Summary
Earlier versions of Foundry Local always called the Foundry model catalog at the eastus endpoint. As usage has grown this single region has started to throttle requests, surfacing as catalog/list/download failures for users. You will see something like the following error message:
foundry_local_sdk.exception.FoundryLocalException: Failed to get model list: Error: System.Net.Http.HttpRequestException: Response status code does not indicate success: 429 (Received too many requests in a short amount of time. Retry again after 1 seconds.).
at System.Net.Http.HttpResponseMessage.EnsureSuccessStatusCode() + 0xf8
at Microsoft.Neutron.AzureFoundry.AzureFoundryApiService.<FetchModelPageFromApiAsync>d__13.MoveNext() + 0x214
--- End of stack trace from previous location ---
at Microsoft.Neutron.AzureFoundry.AzureFoundryApiService.<FetchModelsAsyncImpl>d__18.MoveNext() + 0x1d0
--- End of stack trace from previous location ---
at Microsoft.Neutron.AzureFoundry.AzureFoundryApiService.<FetchAllLocalModelsAsyncImpl>d__16.MoveNext() + 0x5c
--- End of stack trace from previous location ---
at Microsoft.Neutron.AzureFoundry.AzureFoundryApiService.<FetchAllLocalModelsAsync>d__14.MoveNext() + 0x5c
--- End of stack trace from previous location ---
at Microsoft.Neutron.AzureFoundry.AzureFoundryService.<ListCatalogModelsAsync>d__20.MoveNext() + 0xb04
--- End of stack trace from previous location ---
at Microsoft.AI.Foundry.Local.AzureModelCatalog.<GetLatestCatalogAndCachedModelInfo>d__29.MoveNext() + 0x184
--- End of stack trace from previous location ---
at Microsoft.AI.Foundry.Local.AzureModelCatalog.<FetchModelInfoAsync>d__25.MoveNext() + 0x64
--- End of stack trace from previous location ---
at Microsoft.AI.Foundry.Local.BaseModelCatalog`1.<GetCachedInfoAsync>d__20.MoveNext() + 0x1ac
--- End of stack trace from previous location ---
at Microsoft.AI.Foundry.Local.BaseModelCatalog`1.<GetModelsAsync>d__8.MoveNext() + 0x68
--- End of stack trace from previous location ---
at Microsoft.AI.Foundry.Local.NativeInterop.<>c__DisplayClass10_1.<<ExecuteCommandManaged>b__0>d.MoveNext() + 0x74
--- End of stack trace from previous location ---
at Microsoft.AI.Foundry.Local.NativeInterop.<ExecuteWithTracker>d__9.MoveNext() + 0xa8
Fix
Foundry Local 1.2.0 introduces automatic region detection. Traffic is now spread across regions rather than be concentrated on eastus.
Recommended action: upgrade to Foundry Local 1.2.0 or later.
Still seeing errors after upgrading?
You can pin a specific region in the Configuration using:
config = Configuration(
app_name="appname",
additional_settings={"ModelRegistryRegion": "westus"},
log_level="Debug"
)
- Valid values: any Azure region slug (lowercase letters, digits, hyphens), e.g. westus2, westeurope, southeastasia.
- Default is "auto" (region detection as described above).
- Setting an explicit region skips detection and sends all catalog requests directly to that region.
Valid Azure regions
Here are a list of valid Azure regions to try:
"eastus", "australiaeast", "australiasoutheast", "brazilsouth",
"canadacentral", "canadaeast", "centralindia", "centralus",
"eastasia", "eastus2", "francecentral", "germanywestcentral",
"japaneast", "japanwest", "jioindiawest", "koreacentral",
"northcentralus", "northeurope", "norwayeast", "southafricanorth",
"southcentralus", "southeastasia", "swedencentral", "switzerlandnorth",
"uaenorth", "uksouth", "ukwest", "westcentralus", "westeurope",
"westus", "westus2", "westus3", "qatarcentral", "polandcentral",
"southindia", "switzerlandwest", "italynorth", "spaincentral",
"israelcentral", "taiwannorth", "malaysiawest"
If problems persist after trying a different region, please reply with:
- Foundry Local version
- The ModelRegistryRegion value you tried
- A snippet of the log including the Created AzureModelCatalog line and any FetchModelPageFromApiAsync / catalog error lines
Summary
Earlier versions of Foundry Local always called the Foundry model catalog at the eastus endpoint. As usage has grown this single region has started to throttle requests, surfacing as catalog/list/download failures for users. You will see something like the following error message:
Fix
Foundry Local 1.2.0 introduces automatic region detection. Traffic is now spread across regions rather than be concentrated on eastus.
Recommended action: upgrade to Foundry Local 1.2.0 or later.
Still seeing errors after upgrading?
You can pin a specific region in the
Configurationusing:Valid Azure regions
Here are a list of valid Azure regions to try:
If problems persist after trying a different region, please reply with: