Issue with the model used by the agents.

[Passerby1011]

What happened?

I chose GPT-4o, but LibreChat actually uses GPT-4o-mini.

From the backend of my API management, I see that LibreChat is actually calling GPT-4o-mini.

Steps to Reproduce

Normal use of agent conversation

What browsers are you seeing the problem on?

Chrome

Relevant log output

2024-10-21T03:06:03.816Z debug: [BaseClient] Loading history:
{
  conversationId: "fd1ce9ca-05be-4fbc-a0bb-93d71a28e266",
  parentMessageId: "00000000-0000-0000-0000-000000000000",
}
2024-10-21T03:06:03.822Z debug: [BaseClient] Context Count (1/2)
{
  remainingContextTokens: 127492,
  maxContextTokens: 127500,
}
2024-10-21T03:06:03.823Z debug: [BaseClient] Context Count (2/2)
{
  remainingContextTokens: 127492,
  maxContextTokens: 127500,
}
2024-10-21T03:06:03.823Z debug: [BaseClient] tokenCountMap:
2024-10-21T03:06:03.823Z debug: [BaseClient]
{
  promptTokens: 8,
  remainingContextTokens: 127492,
  payloadSize: 1,
  maxContextTokens: 127500,
}
2024-10-21T03:06:03.824Z debug: [api/server/controllers/agents/client.js] chatCompletion
{
  baseURL: undefined,
  // 1 payload(s)
  payload: [{"role":"user","content":"你好"}],
}
2024-10-21T03:06:03.855Z debug: [saveConvo] api/app/clients/BaseClient.js - saveMessageToDatabase #saveConvo
2024-10-21T03:06:05.013Z debug: [spendTokens] conversationId: fd1ce9ca-05be-4fbc-a0bb-93d71a28e266 | Context: message | Token usage:
{
  promptTokens: 502,
  completionTokens: 15,
}
2024-10-21T03:06:05.039Z debug: [saveConvo] api/app/clients/BaseClient.js - saveMessageToDatabase #saveConvo
2024-10-21T03:06:05.044Z debug: [spendTokens] No transactions incurred against balance
2024-10-21T03:06:05.065Z debug: [AgentController] Request closed

Screenshots

Code of Conduct

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

[danny-avila]

This issue can't be reproduced. Make sure you save the selection for the Agent.

EDIT: Actually, I see the issue now. The model value for OpenAI is hardcoded for testing. Will remedy that soon

[Passerby1011]

I'm sure I saved it.

[danny-avila]

I've identified the issue. As the flag to enable this lists, this feature is still experimental. There is still some code in there that was used for testing.

[danny-avila]

This will fix it: #4483

[Dannypeja]

Could it be that we have that issue with Anthropic models currently?
No matter which model is selected (and saved) in the agent creation, It seems to always use sonnett.

[danny-avila]

@Dannypeja yes looks like it, similar issue. Fixed here: 951bb9d