User uploads image or inputs path

Triggers extraction

Reads image, encodes, sends to MCP server

Receives extracted text

Receives image data

Calls Gemini AI for extraction

Processes image and prompt

Returns extracted text