azureOpenAI only working sometimes (?)

[AqujZ1K9]

My configuration with Azure OpenAI works – sort of.

Basic queries, like "Who is the Pope?", take forever.

Sometimes I get a response, sometimes I get parts of the answer. In 90% of the cases, I get "Error connecting to server, try refreshing the page."

I have tested my deployment on Azure with other tools. Same API version, same API key, same deployment names, same resource name. With PSOpenAI, I get a response in about 1-2 seconds.

I'm slowly going crazy...

I'm on v0.7.6. using docker.
I'm using the orginal docker compose, with no changes...

what am I doing wrong ? :/

this is my librechat.yaml.
`---
version: 1.2.1
cache: true
modelSelect: true
parameters: true
sidePanel: true
presets: true
prompts: true
bookmarks: true
multiConvo: true
agents: true
endpoints:
azureAssistants:
capabilities:
- code_interpreter
- actions
- tools
pollIntervalMs: 750
timeoutMs: 180000

azureOpenAI:
assistants: true
titleConvo: true
titleModel: "gpt-4o-mini"
summarize: true
summaryModel: "gpt-4o-mini"
plugins: true
groups:
- group: ""
apiKey: ""
assistants: true
instanceName: ""
version: "2024-07-01-preview"
models:
gpt-4-vision-preview:
deploymentName: "gpt-4-vision-preview"
version: "2024-07-01-preview"
gpt-35-turbo:
deploymentName: "gpt-35-turbo"
version: "2024-07-01-preview"
gpt-4:
deploymentName: "gpt-4"
version: "2024-07-01-preview"
gpt-4-1106-preview:
deploymentName: "gpt-4-1106-preview"
version: "2024-07-01-preview"
gpt-4o:
deploymentName: "gpt-4o"
version: "2024-07-01-preview"
gpt-4o-mini:
deploymentName: "gpt-4o-mini"
version: "2024-07-01-preview"`

[AqujZ1K9]

Debug Log gives me something like this:

`LibreChat | 2025-01-09T22:45:11.274Z debug: [BaseClient] userMessage

LibreChat | {

LibreChat | messageId: "5cd2fd39-5c27-406e-bcdb-610107263135",

LibreChat | parentMessageId: "00000000-0000-0000-0000-000000000000",

LibreChat | conversationId: "0006587d-3d5f-4fee-bf69-a632dc3c5498",

LibreChat | sender: "User",

LibreChat | text: "whos the pope ?",

LibreChat | isCreatedByUser: true,

LibreChat | tokenCount: 9,

LibreChat | }

LibreChat | 2025-01-09T22:45:11.274Z debug: [OpenAIClient] chatCompletion

LibreChat | {

LibreChat | baseURL: "https://librechatopenai.openai.azure.com/openai/deployments/gpt-4o-mini/chat/completions?api-version...

[truncated]",

LibreChat | modelOptions.model: "gpt-4o-mini",

LibreChat | modelOptions.user: "659d4ae1ddc30c905751de17",

LibreChat | modelOptions.stream: true,

LibreChat | // 1 message(s)

LibreChat | modelOptions.messages: [{"role":"user","content":"whos the pope ?"}],

LibreChat | }

LibreChat | 2025-01-09T22:45:11.277Z debug: Making request to https://.openai.azure.com/openai/deployments/gpt-4o-mini/chat/completions?api-version=2024-07-01-preview

LibreChat | 2025-01-09T22:46:41.194Z debug: [AskController] Request closed

LibreChat | 2025-01-09T22:46:41.194Z debug: [AskController] Request aborted on close

LibreChat | 2025-01-09T22:46:41.194Z debug: [AskController] Request closed

LibreChat | 2025-01-09T22:46:41.195Z debug: [AskController] Request aborted on close

LibreChat | 2025-01-09T22:53:14.969Z debug: [AskController]`

[AqujZ1K9]

so this was a Problem with my deployment.
I'm running docker on a machine with an networkt mtu size of 1350.

after adding :

networks:
default:
driver: bridge
driver_opts:
com.docker.network.driver.mtu: 1350

to my docker compose override everything works... :)

[danny-avila]

Great! I was about to comment that it's a networking issue, but I'm glad you figured it out.