Share via

Downstream_error: "Error creating item on conversation" with empty item_id in Foundry Workflow Preview

Pouneh Samavatian 0 Reputation points
2026-05-22T11:43:27.9566667+00:00

Hi community,

I am building a multi-agent workflow in Microsoft Foundry (New) and getting a persistent downstream_error every time I test it in Preview. No matter what changes I make to the workflow, the error remains identical.

**Error message:
**downstream_error: {"error":"Error creating item on conversation", "conversation_id":"conv_NUMBER", "item_id":"", "details":""}
**Workflow structure:
**User's image

Environment:

  • Region: Sweden Central
  • Foundry version: New Foundry (toggle ON)
  • Workflow type: Multi-agent Group Chat workflow
  • Tested: May 2026

Has anyone resolved this? Is this a known platform bug in the Sweden Central region?
Any help appreciated.

Foundry Agent Service
Foundry Agent Service

A fully managed platform in Microsoft Foundry for hosting, scaling, and securing AI agents built with any supported framework or model

0 comments No comments

1 answer

Sort by: Most helpful
  1. AI answer

    2026-05-22T11:43:45.3566667+00:00

    The error message indicates a failure when creating a conversation item in the underlying Foundry conversations API. In workflows, each agent call and message is stored as a conversation item; if item creation fails, the runtime surfaces a generic downstream error like the one shown.

    Based on the available information, focus on these areas:

    1. Verify every agent node is fully configured
    • In the workflow visualizer, open each Agent node (Agent-language-intake, Agent-Analyzer, Agent-Recommender, etc.).
    • Confirm that each node has a valid agent selected and saved.
    • If any node is missing an agent or references a deleted/invalid agent, item creation for that step can fail and cause a downstream error.
    1. Check saved outputs and JSON schemas
    • For nodes that save outputs (for example, capturing agent responses into variables), verify that the saved output schema is valid JSON and matches what the agent actually returns.
    • Invalid or mismatched schemas can cause the workflow run to produce unexpected output or fail.
    1. Validate Power Fx expressions in the if/else node
    • In the If / else condition node, ensure the condition uses correct variable scopes and types.
    • Use System. for system variables and Local. for local variables. For example, if Local.Var_AgentLanguage is a local variable, the expression should use Local. as shown in the screenshot.
    • If there are Power Fx errors like "Name isn't valid" or "Type mismatch", fix them by adding the correct scope prefix or using conversion functions such as Text() or Value().
    1. Confirm workflow changes are saved
    • After editing nodes or conditions, select Save in the visualizer. Foundry does not auto-save; unsaved changes can cause the preview to keep running an older, possibly invalid version of the workflow.
    1. Check role and access
    • Ensure the account used in Sweden Central has at least the Contributor role on the Foundry project. Missing permissions can prevent workflow resources (including conversations) from being created correctly.

    If, after validating agent assignments, JSON schemas, Power Fx expressions, and permissions, the same downstream error persists in Preview, capture the conversation ID from the error and raise a support ticket so the platform team can inspect the failing conversation item creation in that specific region.


    References:

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

    Was this answer helpful?

Your answer

Answers can be marked as 'Accepted' by the question author and 'Recommended' by moderators, which helps users know the answer solved the author's problem.