ai-test-gen
AI-Powered Test Case Generation Framework
92%
Time Saved
$0.002
Cost/TC
207
Test Cases
73% first-pass
Quality
Overview
A hybrid rule-based + LLM pipeline that generates structured manual test cases from Azure DevOps user stories. Combines deterministic scaffolding with Gemini 2.5 Flash for natural-language enrichment, ChromaDB for semantic step matching, and automated upload to ADO Test Plans.
The system first extracts acceptance criteria and UI elements from user stories using spaCy NLP, then applies rule-based templates for deterministic structure. The LLM enriches test steps with natural language and handles edge cases. ChromaDB provides semantic similarity matching to maintain consistent wording across related test cases.
Architecture
Input — Azure DevOps user story with acceptance criteria
NLP Extraction — spaCy extracts entities, actions, UI elements
Rule-Based Scaffold — Deterministic templates generate test structure
Semantic Matching — ChromaDB finds similar existing steps for consistency
LLM Enrichment — Gemini 2.5 Flash refines language and adds edge cases
Output — Structured CSV + automated upload to ADO Test Plans