Fatal error in Google Gemini custom v0.7.8

[holasoygeorgee]

What happened?

When trying to use Google Gemini as a custom in the Yaml it works perfectly, but when trying to use tools an error occurs that crashes the entire app.

Version Information

I use the latest version in npm librechat v0.7.8

Steps to Reproduce

1.- I put my configuration in the yaml
custom:
- name: "google2"
apiKey: user_provided
baseURL: "https://generativelanguage.googleapis.com/v1beta"
models:
default: ["gemini-2.0-flash-exp"]
fetch: true
titleConvo: false
customParams:
defaultParamsEndpoint: 'google'
2.- I used the model and it responded perfectly until I used a tool and it crashed.

What browsers are you seeing the problem on?

google chrome

Relevant log output

2025-07-16 05:29:11 error: [api/server/controllers/agents/client.js #sendCompletion] Operation aborted No stepId found for tool_call_id undefined
2025-07-16 05:29:11 error: [api/server/controllers/agents/client.js #sendCompletion] Unhandled error type No stepId found for tool_call_id undefined
2025-07-16 05:29:11 info: [BaseClient] Debug - Checking if user provides API key
2025-07-16 05:29:11 info: [BaseClient] User API key check result:
2025-07-16 05:29:12 error: There was an uncaught error: Aborted

Screenshots

Code of Conduct

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

[danny-avila]

v0.7.8 is not the latest version, please update:

Updating Instructions (docker):
https://www.librechat.ai/docs/local/docker#update-librechat

[holasoygeorgee]

I updated and the error keeps appearing and Librechat keeps crashing. I tried it in Docker and the same problem persists.

[danny-avila]

Use google directly, you don't need a custom endpoint especially if you are using the official API

[holasoygeorgee]

Yes, it's perfect there, but I wanted users to also have the option to add their Google API since it's free. However, I can't have both the global API and the user_provided API at the same time. I tried using the endpoints, but I got that error in the tools. I tried to fix it, but the error is very deep; the server crashes completely if I use a tool. If you could give me a guide on how to fix this, I would be very grateful.

[danny-avila]

It's up to Google to improve their OpenAI-spec compatibility. You can also look into PortKey or LiteLLM to help serve Google in OpenAI-spec.