azure_ai_foundry_tools

Vision and Document Intelligence clients for the Azure AI Foundry Rust SDK.

Features

• Vision — Image Analysis 4.0: tags, captions, object detection, OCR, smart crops, people
• Document Intelligence — Document Intelligence v4.0: OCR, layout, invoices, receipts, ID documents, business cards
• Tracing — Full instrumentation with tracing spans

Installation

[dependencies]
azure_ai_foundry_core = "0.8"
azure_ai_foundry_tools = "0.8"
tokio = { version = "1", features = ["full"] }

Usage

Analyze an Image

use azure_ai_foundry_core::client::FoundryClient;
use azure_ai_foundry_core::auth::FoundryCredential;
use azure_ai_foundry_tools::vision::{self, ImageAnalysisRequest, VisualFeature};

# async fn example() -> Result<(), Box<dyn std::error::Error>> {
let client = FoundryClient::builder()
    .endpoint("https://your-resource.services.ai.azure.com")
    .credential(FoundryCredential::api_key("your-key"))
    .build()?;

let request = ImageAnalysisRequest::builder()
    .url("https://example.com/image.jpg")
    .features(vec![VisualFeature::Tags, VisualFeature::Caption])
    .try_build()?;

let result = vision::analyze(&client, &request).await?;
if let Some(caption) = &result.caption_result {
    println!("Caption: {} ({:.1}%)", caption.text, caption.confidence * 100.0);
}
# Ok(())
# }

Analyze a Document

use azure_ai_foundry_core::client::FoundryClient;
use azure_ai_foundry_core::auth::FoundryCredential;
use azure_ai_foundry_tools::document_intelligence::{
    self, DocumentAnalysisRequest, PREBUILT_READ,
};

# async fn example() -> Result<(), Box<dyn std::error::Error>> {
let client = FoundryClient::builder()
    .endpoint("https://your-resource.services.ai.azure.com")
    .credential(FoundryCredential::api_key("your-key"))
    .build()?;

let request = DocumentAnalysisRequest::builder()
    .model_id(PREBUILT_READ)
    .url_source("https://example.com/document.pdf")
    .try_build()?;

let operation = document_intelligence::analyze(&client, &request).await?;
let result = document_intelligence::poll_until_complete(
    &client,
    &operation.operation_location,
    std::time::Duration::from_secs(2),
    60,
).await?;

if let Some(ar) = &result.analyze_result {
    println!("Extracted text: {:?}", ar.content);
}
# Ok(())
# }

Supported Models (Document Intelligence)

Tracing Spans

Related Crates

• azure_ai_foundry_core — Core types, authentication, and HTTP client
• azure_ai_foundry_models — Chat completions and embeddings
• azure_ai_foundry_agents — Agent Service client

License

This project is licensed under the MIT License.