Azure AI Foundry SDK File Attachment Issue

Problem Summary

We are unable to attach files to Azure AI Foundry agents using the Python SDK (`azure-ai-projects` and `azure-ai-agents`). **Every approach to file attachment results in a generic HTTP 500 error** from the Azure API:

```

HttpResponseError: (None) The server had an error processing your request.

Sorry about that! You can retry your request, or us through our help

center at oai-assistants@ if you keep seeing this error.

```

This occurs consistently across:

- Code Interpreter file attachments

- Vision/image file references

- All documented attachment patterns from Microsoft's official documentation

Environment Details

- **SDK Versions:**

- `azure-ai-projects`: 1.0.0

- `azure-ai-agents`: 1.1.0

- **Python**: 3.12

- **Azure Region**: East US 2

- **Model**: gpt-4o, gpt-5-mini

- **File Upload**: Files upload successfully to Azure AI Foundry (status: "processed")

- **Agent Creation**: Agents create successfully without file attachments

What We've Tried

Files Upload Successfully But Cannot Be Connected to Agents

While file upload and agent creation work individually, **we cannot link files to agents** - every attempt to connect them fails with HTTP 500:

# 1. File upload - No errors, but agent cannot access the file
with open("data.csv", 'rb') as f:
    file_response = openai_client.files.create(file=f, purpose="assistants")
# Result: File ID returned, file status = "processed"
# Problem: File exists on Azure but no way to connect it to an agent

# 2. Agent creation without files - No errors, but cannot attach files later
agent = project_client.agents.create_agent(
    model="gpt-4o",
    name="test-agent",
    instructions="You are helpful",
    tools=[CodeInterpreterToolDefinition()]
)
# Result: Agent created successfully
# Problem: Agent exists but cannot access any files

❌ All Attempts to Connect Files to Agents Fail (HTTP 500)

Attempt 1: Message Attachments (per Microsoft docs)

from azure.ai.agents.models import MessageAttachment, CodeInterpreterToolDefinition

attachment = MessageAttachment(
    file_id=file_id,
    tools=[CodeInterpreterToolDefinition()]
)

message = project_client.agents.messages.create(
    thread_id=thread_id,
    role="user",
    content="Analyze this file",
    attachments=[attachment]  # ❌ HTTP 500
)

**Request ID**: `649ec8ad5b8941804068be4f41ccb680`

Attempt 2: Thread Creation with Tool Resources

from azure.ai.agents.models import ToolResources, CodeInterpreterToolResource

tool_resources = ToolResources(
    code_interpreter=CodeInterpreterToolResource(file_ids=[file_id])
)

thread = project_client.agents.threads.create(
    tool_resources=tool_resources  # ❌ HTTP 500
)

**Request ID**: `17501cc0daa8e0a4a7bcc95945231b10`

Attempt 3: Thread Update with Tool Resources

project_client.agents.threads.update(
    thread_id=thread_id,
    tool_resources=tool_resources  # ❌ HTTP 500
)

**Request ID**: `418f1aeedb94b55504fcee444ae471ee`

Attempt 4: Agent Creation with Tool Resources (Microsoft's recommended approach)

Following the official documentation at: https://learn.microsoft.com/en-us/azure/ai-foundry/agents/how-to/tools/code-interpreter-samples

from azure.ai.agents.models import (
    ToolResources,
    CodeInterpreterToolResource,
    CodeInterpreterToolDefinition
)

# Upload file
file = openai_client.files.create(file=f, purpose="assistants")

# Create tool resources
tool_resources = ToolResources(
    code_interpreter=CodeInterpreterToolResource(file_ids=[file.id])
)

# Create agent with files
agent = project_client.agents.create_agent(
    model="gpt-4o",
    name="test-agent",
    instructions="You are helpful",
    tools=[CodeInterpreterToolDefinition()],
    tool_resources=tool_resources  # ❌ HTTP 500
)

**Request ID**: `c2e1a86cfa0dd09d4448092a3139dae5`

Attempt 5: Vision with Image File Content (no code_interpreter)

message_content = [
    {
        "type": "text",
        "text": "Please describe this image."
    },
    {
        "type": "image_file",
        "image_file": {
            "file_id": file_id
        }
    }
]

message = project_client.agents.messages.create(
    thread_id=thread_id,
    role="user",
    content=message_content  # ❌ HTTP 500
)

**Request ID**: `401e53a5802d1e203e6fe92d74ad59f1`

Summary Table

| Approach | API Method | Code Interpreter | Result |

|----------|-----------|------------------|--------|

| Message attachments | `messages.create(attachments=[...])` | Yes | ❌ HTTP 500 |

| Thread creation | `threads.create(tool_resources=...)` | Yes | ❌ HTTP 500 |

| Thread update | `threads.update(tool_resources=...)` | Yes | ❌ HTTP 500 |

| Agent creation | `agents.create_agent(tool_resources=...)` | Yes | ❌ HTTP 500 |

| Vision image_file | `messages.create(content=[image_file])` | No | ❌ HTTP 500 |

Questions:

1. **Has anyone successfully attached files to Azure AI Foundry agents using the Python SDK?**
2. **Is there a specific Azure configuration, permission, or feature flag required for file attachments?**
3. **Are file attachments only available in certain Azure regions or subscription tiers?**
4. **Is this a known issue with the current SDK version?**
5. **Are there any alternative approaches we haven't tried?**

Documentation followed: https://learn.microsoft.com/en-us/azure/ai-foundry/agents/how-to/tools/code-interpreter-samples?pivots=python

Yes this was AI written - but human validated!