Azure AI services
A group of Azure services, SDKs, and APIs designed to make apps more intelligent, engaging, and discoverable.
3,632 questions
Sudden Persistent TypeError in Azure AI Foundry (UI)
![](data:image/svg+xml, %3Csvg xmlns='http://www.w3.org/2000/svg' height='64' class='font-weight-bold' style='font: 600 30.11764705882353px "SegoeUI", Arial' width='64'%3E%3Ccircle fill='hsl(243.2, 16%, 33%)' cx='32' cy='32' r='32' /%3E%3Ctext x='50%25' y='55%25' dominant-baseline='middle' text-anchor='middle' fill='%23FFF' %3EFA%3C/text%3E%3C/svg%3E)
Fatima Abdillahi
0
Reputation points
Hello,
I'm experiencing an unexpected TypeError with my Azure AI agents that began a few days ago. My established workflow, which was previously functioning without issues, involves two agents:
Agent 1: Handles structuring data into a JSON object. Agent 2: Receives this JSON data from Agent 1, performs research using a connected Logic App tool, and summarizes the findings.
The error arises when Agent 1 attempts to pass the JSON data to Agent 2. Here's a sample of the data being sent to Agent 2:
enrich_data: [{"name": "Toronto Restaurant Market", "website": "https://torontorestaurantforsale.com/toronto-restaurant-market/", "address": "Toronto, Ontario, Canada"}]
Agent 2's prompt is designed to summarize this incoming data:
If you received an array from the orchestrator agent,
- summarize the data into 1 sentence
- return the summary back to the orchestrator agent
Notably, the TypeError occurs even when I've stripped down Agent 2's logic to simply summarize and return the data, removing any tool integration. This indicates the error is happening at the input reception stage, before actual data processing.
Given that this workflow was stable until recently, I'm concerned there might be a new bug affecting Azure AI Foundry agents. Any insights or assistance would be greatly appreciated.
Azure AI services
{count} votes
-
Jerald Felix 2,180 Reputation points2025-06-29T07:34:51.3833333+00:00 Hi Fatima,
A “
TypeError” at the very first hand-off between two agents almost always means the runtime can’t find, or can’t coerce, one of the input parameters that Agent 2 is expecting. Since 26-27 June a service update to Azure AI Foundry’s Agent Service began strictly validating the input-parameter schema; variables that used to be accepted implicitly must now be declared explicitly. Here’s the quickest way to get your flow working again.1 Add the missing input parameter to Agent 2
- Open Agent 2 in the portal.
Select Parameters & schema ▶ Input parameters.
Add a new parameter
Name Type Item-type Required enrich_dataarrayobject✔︎ Save & Publish Agent 2.
(If you prefer to accept any structure while you debug, set the type to
anyfirst, then tighten it later.)Why it matters
The latest runtime validates every key in the JSON passed to an agent against the declared schema. If the key isn’t present you’ll see a TypeError before the prompt even runs. ([learn.microsoft.com](https://learn.microsoft.com/en-us/azure/ai-foundry/how-to/flow-develop"How to build with prompt flow - Azure AI Foundry | Microsoft Learn"))2 Confirm Agent 1 is emitting valid JSON
Make sure Agent 1 wraps the array in a JSON object that matches the schema:
{ "enrich_data": [ { "name": "Toronto Restaurant Market", "website": "https://torontorestaurantforsale.com/toronto-restaurant-market/", "address": "Toronto, Ontario, Canada" } ] }Avoid the YAML-style
enrich_data: [...]string you showed in the post—it can be parsed as a plain string.
If you’re building Agent 1 in a Prompt Flow, usejson.dumps()or the portal’s Structured output helper so the payload is guaranteed to be legal JSON. ([learn.microsoft.com](https://learn.microsoft.com/en-us/azure/ai-services/openai/how-to/structured-outputs?utm_source=chatgpt.com"How to use structured outputs with Azure OpenAI ... - Learn Microsoft"))3 Update the hand-off mapping (or Orchestrator)
If you are invoking Agent 2 via:
agent_client.invoke( agent_id=AGENT2_ID, inputs=agent1_output # <-- must be a dict, not raw list )double-check that
inputs=receives the object above, not the bare array. Passing a list causesTypeError: list indices must be integers or slices, not str.4 Test with a clean run
In the playground, select Conversation ▶ Reset (this clears any cached state).
Trigger the flow again.
If it still fails, open Developer tools ▶ Network and copy the GUID in the failing
/invocationscall—Azure support will ask for this.5 Optional hardening tips
Goal Technique Guarantee well-formed JSON Enable JSON mode on the LLM call or wrap with Structured Outputs. ([learn.microsoft.com](https://learn.microsoft.com/en-us/azure/ai-services/openai/how-to/json-mode?utm_source=chatgpt.com"Learn how to use JSON mode - Azure OpenAI")) Catch schema issues early Add a tiny Python tool in Agent 2 that simply does json.loads()on the incoming blob; the trace view will show you exactly what the runtime received. ([learn.microsoft.com](https://learn.microsoft.com/en-us/azure/ai-foundry/how-to/develop/trace-application?utm_source=chatgpt.com"How to trace AI applications using OpenAI SDK - Azure AI Foundry"))Prevent future breaking changes Keep an Input/Output contract document for each agent and version control it; re-run your contract tests after each “What’s new” notice from the Foundry team. 6 When to escalate
If Agent 2 already has
enrich_datadeclared and the payload matches the schema but you still get a TypeError, open an Azure support ticket and attach:Subscription + Region
Agent IDs and version
UTC timestamp of the failure
Correlation ID from the failing request
This lets engineering pull server-side logs and confirm whether a regression was introduced in the weekend rollout.
Give the above a try adding the explicit
enrich_dataparameter fixes the vast majority of sudden TypeErrors we’ve seen since the 27 June update. Let me know how it goes!Best Regards,
Jerald Felix
-
Fatima Abdillahi 0 Reputation points
2025-06-30T04:46:51.5233333+00:00 Thanks for your reply Jerald. I’ll definitely try these out and let you know how it goes.
-
Fatima Abdillahi 0 Reputation points
2025-06-30T20:59:25.0733333+00:00 Hi Jerald,
I can't find Select Parameters & schema ▶ Input parameter in the foundry portal. I've checked the project settings and the agent settings as well. I also can't find any relevant documentation online.
Can you please advise?
Note: I'm not using an API, just the the foundry portal. And my project type is
Project (Default)(AI Foundry) -
Manas Mohanty 6,115 Reputation points Microsoft External Staff Moderator2025-07-02T05:02:20.5233333+00:00 Sorry for the late response.
Could you share the full error trace of type error. Shall try to connect with product group post you share the full error trace, and a minimal replication is done at myside.
Thank you.
-
Manas Mohanty 6,115 Reputation points Microsoft External Staff Moderator
2025-07-04T05:40:31.6233333+00:00 It would have been great for debugging if you have shared more error details.
Went through the case description earlier. Even you if remove the prompt in child agent mentioning to parse the array (getting Json inputs from parent agent), issue persisted for you.
Might have been delay in changes of connected agent. You can remove and connect a new agent with updated instruction in that case
Could you use the below prompt in child agent
"Summarize the input received and return the summary "
Please let us know if connecting a new agent and using updated prompt fixes the issue.
Thank you.
-
Manas Mohanty 6,115 Reputation points Microsoft External Staff Moderator
2025-07-07T03:28:12.0966667+00:00 Please let us know if you can accept the above answer or the issue still persists for you with new agent and updating prompt instruction
(looks like stricter schema validation implemented in new versions)
Thank you.
Sign in to comment
Sign in to answer