Context
This project involves an automated workflow to transform raw client requirements into well-defined product documentation. The process starts by extracting project requirements from email threads, where various departments share desired product features. Through a sequential pipeline, this information is summarized, and Business Requirement Documents (BRDs), user stories, and detailed test cases are produced. Leveraging Large Language Models (LLMs) orchestrated in Crew AI, this system enables efficient and structured document creation, ensuring each phase builds accurately on the previous one for a cohesive end-to-end product definition.
Requirements
Automated Document Creation: Seamlessly transform client and departmental inputs into product documentation across stages.
Sequential Workflow: Each stage's output informs the next, producing requirements summaries, BRDs, user stories, and test cases in sequence.
LLM-based Document Generation: Employ Large Language Models for summarization, requirement extraction, and structured documentation creation.
Testing and Coverage: Define functional, non-functional, and edge case test cases to ensure comprehensive product testing.
Collaborative Agent System: Utilize multiple LLM agents working cohesively to enhance the accuracy and consistency of the documentation pipeline.
Approach
Requirement Extraction:Extract product requirements from email threads where various departments contribute feature requests and specifications.
Summarization and Analysis:Use LLM agents to summarize the raw requirements, distilling key points and aligning with project objectives for deeper understanding.
BRD Creation:Leverage summarized requirements to generate a comprehensive Business Requirement Document (BRD), detailing scope, objectives, and high-level specifications.
User Story Development:Based on the BRD, generate user stories to capture specific use cases and expected user interactions with the product.
Test Case Generation:Create functional, non-functional, and edge case test cases from user stories to ensure thorough validation, considering all possible usage scenarios.
Sequential LLM Agent Pipeline:Implement a Crew AI system with Mixtal 7x8B models, where each LLM agent specializes in a stage and passes outputs to the next, facilitating streamlined and coordinated document generation.
Technologies Used
LLM Models: Mixtal 7x8B models for high-quality, contextual document generation.
Orchestration: Crew AI to manage and coordinate LLM agents across stages.
Automated Document Pipeline: Sequential pipeline design for requirement extraction, summarization, BRD creation, user story formulation, and test case generation.