IBM has introduced a tutorial on building an AI research agent for image analysis using Granite 3.2 Reasoning and Vision models. The agent leverages agentic AI workflows and retrieval-augmented generation (RAG) to extract insights from images, supporting use cases like analyzing architecture diagrams, business dashboards, and historical photos.

The system operates locally using Open WebUI and Ollama for privacy and efficiency. It first identifies key research topics from an image, then runs parallel research agents to gather insights from web sources and user documents, ultimately generating a comprehensive report.

The tutorial provides code examples and setup instructions, allowing developers to implement their own AI-powered image research agent. More details are available on IBM’s GitHub repository and the Granite Playground.