my foundry agent give error when i ask to draw chart?

Hi @Ashish Kumar Jha welcome to the Microsoft Q&A community.

Yes the behavior you describe strongly points to a Code Interpreter / chart-generation path problem, rather than a general agent or model problem.

Microsoft's current Foundry documentation explicitly says Code Interpreter is the component responsible for running Python and generating charts.https://learn.microsoft.com/th-th/azure/foundry/agents/how-to/tools/code-interpreter?view=foundry&utm_source=chatgpt.com&pivots=python

What I would check first

Because ordinary questions work but a chart request produces HTTP 500, isolate the failure in this order:

1. Ask for a calculation
   • "Calculate 12345 * 6789 using Python."
2. Ask it to execute Python without creating a file
   • "Use Python to calculate the mean of [1,2,3,4,5]."
3. Ask for a chart without an uploaded file
   • "Use Python to create a simple matplotlib line chart for x=[1,2,3], y=[2,4,6]. Save it as chart.png."
4. Then test with your actual CSV/XLSX input. If #1/#2 work but #3 fails, you've narrowed it specifically to Code Interpreter's graphical/file-output path.

There is precedent for exactly this issue

Microsoft Q&A has reports of Azure AI Foundry returning server_error specifically when code_interpreter generates graphs. One report describes calculations working while graph generation fails, including cases where an image ID is produced but the corresponding file isn't available.

So I would not immediately rewrite your agent code based only on this 500.

Also check your model/deployment

Microsoft recommends checking the current regional and model availability for Code Interpreter because support varies by model/region.https://learn.microsoft.com/th-th/azure/foundry/agents/how-to/tools/code-interpreter?view=foundry&utm_source=chatgpt.com&pivots=python

If you're using something like gpt-4.1-mini, there's an especially interesting precedent: a Microsoft Q&A report found that Code Interpreter calculations worked but graph generation failed, while gpt-4.1 worked.https://learn.microsoft.com/en-us/answers/questions/5641829/server-error-occurs-in-azure-ai-foundry-when-gener?utm_source=chatgpt.com

So, as a diagnostic, try the same agent/tool configuration with another supported model deployment.

Verify your Code Interpreter configuration

For the current Foundry Agents API, the configuration should conceptually contain a Code Interpreter tool, e.g.:

"tools": [
  {
    "type": "code_interpreter",
    "container": {
      "type": "auto"
    }
  }
]

If you're using a file, the container can include the uploaded file IDs. Microsoft's current REST example uses this pattern for a CSV → Python → PNG workflow.https://learn.microsoft.com/bs-latn-ba/azure/foundry/agents/how-to/tools/code-interpreter?utm_source=chatgpt.com&pivots=python

The expected workflow is:

user prompt → agent → Code Interpreter → Python/matplotlib → generated file → response/file citation

If the request dies with HTTP 500 before you get a normal tool result, that is useful evidence that the failure is occurring on the Azure/Foundry service side rather than being a Python exception in your application.

One important distinction

If your error is literally:

openai.InternalServerError: Error code: 500
{
    'error': {
        'message': 'The server had an error processing your request...',
        'type': 'server_error'
    }
}

then this is not the same as Python code inside Code Interpreter throwing an exception.

A Python exception would normally be returned as part of the agent/tool execution, allowing the agent to potentially retry or correct the code. Code Interpreter is specifically designed to iteratively modify and rerun failed Python code.https://learn.microsoft.com/en-us/azure/foundry-classic/agents/how-to/tools-classic/code-interpreter?view=foundry-classic&utm_source=chatgpt.com&pivots=python

Your error is occurring at the API request level.

What I'd need to diagnose your case precisely

Please paste these 4 things (remove your endpoint/key/project IDs):

1. Model/deployment:
   e.g. gpt-4.1 / gpt-4.1-mini / ...

2. Foundry SDK/package versions:
   e.g. openai==...
        azure-ai-projects==...
        azure-ai-agents==...

3. Agent creation code:
   The part where you create/configure the agent and Code Interpreter.

4. Request/run code:
   The part where you send the user prompt to the agent.

Also tell me whether this works:

"Use Python to calculate 2 + 2"

but this fails:

"Use Python to create a simple bar chart of [10,20,30]."

With those details, I can pinpoint whether you're dealing with model compatibility, Foundry API/SDK version mismatch, Code Interpreter configuration, generated-file handling, or an Azure-side service issue.

N/B: I have generated the above answer using co-pilot as an AI tool. Also I have validated and updated the AI output.