Custom Azure Endpoints not working anymore after LibreChat update

[asieverding]

What happened?

Hi, I have some custom endpoints with Azure Open AI models configured. I use custom endpoints for these models so I can change the icon to a custom one. However, I updated LibreChat from v0.7.8 to v0.8.0-rc2 and after the update the endpoints are not working anymore. I get the following error message, when I start to chat with the model/endpoint:

Something went wrong. Here's the specific error message we encountered: An error occurred while processing the request: 404 Resource not found Troubleshooting URL: https://js.langchain.com/docs/troubleshooting/errors/MODEL_NOT_FOUND/

The configuration has worked fine before the update. Here is the Part of my config:

custom:
- name: "Hermann-generative"
  apiKey: "${AZURE_OPENAI_API_KEY}"
  baseURL: "https://*****.openai.azure.com/openai/deployments/gpt-5-chat/chat/completions?api-version=2025-01-01-preview"
  iconURL: "/assets/hermannLogo.svg"
  models:
    default: ["gpt-5-chat"]
    fetch: false
    userIdQuery: false
  titleModel: "title-gpt"
  modelDisplayLabel: "Hermann"
  forcePrompt: false
  directEndpoint: true

Version Information

librechat latest 7c1b4cab81ab 39 minutes ago 1.17GB
ghcr.io/danny-avila/librechat-rag-api-dev-lite latest 2f812738600a 16 hours ago 1.67GB
ghcr.io/danny-avila/librechat-rag-api-dev-lite e7da8e6aca0f 3 months ago 1.31GB

Steps to Reproduce

1. Configure a custom endpoint with an Azure Open AI model with directEndpoint and api version in query string.
2. Run LibreChat

What browsers are you seeing the problem on?

Firefox, Chrome, Safari, Microsoft Edge, Mobile (iOS), Mobile (Android)

Relevant log output

Error logs are empty.

Debug logs:
2025-08-18T09:44:58.260Z debug: Web search serperApiKey: Using environment variable SERPER_API_KEY (not set in environment, user provided value)
2025-08-18T09:44:58.261Z debug: Web search searxngInstanceUrl: Using environment variable SEARXNG_INSTANCE_URL (not set in environment, user provided value)
2025-08-18T09:44:58.261Z debug: Web search searxngApiKey: Using environment variable SEARXNG_API_KEY (not set in environment, user provided value)
2025-08-18T09:44:58.261Z debug: Web search firecrawlApiKey: Using environment variable FIRECRAWL_API_KEY (not set in environment, user provided value)
2025-08-18T09:44:58.261Z debug: Web search firecrawlApiUrl: Using environment variable FIRECRAWL_API_URL (not set in environment, user provided value)
2025-08-18T09:44:58.261Z debug: Web search jinaApiKey: Using environment variable JINA_API_KEY (not set in environment, user provided value)
2025-08-18T09:44:58.262Z debug: Web search cohereApiKey: Using environment variable COHERE_API_KEY (not set in environment, user provided value)
2025-08-18T09:44:58.262Z warn: Default value for CREDS_KEY is being used.
2025-08-18T09:44:58.262Z warn: Default value for CREDS_IV is being used.
2025-08-18T09:44:58.263Z warn: Default value for JWT_SECRET is being used.
2025-08-18T09:44:58.263Z warn: Default value for JWT_REFRESH_SECRET is being used.
2025-08-18T09:44:58.263Z info: Please replace any default secret values.
2025-08-18T09:44:58.265Z info: ​

    For your convenience, use this tool to generate your own secret values:
    https://www.librechat.ai/toolkit/creds_generator

    ​
2025-08-18T09:44:58.266Z warn: The `CHECK_BALANCE` environment variable is deprecated. Please use the `balance` field in the `librechat.yaml` config file instead.
More info: https:/... [truncated]
2025-08-18T09:44:58.808Z info: RAG API is running and reachable at http://rag_api:8000.
2025-08-18T09:44:59.127Z info: Turnstile is DISABLED (no siteKey provided).
2025-08-18T09:44:59.127Z info: Outdated Config version: 1.1.5
Latest version: 1.2.8

      Check out the Config changelogs for the latest options and features added.

      https://... [truncated]
2025-08-18T10:57:16.654Z debug: [BaseClient] Loading history:
{
  conversationId: "7a05220a-0f79-4106-b1cc-c48badeacb10",
  parentMessageId: "00000000-0000-0000-0000-000000000000",
}
2025-08-18T10:57:16.932Z debug: [BaseClient] Context Count (1/2)
{
  remainingContextTokens: 359992,
  maxContextTokens: 360000,
}
2025-08-18T10:57:16.933Z debug: [BaseClient] Context Count (2/2)
{
  remainingContextTokens: 359992,
  maxContextTokens: 360000,
}
2025-08-18T10:57:16.933Z debug: [BaseClient] tokenCountMap:
{
  a7213d29-8750-44c2-9dd0-1dcddb5e8373: 5,
}
2025-08-18T10:57:16.935Z debug: [BaseClient]
{
  promptTokens: 8,
  remainingContextTokens: 359992,
  payloadSize: 1,
  maxContextTokens: 360000,
}
2025-08-18T10:57:16.936Z debug: [BaseClient] tokenCountMap
{
  a7213d29-8750-44c2-9dd0-1dcddb5e8373: 5,
}
2025-08-18T10:57:16.936Z debug: [BaseClient] userMessage
{
  messageId: "a7213d29-8750-44c2-9dd0-1dcddb5e8373",
  parentMessageId: "00000000-0000-0000-0000-000000000000",
  conversationId: "7a05220a-0f79-4106-b1cc-c48badeacb10",
  sender: "User",
  text: "test",
  isCreatedByUser: true,
  tokenCount: 5,
}
2025-08-18T10:57:18.856Z debug: [spendTokens] conversationId: 7a05220a-0f79-4106-b1cc-c48badeacb10 | Context: message | Token usage:
{
  promptTokens: 8,
  completionTokens: 4,
}
2025-08-18T10:57:19.148Z debug: [spendTokens] No transactions incurred against balance

