Generative AI Apps & Agents with Foundry
Exam objectives:
- Create effective system and user prompts for generative AI models
- Deploy a model and interact with it in the Foundry portal
- Create a lightweight chat client application using the Foundry SDK
- Create and test a single-agent solution in the Foundry portal
- Create a lightweight client application for an agent
Overview
The largest and most heavily tested area of AI-901 is building generative AI applications and agents using Microsoft Foundry (Azure AI Foundry). You need to be able to perform these tasks in the portal and understand the Python SDK code that powers client applications.
Key Concepts
The Foundry Workflow
1. Create an Azure AI Foundry project
↓
2. Deploy a model (model catalog → deploy)
↓
3. Test in the Foundry playground
↓
4. Build a client app using the Foundry SDK
↓
5. (Optional) Add tools → create an agent
Prompts
| Type | Role | Example |
|---|---|---|
| System prompt | Sets the model's persona, tone, and constraints for the whole session | "You are a helpful Azure certification tutor. Answer only questions about Azure." |
| User prompt | The end-user's input | "What is the difference between AI-901 and AI-102?" |
| Assistant message | A previous model response — used in multi-turn conversations | — |
Effective prompt techniques:
- Be specific and unambiguous
- Provide context and examples (few-shot prompting)
- Use the system prompt to constrain behavior
- Break complex tasks into steps (chain-of-thought)
Chat Client App — Foundry SDK Pattern
from azure.ai.inference import ChatCompletionsClient
from azure.ai.inference.models import SystemMessage, UserMessage
from azure.core.credentials import AzureKeyCredential
client = ChatCompletionsClient(
endpoint="https://<your-project>.services.ai.azure.com/models",
credential=AzureKeyCredential("<your-key>")
)
response = client.complete(
model="<deployment-name>",
messages=[
SystemMessage("You are a helpful assistant."),
UserMessage("Explain Azure AI Foundry in one sentence."),
],
temperature=0.7,
max_tokens=200,
)
print(response.choices[0].message.content)
Agents
An AI agent is a model + a set of tools that it can call to complete multi-step tasks.
| Concept | Description |
|---|---|
| Tool | A function the agent can call — e.g., search, run code, call an API |
| Instructions | A system-level prompt that defines the agent's goal and behavior |
| Thread | A conversation session with the agent |
| Run | A single execution of the agent on a thread |
# Agent client app pattern
from azure.ai.projects import AIProjectClient
from azure.identity import DefaultAzureCredential
project_client = AIProjectClient(
endpoint="https://<your-project>.services.ai.azure.com",
credential=DefaultAzureCredential()
)
agent = project_client.agents.get_agent("<agent-id>")
thread = project_client.agents.create_thread()
project_client.agents.create_message(thread.id, role="user", content="What is today's weather?")
run = project_client.agents.create_and_process_run(thread.id, agent.id)
messages = project_client.agents.list_messages(thread.id)
print(messages.data[0].content[0].text.value)
Azure Services & Foundry Features
| Feature | Where in Foundry | Purpose |
|---|---|---|
| Model catalog | Foundry portal → Model catalog | Browse and deploy models |
| Playground | Foundry portal → Playground | Test prompts interactively |
| Agents | Foundry portal → Agents | Build and test single-agent solutions |
| Foundry SDK | azure-ai-inference, azure-ai-projects packages | Build client apps in Python |