In Microsoft Foundry Workflow

Question

In Microsoft Foundry Workflow

Suraj Panker 15

When i am runing workflow then i grtting below error

downstream_error{"error":"Error creating item on conversation","conversation_id":"conv_719a6b12d5ab869a00AY5r6HNCOrLq2SEp7g2dSu8GW9V8Ygjf","item_id":"","details":""}

Anshika Varshney 11,585 Reputation points Microsoft External Staff Moderator

2026-05-22T08:00:32.52+00:00

Hi @Suraj Panker

Thanks for sharing the details. This error usually means the workflow is failing when it tries to add data to an existing conversation.

To troubleshoot this, you can check a few things step by step.

First, verify the conversation id Make sure the conversation id used in your workflow is valid and still exists. You can try retrieving it using the API. If it is not found, the workflow will fail when trying to add items.

Second, check how the item is being created. The workflow step that adds data to the conversation should include a valid items array in the request body. If the items field is missing or empty, the request will fail.

Third, confirm you are using the correct endpoint Make sure the workflow is using the correct Foundry project endpoint consistently for creating the conversation adding items retrieving data

If there is a mismatch, it can result in this error.

Fourth, simplify the request If your workflow includes extra parameters, try removing them and send only the required fields. This helps confirm whether the issue is caused by request format or payload.

Fifth, test outside the workflow Try running the same request manually using a REST client. This helps you see the full error response, as sometimes workflows do not show detailed error messages.

Sixth, try with a new conversation If the issue continues only for one conversation id, create a new conversation and test again. Sometimes the existing conversation might be in an invalid state.

This type of error generally happens due to invalid conversation id, incorrect request format, or endpoint mismatch.

For more details, you can also refer to the official workflow concept here Microsoft Foundry workflow documentation

I hope this helps. Let me know if you need help checking your workflow configuration.
Thankyou!
Suraj Panker 15 Reputation points

2026-05-22T08:05:08.5566667+00:00
@Anshika Varshney this is actually problem at the platform level, not any issue during workflow runtime
Parisi, Roberta 0 Reputation points

2026-05-22T11:15:18.41+00:00

I am having this issue as well.

In my case, the same workflow version was working correctly around 4 days ago, and no recent changes were made to any part of the pipeline, configuration, agents, or prompts.

I also tried creating a completely new workflow containing only a simple Set Variable node, but it failed with the same error message. This makes it unlikely that the issue is related to a specific conversation ID, payload format, agent configuration, or endpoint mismatch.

Additionally, I tested the individual agents used in my original workflow separately, and they work perfectly. The failure seems to happen only when running them through the Workflows section in Microsoft Foundry.

Based on this, it looks more like a platform-level issue with Foundry Workflows rather than a problem in the workflow logic itself.
Anshika Varshney 11,585 Reputation points Microsoft External Staff Moderator

2026-05-24T08:59:04.1366667+00:00
Hi @Suraj Panker

Thanks for your update. Understood if you suspect this is a platform-level issue, it is still important to clearly isolate whether the problem is coming from the workflow configuration or the backend service.

Here are some practical troubleshooting steps you can follow to confirm this.

Step 1: Test the same API call outside the workflow

Try calling the same Conversations API directly using a REST client.

Use the same endpoint, conversation id, and request body

If it fails the same way, it strongly indicates a platform issue

If it works, then the issue is likely in the workflow configuration

Step 2: Validate with a new conversation

Create a fresh conversation and update your workflow to use the new id.

If the new conversation works, the issue may be with the old conversation state

If both fail, it points more toward a platform-level behavior

Step 3: Check request payload from workflow

Review the step where the workflow creates the conversation item.

Make sure:

items array is present and not empty

structure matches expected schema

no null or invalid fields

Sometimes workflow transformations may silently break the payload.

Step 4: Compare workflow vs direct request

Log or capture the actual request sent by the workflow and compare it with your manual API call.

Check for differences in:

request body

headers

endpoint path

API version

Even small differences can cause failures.

Step 5: Test minimal request

Simplify the request to the minimum required fields.

Remove optional parameters

Send only basic items content

If minimal request works but full request fails, then the issue is payload-related, not platform.

Step 6: Check consistency across runs

Run the workflow multiple times.

If failures are random or intermittent, this may indicate backend instability

If failures are consistent, then it is either configuration or a systematic platform issue

Step 7: Test across multiple workflows

Try the same logic in another workflow or a simple test workflow.

If all workflows fail similarly, it supports platform-level issue

If only one workflow fails, it is likely configuration-specific

Step 8: Check timing and retry behavior

Sometimes backend services may fail temporarily.

Retry the same workflow after some time

If it succeeds later without changes, it may be a transient platform issue

Step 9: Capture request details

If issue still persists, collect:

timestamp of failure

request or correlation id if available

exact request payload

This helps confirm whether it is platform-side behavior.

You can conclude it is a platform-level issue if:

Same request fails outside workflow

New conversations also fail

Issue is reproducible across workflows

Payload and configuration are validated

Otherwise, it is usually related to workflow setup or request formatting.

I hope this helps you clearly isolate the root cause. Do let me know if you have any further queries.

Thankyou!
Anshika Varshney 11,585 Reputation points Microsoft External Staff Moderator

2026-05-25T17:33:54.35+00:00

Hi @Suraj Panker

Did you get any chance to review the response.