Screenshots

No response

Code of Conduct

• I agree to follow this project's Code of Conduct

[asieverding]

Yes please! How can I get my custom endpoints with Azure Open AI models back to work?

[asieverding]

The model names and api versions in the backend (Azure AI Foundry) haven't changed at all. If I roll back to Version v0.7.8 of LibreChat, the custom endpoints work like a charm. But in v0.8.0-rc2 of LibreChat I get the 404 not found error. I can't find any solution for my problem in the ProblemMap.

[danny-avila]

It is recommended to use Azure Endpoint configuration and not custom endpoints for Azure.
See:

• https://www.librechat.ai/docs/configuration/azure
• Azure config help #9009 (comment)

On v0.7.8 it was never intended for azure to work via custom endpoints and there could be more than one reason why it's not working now.

[uhuynhnt]

I encountered the same problem after upgrading from v0.7.8 to v0.8.0-rc2.

After analysing the issue, I found that directEndpoint: true is not working correctly in the latest version.

Background: I use Azure API Management to centralize all requests to Azure AI Foundry and therefore using Custom Endpoint is a must.

Here is my config (setting directEndpoint: true)

version: 1.2.8

endpoints:
  custom:
    - name: 'APIM'
      apiKey: '${AZURE_OPENAI_API_KEY}'
      baseURL: 'https://***.azure-api.net/openai***/deployments/gpt-5/chat/completions?api-version=2025-01-01-preview'
      directEndpoint: true
      models:
        default: [ "gpt-5" ]
        fetch: false
      modelDisplayLabel: 'APIM'

Here is the actual URL that is being sent from LibreChat https://***.azure-api.net/openai***/deployments/gpt-5/chat/completions?api-version=2025-01-01-preview/chat/completions.

This of course leading to the 404 Model Not Found error because Azure AI Foundry cannot recognize the strange URL being sent.

[keepthegoal]

Having issue with directEndpoint: true as well

Here's my custom endpoint which I try to set up for Anthropic models in Bedrock via corporate proxy:

name: "AWS Bedrock Models' 
iconURL: "anthropic"
apiKey: "${AI_BEDROCK_GATEWAY_KEY}"
baseURL: 'https://bedrock-gateway-beta.apps.ai-dev.abc.cloud/invoke' 
models:
   default:
      "us-anthropic.claude-3-7-sonnet-20250219-v1:0"
      "us-anthropic.claude-3-5-sonnet-20241822-v2:0"
fetch: false 
titleConvo: true 
titleModel: 'title-endpoint' 
directEndpoint: true 
addParams:
# safe_prompt: true
    max_tokens: 2048
    anthropic_version: "bedrock-2023-05-31"
    model_id: "us.anthropic.claude-3-7-sonnet-20250219-v1:0" 
dropParams:
 - "model"
 - "stream" 
headers:
X-AI-USAGE-TRACKING-USER: "email={{LIBRECHAT_USER_EMAIL}}"
Authorization: "Bearer ${AI_BEDROCK_GATEWAY_KEY}"

With this configutation, I see the following url in request:
https://bedrock-gateway-beta.apps.ai-dev.abc.cloud/invoke/chat/completions

Actually I'm trying to simulate this request in custom endpoint:

headers = {"Content-Type": "application/json", "Authorization": "Bearer someApiKey"}
url = "https://bedrock-gateway-beta.apps.ai-dev.abc.cloud/invoke"
"messages": [{"role": "user", "content": "Hello, world"}l,
"max_tokens": 1024,
"anthropic_version": "bedrock-2023-05-31"
"model_id": "us-anthropic.claude-3-7-sonnet-20258219-v1:0"

Unfortunately, I can not use pre configured provider as described here:
https://www.librechat.ai/docs/configuration/pre_configured_ai/bedrock
So the only option is using custom endpoint

Working with librechat:0.7.9 image

[danny-avila]

I noticed from a few configurations being shared here and on discord, that directEndpoint is being used but was not implemented in the streamlining update to agents framework. I'm looking to see if we can still easily support what this field did in the past.

[uhuynhnt]

Yes, please. I see the Custom Endpoint feature one of the most powerful aspects of LibreChat, as it enables seamless integration of custom agents and endpoints. I currently maintain around 10 configurations, which worked very well prior to the recent upgrade.

[uhuynhnt]

FYI the directEndpoint is working again for me with latest version v0.8.0-rc3 and config version 1.2.9

[asieverding]

It's working for me also again :)