How it Works
OnboardGhost uses a sophisticated pipeline of static analysis, large language models, and vector retrieval to build a comprehensive map of any codebase in seconds.
Repository Analysis
We fetch your repository using the GitHub API. Our system analyzes the file tree and extracts critical metadata—dependencies, configuration files, and README documentation—to understand the project's skeleton.
AI Reasoning with Gemini
The raw data is processed by Google Gemini Pro. It identifies the tech stack, determines the project purpose, and detects database or third-party service requirements. It essentially 'reads' the project for you.
Dynamic Roadmap Generation
Based on the analysis, we generate a structured onboarding roadmap. This isn't a generic template; it's a tailormade sequence of tasks ranging from environment setup to making your first feature contribution.
The Ghost Mentor (RAG)
We use Retrieval-Augmented Generation (RAG) to power our mentor. Critical code snippets are embedded and stored in Upstash Vector. When you ask a question, the mentor retrieves the exact code context needed to give an accurate answer.
Persistent Progress
Every task you complete and every analysis you run is stored securely in Firebase. This allows you to pick up exactly where you left off across any device, building a permanent record of your codebase knowledge.
Privacy First
We never store your entire source code permanently. We only analyze what's necessary to build your roadmap and mentor, using secure vector embeddings for reasoning.