Extracts an illustrative invoice into a typed InvoiceData record using the structured-response API.
Install the .NET 10 SDK and .NET 8 runtime. Windows/Linux/macOS require a compatible platform-specific llama-server runtime and an application-owned GGUF chat model. See runtime setup and the license.
No MCP or Document Editor is required to execute this focused API example. Use within solutions under the repository's TX Text Control license terms.
From the repository root, after provisioning the runtime:
dotnet run --project samples/LocalStructuredOutput -- "/path/to/chat-model.gguf"The invoice number, date, customer and total are printed. Use a chat model supporting the runtime's JSON/schema constraints; always validate extracted data before business use.
The single argument is the GGUF path. Use an absolute path, especially when running from another directory. Set TXTEXTCONTROL_AI_LLAMA_SERVER to an installed executable or install a managed runtime explicitly using the package API described in the runtime guide. Startup does not download missing runtimes by default.
Model files are not included. Missing-runtime, unsupported-model and insufficient-memory errors are separate from NuGet restore errors. Try a smaller model/context if memory is limited. See the complete AI API guide for LocalModelOptions, cancellation, disposal and session behavior.