Thankyou!

1 answer

Your answer

Anshika Varshney 11,585 Reputation points Microsoft External Staff Moderator

2026-05-22T08:00:32.52+00:00

Hi @Suraj Panker

Thanks for sharing the details. This error usually means the workflow is failing when it tries to add data to an existing conversation.

To troubleshoot this, you can check a few things step by step.

First, verify the conversation id Make sure the conversation id used in your workflow is valid and still exists. You can try retrieving it using the API. If it is not found, the workflow will fail when trying to add items.

Second, check how the item is being created. The workflow step that adds data to the conversation should include a valid items array in the request body. If the items field is missing or empty, the request will fail.

Third, confirm you are using the correct endpoint Make sure the workflow is using the correct Foundry project endpoint consistently for creating the conversation adding items retrieving data

If there is a mismatch, it can result in this error.

Fourth, simplify the request If your workflow includes extra parameters, try removing them and send only the required fields. This helps confirm whether the issue is caused by request format or payload.

Fifth, test outside the workflow Try running the same request manually using a REST client. This helps you see the full error response, as sometimes workflows do not show detailed error messages.

Sixth, try with a new conversation If the issue continues only for one conversation id, create a new conversation and test again. Sometimes the existing conversation might be in an invalid state.

This type of error generally happens due to invalid conversation id, incorrect request format, or endpoint mismatch.

For more details, you can also refer to the official workflow concept here Microsoft Foundry workflow documentation

I hope this helps. Let me know if you need help checking your workflow configuration.
Thankyou!
Suraj Panker 15 Reputation points

2026-05-22T08:05:08.5566667+00:00

@Anshika Varshney this is actually problem at the platform level, not any issue during workflow runtime
Parisi, Roberta 0 Reputation points

2026-05-22T11:15:18.41+00:00

I am having this issue as well.

In my case, the same workflow version was working correctly around 4 days ago, and no recent changes were made to any part of the pipeline, configuration, agents, or prompts.

I also tried creating a completely new workflow containing only a simple Set Variable node, but it failed with the same error message. This makes it unlikely that the issue is related to a specific conversation ID, payload format, agent configuration, or endpoint mismatch.

Additionally, I tested the individual agents used in my original workflow separately, and they work perfectly. The failure seems to happen only when running them through the Workflows section in Microsoft Foundry.

Based on this, it looks more like a platform-level issue with Foundry Workflows rather than a problem in the workflow logic itself.
Anshika Varshney 11,585 Reputation points Microsoft External Staff Moderator

2026-05-25T17:33:54.35+00:00

Hi @Suraj Panker

Did you get any chance to review the response.

Thankyou!

Answer 1

AI answer

The error indicates that the workflow failed when trying to create an item in an existing conversation via the Foundry Conversations API.

To troubleshoot, validate each of the following against the workflow configuration and the API contract:

Verify the conversation ID
- Ensure conversation_id (conv_719a6b12d5ab869a00AY5r6HNCOrLq2SEp7g2dSu8GW9V8Ygjf in the error) actually exists by calling:
  - GET {endpoint}/openai/v1/conversations/{conversation_id}
- If this call does not return HTTP 200 with an OpenAI.ConversationResource, recreate the conversation first using:
  - POST {endpoint}/openai/v1/conversations
Check the item creation call
- The workflow step that adds to the conversation must call:
  - POST {endpoint}/openai/v1/conversations/{conversation_id}/items
- Required body:
```
     {
       "items": [
         { /* OpenAI.InputItem */ }
       ]
     }
```
- Confirm:
  - items is present and is a non‑empty array.
  - Each element matches the OpenAI.InputItem schema expected by the Foundry project (correct fields, types, and item role/type).
Confirm endpoint and project
- For Foundry projects, the endpoint must be of the form:
  - https://{ai-services-account-name}.services.ai.azure.com/api/projects/{project-name}
- Ensure the workflow is using the same project endpoint for:
  - Creating the conversation.
  - Creating conversation items.
  - Listing/retrieving items.
Validate query parameters
- If the workflow passes include or other query parameters, temporarily remove them and test with the minimal valid call:
  - POST {endpoint}/openai/v1/conversations/{conversation_id}/items with only the required items body.
- For listing or debugging items, use:
  - GET {endpoint}/openai/v1/conversations/{conversation_id}/items?limit=20&order=desc
Inspect API error details
- The workflow error shows details:"", which means the workflow did not surface the underlying API error.
- Run the same request outside the workflow (e.g., via a REST client) and check the full JSON error from the API (HTTP default response with ApiErrorResponse or generic object).
- Use the apim-request-id header from the failed call for deeper troubleshooting if needed.
Recreate the conversation if corrupted
- If item creation consistently fails on this conversation but works on a new one:
  - Create a new conversation with POST /openai/v1/conversations.
  - Point the workflow to the new conversation_id.

If, after these checks, item creation still fails, capture:

The exact HTTP request (method, URL, headers, and body) used to create the item.
The full API response body and status code. Then compare them against the documented Create conversation items and Get/List conversation items operations to identify any schema or parameter mismatch.

References:

AI-generated content may be incorrect. Read our transparency notes for more information.

Share via

In Microsoft Foundry Workflow

1 answer

Your answer