langchain, to protect against prompt injection, ensuring the integrity of the data and insights generated.
Log in or sign up for Devpost to join the conversation.