Skip to content

README Missing Local Testing Steps for Vercel AI SDK Node.js Sample #242

@smita-MSFT

Description

@smita-MSFT

Description

The README only states:

"To set up and test this agent, refer to the Configure Agent Testing guide for complete instructions."
However, the README does not include the required steps for configuring agentic authentication or ensuring that the agentic identity is provisioned during setup.
After following the documented steps — installing dependencies, configuring environment variables, running a365 config init, and executing a365 setup all — the agent fails to respond when tested through Agents Playground.
During local testing, authentication failures occur and the agent cannot process messages.
Inspection of a365.generated.config.json shows that both AgenticAppId and AgenticUserId remain null, indicating that the agentic identity was never provisioned.
No error or warning is displayed during a365 setup all to indicate that this identity setup is incomplete.

Expected behavior

After completing the documented setup steps for the Vercel AI SDK Node.js sample agent, local testing using Agents Playground should work successfully.

Specifically:

Running a365 setup all should provision the required agentic identity automatically.

The generated configuration file (a365.generated.config.json) should contain valid values for:

AgenticAppId
AgenticUserId

When executing:

npm run test-tool

Agents Playground should open successfully and allow sending messages to the agent.

When a message is sent (for example: "hi"):

The agent should receive the request from Agents Playground.

Authentication should complete successfully.

The request should be forwarded to the agent backend.

The agent should process the message using the configured LLM.

The agent should return an AI-generated response in the Agents Playground chat interface.

SDK Version

Node.js v22.14.0

Language/Runtime

Node.js v22.14.0

OS

windows 11

How to Reproduce

Clone the Agent365-Samples repository and navigate to the Vercel AI SDK Node.js sample agent.

Install dependencies.

npm install

Initialize the Agent 365 configuration.

a365 config init

Configure the .env file with the required settings such as:

Azure OpenAI credentials

Anthropic API key

Agentic authentication enabled

Example:

USE_AGENTIC_AUTH=true
ANTHROPIC_API_KEY=

Run the full setup command.

a365 setup all

The CLI completes successfully and reports that the infrastructure and agent blueprint were created.

Verify the generated configuration file.

Open:

a365.generated.config.json

Observe that the following fields remain null:

"AgenticAppId": null
"AgenticUserId": null

Run the local testing command.

npm run test-tool

Agents Playground launches at:

http://localhost:56150

Send a test message in the playground (for example: "hi").

Output

Agents Playground fails to communicate with the agent and displays an authentication error.

Example error shown in the playground:

Error communicating with agent: Failed to acquire token. User may need to sign in.

The agent does not respond to messages sent from the playground.

Additionally, the generated configuration file still contains:

AgenticAppId: null
AgenticUserId: null

Screenshots

No response

Code of Conduct

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions