What problem does it solve?
This Skill documents how to use the @azure/ai-projects TypeScript SDK to remove friction when creating, configuring, and operating Azure AI Foundry projects by providing clear patterns for agents, connections, deployments, datasets, indexes, and evaluations.
Core Features & Use Cases
- Project Client Setup: Authenticate with DefaultAzureCredential and instantiate AIProjectClient using AZURE_AI_PROJECT_ENDPOINT.
- Agent Lifecycle: Create versioned agents, attach tools (code interpreter, file search, web search, functions), run agents via the OpenAI client, and clean up versions and conversations.
- Connections & Deployments: Discover and retrieve project connections (with credentials), list and filter model deployments, and use connections to create direct SDK clients (OpenAI, Search).
- Data & Evaluation Workflows: Upload datasets (files and folders), create and manage indexes, run built-in and custom evaluators, and poll for evaluation results.
- Use Case: Build a reproducible RAG pipeline that uploads documents, indexes content, creates an agent using file_search, and evaluates output groundedness against a dataset.
Quick Start
Create an AIProjectClient with DefaultAzureCredential, list deployments in your project, and retrieve an OpenAI client for generating responses.
