Add input_audio support to @langchain/google-common

[hntrl]

Privileged issue

• I am a LangChain maintainer, or was asked directly by a LangChain maintainer to create an issue here.

Issue Content

@langchain/google-common (used by @langchain/google-vertexai) does not support the OpenAI-standardized input_audio message content type. This causes an Unsupported type "input_audio" error when using Gemini models with prompts containing audio variables, particularly those from LangChain Hub or OpenAI-compatible tools.

Root Cause

The messageContentComplexToPart function in @langchain/google-common/dist/utils/gemini.js (around line 290) only handles text, image_url, media, and reasoning content types. It lacks support for input_audio.

Proposed Solution

Add input_audio case to the messageContentComplexToPart function to convert OpenAI-format audio to Gemini's inlineData format:

case "input_audio":
    if ("input_audio" in content) {
        return {
            inlineData: {
                mimeType: `audio/${content.input_audio.format}`,
                data: content.input_audio.data
            }
        };
    }
    break;