You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Thanks for creating content for Ignite! Please follow this guidance so all ignite repos have the same look and feel.
4
-
5
-
## Put your docs in the docs folder
6
-
7
-
This repo is configured for a user to view the docs through mkdocs.
8
-
9
-
1. update the `mkdocs.yml` file to reference your session. Look particularly for the `site_name`, `site_author`, and `repo_name` tags.
10
-
1. Subfolders in the docs folder will show up as tabs in the navigation bar.
11
-
12
-
## Guidance
13
-
14
-
1. Do not turn on github pages for this repo, traffic to github pages isn't captured in our metrics.
15
-
1. Do not store slides or recordings in this repo
16
-
17
-
## Update the badges!
18
-
1. if the foundry discord and github links aren't appropriate for your session you can remove or change them.
19
-
2. See shields.io for how to build your own dynamic or static badge for your community destinations, or other sample repos for how they've done
20
-
3. Please limit your repo to one row of badges under the session title.
21
-
22
-
23
-
## Update Readme
24
-
25
-
1. Fill out the content below in this file, below the banner graphic, including the session code.
26
-
2. Please embed links to Learn with your campaign codes!
27
-
3. Add reources for your session to the Resources and Next Steps table
28
-
29
-
## Improve SEO
30
-
31
-
1. Update the Repo Info for this repo
32
-
1. Click the gear icon⚙️ in the upper right.
33
-
1. Set a good description of this repo.
34
-
1. Add the technologies that you're using in this session.
35
-
36
-
## Questions?
37
-
Send them to [Mike Kinsman](mailto:mikki@microsoft.com) and [Laurent Bugnion](lbugnion@microsoft.com), or surface your question through your content lead.
38
-
39
-
## Finished?
40
-
- Remove the folders that you dont need or didn't use, please dont leave them empty.
## 🔥BRKXXX: Learn how to observe, manage, and scale agentic AI apps using Azure
7
+
## Learn how to observe, manage, and scale agentic AI apps using Azure
50
8
51
9
[](https://discord.com/invite/ByRwuEEgH4)
52
10
[](https://aka.ms/foundry/forum)
53
11
54
12
### Session Description
55
13
56
-
*Add Session Description*
14
+
This hands-on workshop will provide participants with the skills to effectively manage, govern, and scale agentic AI applications using Azure and Azure AI Foundry. The session will cover observability capabilities, model management policies, agent functionalities, and governance strategies. Participants will engage in practical exercises to apply these concepts in real-world scenarios.
15
+
16
+
-**Level:** 300-400
17
+
-**Duration:** 4 hours
18
+
19
+
### Application Scenario
57
20
21
+
Imagine this. You are an AI engineer at Zava, an enterprise retail store specializing in home improvement goods for DIY enthusiasts. Your team is building **Cora**, a shopping assistant AI to answer customer queries in-store and online. You have three requirements:
22
+
23
+
1. The solution must reflect the Zava brand with a custom tone & style
24
+
1. It should be cost-effective to deploy given the simple, narrow task
25
+
1. It should support end-to-end observability to ensure trustworthy AI
58
26
59
27
### 🧠 Learning Outcomes
60
28
61
-
By the end of this session, learners will be able to:
29
+
By the end of this session, you will be able to
62
30
63
-
-
64
-
-
65
-
-
31
+
- Build and deploy an agentic AI retail chatbot on Azure AI Foundry
32
+
- Evaluate the quality, safety & agentic efficacy of chatbot operation
33
+
- Fine-Tune the chatbot model to customize tone & style of response
34
+
- Distill chatbot behavior to a smaller model for cost-effective operation
35
+
- Trace and monitor chatbot operations to detect & debug performance issues
36
+
- Understand how Azure AI Foundry enables end-to-end observability for AI
66
37
67
38
### 💻 Technologies Used
68
39
69
-
1.
70
-
1.
71
-
1.
40
+
- Azure Open AI models
41
+
- Azure AI Foundry
42
+
- Azure AI Evaluation SDK
43
+
- Supervised Fine Tuning
44
+
- Model Distillation
45
+
- Azure Monitor
46
+
- Azure Application Insights
47
+
- Azure AI Search
72
48
73
49
### 📚 Resources and Next Steps
74
50
@@ -87,21 +63,26 @@ By the end of this session, learners will be able to:
We will provision the AZD "Get Started with AI Agents" template as our infrastructure. These notes correspond to [this commit](https://github.com/nitya/get-started-with-ai-agents/commit/c63e1384ae6a295b3d57378fab02ac4227bf835d) on the AZD template repo. My goal is to document the usage and customization of this template to support various projects - as pre-provisioned infrastructure.
