AI Test Scenario Generator MVP

I need a first-cut platform that reads Jira Cloud tickets and instantly turns each one into ready-to-run, step-by-step test cases with clear expected results. In this initial version the spotlight is entirely on “creating test scenarios from Jira tickets,” leaving PR analysis and high-level coverage dashboards for later iterations.

Here is the workflow I have in mind:

• The service connects to Jira (OAuth or PAT) and pulls the full ticket content—description, acceptance criteria, every comment, plus any linked tickets—so the model has maximum context.
• An LLM-driven component analyses that context, distils business rules, edge cases and negative paths, and produces concise but thorough manual test cases in a structured JSON (or Markdown) format: title, pre-conditions, ordered steps, expected results.
• The response is surfaced through a simple web UI and a REST endpoint so QA engineers can copy, export, or push them straight into their test-management tool.

Deliverables
1. Running prototype (Docker preferred) that I can point at a sample Jira project and receive generated test cases.
2. Source code with clear comments and a lightweight README explaining model choice, prompt strategy, and how to retrain or swap the model.
3. Short loom or screenshot walkthrough proving end-to-end flow on at least three example tickets.

Acceptance criteria
• Given a ticket containing description, acceptance criteria, comments, and links, the system returns a minimum of three distinct test cases, each with numbered steps and explicit expected results.
• Tests cover both positive and negative paths present in ticket conversations.
• Turnaround time per ticket stays under 15 seconds on the sample dataset.

Tech is flexible, though Python, Node.js, LangChain, OpenAI or similar LLM APIs, vector stores, and Jira REST SDK are very welcome. Clean code, modular architecture, and maintainable prompt design will influence the selection.