SRS Generator (AI Agent)
Turns a one-paragraph project description into a professional, UML-diagrammed SRS document in under 2 minutes.
SWE (AI/MLOps)LLM AgentsAutomationStreamlitDocument Generation
An n8n-inspired, multi-agent pipeline that automates Software Requirements Specification writing: given a project name and a short description, a manager coordinates dedicated agents per section (Introduction, System Features, Overall Description, System Models) that call the Gemini API to generate content, assemble it into a formatted .docx via python-docx, and embed Use Case, Sequence, and Class diagrams generated through PlantUML, cutting a task that normally takes software teams 4 to 8 hours down to under 2 minutes.
What I built
- Built a modular, n8n-inspired multi-agent pipeline (a manager coordinating an IntroductionAgent, SystemFeaturesAgent, and others) where each agent owns one SRS section.
- Used the Gemini API to generate section content and python-docx to assemble it into a formatted, professional .docx document with headings and structured text.
- Auto-generated and embedded Use Case, Sequence, and Class UML diagrams via PlantUML based on the project description.
- Added retry logic (max 5 retries) and structured logging (loguru) around every agent call for resilience against LLM output failures.
- Built the interface in Streamlit: enter a name, project description, and file name, then download the generated .docx.
Tech stack
SWE (AI/MLOps)LLM AgentsAutomationStreamlitDocument Generation