4
+
5
+
<br/>
6
+
7
+
## 0. Fork Template & Launch Codespaces
8
+
9
+
1. Visit the [Get Started With AI Agents](https://github.com/Azure-Samples/get-started-with-ai-agents) repo
10
+
1. Fork it to your profile to get a custom copy (.e., [my fork-branch](https://github.com/nitya/get-started-with-ai-agents/tree/sep2025-pdy123-infra))
11
+
1. Launch GitHub Codespaces on your fork
12
+
1. Wait till the browser VS Code loads - and an active terminal appears.
13
+
14
+
_You are ready to deploy the template_
15
+
16
+
17
+
## 1. Set AZD Env Variables
18
+
19
+
The azd template can be customized in many ways - the easiest starting point is to use environment variables to customize the resource deployment paramters. Run these commands in the Codespaces terminal. _Note: Pick unique names for the last 2 variables that set the Foundry project and resource names in Azure AI Foundry Portal_
20
+
21
+
```bash
22
+
azd env set AZURE_AI_AGENT_DEPLOYMENT_CAPACITY 100
23
+
azd env set ENABLE_AZURE_MONITOR_TRACING true
24
+
azd env set AZURE_TRACING_GEN_AI_CONTENT_RECORDING_ENABLED true
25
+
azd env set USE_AZURE_AI_SEARCH_SERVICE true
26
+
azd env set USE_APPLICATION_INSIGHTS true
27
+
azd env set AZURE_AIPROJECT_NAME nitya-pdy123-aiproj-v1
28
+
azd env set AZURE_AISERVICES_NAME nitya-pdy123-aisvcs-v1
29
+
```
30
+
31
+
## 2. Deploy it
32
+
33
+
```bash
34
+
azd up
35
+
```
36
+
37
+
Note that we did not do any pre-authentication step. So we'll get prompted to login using a device-code workflow like this:
38
+
39
+
```bash
40
+
? Would you like to log in now? Yes
41
+
42
+
Start by copying the next code: XXXXXXX
43
+
Then press enter and continue to log in from your browser...
44
+
```
45
+
46
+
Next you'll be prompted to select your subscription and the region for deployment. I used `East US2`. Then just let the provisioning complete. **This can take 10-15 minutes given our added requirements** - the default (basic) template provisions in 4-6 minutes. _The output looks like this - the first section provisions resources, the second deploys the app_.
47
+
48
+
```bash
49
+
You can view detailed progress in the Azure Portal:
Click on the **Endpoint** URL provided in the output - or visit the Azure Portal and look for the Azure Container Apps resource and click the _Application Url_. You should see an agent application like this:
6
+
7
+
🌟 | Try asking a question like `Tell me about camping tents` - and see what happens.
8
+
9
+
10
+
11
+
12
+
## 2. Visit Agent Dashboard
13
+
14
+
Click the `Build & deploy agents with Azure AI Foundry` link (bottom left) to navigate to the underlying Agent on the Azure AI Foundry portal. You should see something like this:
15
+
16
+
🌟 | Select the agent name to see its details. Explore the integrated tools (Knowledge & Actions). Update the Agent description and instructions.
17
+
- Agent name is helpful identification. _Change from "agent-template-assistant" to "cora-for-zava"_. Note how the changes are automatically saved. However, the container app will need to be updated separately.
18
+
- Description is for human users. It does not affect agent behavior. Example: _Cora is a customer support agent for the Zava retail enterprise. Cora helps DIY enthusiasts find and purchase home improvement products in Zava stores._
19
+
- Instructions are for agents. They guide agent actions. Example: _Use AI Search always. Avoid to use base knowledge._ - we'll modify this later.
20
+
21
+
22
+
23
+
Click **Try in playground** to activate the agent for interactive exploration.
24
+
- See how changes made to description or instructions here are _automatically saved_ so they will be visible to agent detailed view in earlier dashboard.
25
+
- Try asking [a sample question](https://github.com/Azure-Samples/get-started-with-ai-agents/blob/main/docs/sample_questions.md) like _What's the best tent under $200 for two people, and what features does it include?_ 🚨 - fix Azure AI Search. It is not grounded.
26
+
27
+
28
+
29
+
Click the **View Run Info** button.
30
+
- Note how this opens up a view into the traces for agent execution
31
+
- Explore the waterfall diagram - see how each step provides token counts & time
32
+
- Click on a _Run step_ and a _Tool_ item - how is the output different?
33
+
- Click the _Evaluations_ tab - what does it tell us about quality & safety?
34
+
35
+
36
+
37
+
- Close the tracing view - can you see the same token counts and latency metrics in the main view?
38
+
- Hover over one of the numbers e.g., tokens - see the breakdown in token costs
0 commit comments