03.3-vision-extract

Chapter 3.3 -- Vision: Receipt Image to Structured Object

Companion code for Generative AI in .NET, Chapter 3 sections 3.3.1 -- 3.3.4 (and the expense-processor walkthrough in 3.3.5).

Sends a receipt image to a multimodal IChatClient and parses the response into a strongly-typed Receipt record (vendor, date, total, line items). The demo combines two patterns from Chapter 2 -- multimodal input (DataContent) and structured output (GetResponseAsync<T>()).

Run it

export OPENAI_API_KEY=sk-...
dotnet run --project samples/ch03-rag/03.3-vision-extract -- /path/to/receipt.jpg

Manuscript reference

• Manuscript/Chapter-03.md, sections 3.3.1 (multimodal chat) and 3.3.4 (structured extraction).
• Figure 3.10 ("Expense Report Processor").

Prerequisites

• .NET 9 SDK and an OpenAI API key (or any vision-capable IChatClient provider).