Claude 3.7 not work [REQUIRES UPDATE]

[woojuini]

What happened?

When set up and applied as shown below

.env
ANTHROPIC_MODELS=claude-3-7-sonnet-20250219,claude-3-5-sonnet-20241022

3.5 works fine but 3.7 got error...

Something went wrong. Here's the specific error message we encountered: 
Operation failed after 3 attempts: 400 
{"type":"error","error":{"type":"invalid_request_error","message":"prompt: Input should be a valid string"}}

Version Information

v0.7.7-rc1

Steps to Reproduce

.

What browsers are you seeing the problem on?

No response

Relevant log output

2025-02-25T05:39:17.244Z debug: Making request to https://api.anthropic.com/v1/complete
2025-02-25T05:39:17.343Z debug: [saveConvo] api/app/clients/BaseClient.js - saveMessageToDatabase #saveConvo
2025-02-25T05:39:17.730Z warn: User: 626d9afc | Anthropic Request 1 failed: 400 {"type":"error","error":{"type":"invalid_request_error","message":"prompt: Input shou... [truncated]
2025-02-25T05:39:18.081Z debug: Making request to https://api.anthropic.com/v1/complete
2025-02-25T05:39:18.281Z warn: User: 626d9afc | Anthropic Request 2 failed: 400 {"type":"error","error":{"type":"invalid_request_error","message":"prompt: Input shou... [truncated]
2025-02-25T05:39:18.984Z debug: Making request to https://api.anthropic.com/v1/complete
2025-02-25T05:39:19.189Z warn: User: 626d9afc | Anthropic Request 3 failed: 400 {"type":"error","error":{"type":"invalid_request_error","message":"prompt: Input shou... [truncated]
2025-02-25T05:39:19.192Z error: [handleAbortError] AI response error; aborting request: Operation failed after 3 attempts: 400 {"type":"error","error":{"type":"invalid_request_error"... [truncated]
2025-02-25T05:39:19.201Z debug: [AskController] Request closed

Screenshots

.

Code of Conduct

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

[fuegovic]

You need to apply today's update for 3.7 to work
#6008

[i40west]

With today's update, and "thinking" enabled, all requests return the error:

Something went wrong. Here's the specific error message we encountered:
Operation failed after 3 attempts: 400 {"type":"error","error":{"type":"invalid_request_error",
"message":"`top_k` must be unset when thinking is enabled. Please consult our documentation at https://docs.anthropic.com/en/docs/build-with-claude/extended-thinking#important-considerations-when-using-extended-thinking"}}

Is there a way to unset top_k?

[EinMaulwurf]

I have the same problem, when trying to remove the top_k value in the sidebar, it's automatically set to 1. And thus the request fails.

[danny-avila]

Top K is not set automatically:

Make sure you are updated (instructions below), and simply start a new chat without setting Top K. LibreChat is not opinionated with the parameters and does not set it for you, nor will it try to remove params if they were set manually based on model support.

The only exception is max output tokens, since it is required when using reasoning.

[i40west]

simply start a new chat without setting Top K

This is the tricky part. How do you start a new chat without that set? Checking my history, it's been set on every chat I've ever done, and I don't remember ever setting or changing it.

[EinMaulwurf]

For me it worked by un-pin the preset I use and open a new chat. Then select Anthropic and Claude 3.7. When I opened the sidebar, it still showed a Top-K of 5, but the reasoning worked. I could than set my system prompt and save this as a preset (all through the side panel).

[i40west]

That does seem to work. In the sidebar it says top_k is set to 5 but I guess that's just in the UI.

[danny-avila]

Updating Instructions (docker):

# Linux command to Remove all existing images
docker images -a | grep "librechat" | awk '{print $3}' | xargs docker rmi

# Windows Powershell
docker images -a | findstr "librechat" | ForEach-Object { docker rmi $_.Split()[2] }

Then follow the directions here: https://www.librechat.ai/docs/local/docker#update-librechat

[divyam234]

Or use watchtower to automatically update image