Wrong model selected by OpenAIClient?

[Originalimoc]

What happened?

modelOptions.model get mismatched, I selected Qwen2.5-VL-7B-Instruct but received marco-o1-uncensored-exl2-8.0bpw on API:

2025-02-10T03:52:48.197Z debug: [AskController]
{
  text: "What is JP char?",
  conversationId: "2326d069-43b6-44fd-b127-3e3746457a08",
  endpoint: "Local-Relay-All",
  endpointType: "custom",
  resendFiles: true,
    modelOptions.model: "Qwen2.5-VL-7B-Instruct",
  modelsConfig: "exists",
  attachments: [object Promise],
}
2025-02-10T03:52:48.198Z debug: [BaseClient] Loading history:
{
  conversationId: "2326d069-43b6-44fd-b127-3e3746457a08",
  parentMessageId: "00000000-0000-0000-0000-000000000000",
}
2025-02-10T03:52:48.205Z debug: [BaseClient] Context Count (1/2)
{
  remainingContextTokens: 3828,
  maxContextTokens: 4095,
}
2025-02-10T03:52:48.205Z debug: [BaseClient] Context Count (2/2)
{
  remainingContextTokens: 3828,
  maxContextTokens: 4095,
}
2025-02-10T03:52:48.205Z debug: [BaseClient] tokenCountMap:
{
  eba6c2c4-113d-4ef0-80c9-114f0ac54afe: 264,
}
2025-02-10T03:52:48.205Z debug: [BaseClient]
{
  promptTokens: 267,
  remainingContextTokens: 3828,
  payloadSize: 1,
  maxContextTokens: 4095,
}
2025-02-10T03:52:48.205Z debug: [BaseClient] tokenCountMap
{
  eba6c2c4-113d-4ef0-80c9-114f0ac54afe: 264,
  instructions: undefined,
}
2025-02-10T03:52:48.205Z debug: [BaseClient] userMessage
{
  messageId: "eba6c2c4-113d-4ef0-80c9-114f0ac54afe",
  parentMessageId: "00000000-0000-0000-0000-000000000000",
  conversationId: "2326d069-43b6-44fd-b127-3e3746457a08",
  sender: "User",
  text: "What is JP char?",
  isCreatedByUser: true,
  // 1 image_url(s)
  image_urls: [{"type":"image_url","image_url":{"url":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAACkAAAA3CAYAAA... [truncated]],
  tokenCount: 264,
}
2025-02-10T03:52:48.206Z debug: [OpenAIClient] chatCompletion
{
  baseURL: "http://172.22.22.172:8788/v1",
    modelOptions.model: "marco-o1-uncensored-exl2-8.0bpw",
    modelOptions.user: "64ba024578db8649d72687ff",
    modelOptions.stream: true,
    // 1 message(s)
    modelOptions.messages: [{"role":"user","content":[{"type":"text","text":"What is JP char?"},{"type":"image_url","image_url":... [truncated]],
}
2025-02-10T03:52:48.206Z debug: [OpenAIClient] chatCompletion: dropped params
{
  dropParams: ,
    modelOptions.model: "marco-o1-uncensored-exl2-8.0bpw",
    modelOptions.user: "64ba024578db8649d72687ff",
    modelOptions.stream: true,
    // 1 message(s)
    modelOptions.messages: [{"role":"user","content":[{"type":"text","text":"What is JP char?"},{"type":"image_url","image_url":... [truncated]],
    modelOptions.max_tokens: 4000,
}
2025-02-10T03:52:48.207Z debug: Making request to http://172.22.22.172:8788/v1/chat/completions
2025-02-10T03:52:48.210Z debug: [saveConvo] api/app/clients/BaseClient.js - saveMessageToDatabase #saveConvo


2025-02-10T03:53:19.639Z debug: [OpenAIClient] chatCompletion response
{
  object: "chat.completion",
  usage: null,
  id: "chatcmpl-9465c49e2f5842ffbfc4889e25b14701",
  // 1 choice(s)
  choices: [{"message":{"tool_calls_json":null,"role":"assistant","content":"<Thought>\nAlright, I need to figur... [truncated]],
  created: 1739159599,
  model: "marco-o1-uncensored-exl2-8.0bpw",
}
2025-02-10T03:53:19.641Z debug: [spendTokens] conversationId: 2326d069-43b6-44fd-b127-3e3746457a08 | Context: message | Token usage:
{
  promptTokens: 267,
  completionTokens: 1352,
}
2025-02-10T03:53:19.646Z debug: [spendTokens] No transactions incurred against balance
2025-02-10T03:53:19.649Z debug: [saveConvo] api/app/clients/BaseClient.js - saveMessageToDatabase #saveConvo
2025-02-10T03:53:19.664Z debug: [AskController] Request closed

New Incognito Window still so not a broswer cache problem.

Version Information

ghcr.io/danny-avila/librechat-dev latest c2564c51d4f5 7 hours ago 866MB

Steps to Reproduce

A custom endpoint with a lot of models which are fetched. Image query(text query is fine...?).

What browsers are you seeing the problem on?

Chrome

Code of Conduct

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

[danny-avila]

It selects the first vision model as a fallback if the current model is not. I will need to add Qwen2.5-VL as a valid vision model