Add ability to set outputDimensionality for Gemini embeddings

OSS Product Help LangChain
1

Problem:

  • Dimension cannot be specified dynamically. We must hard-code DB schemas (e.g., pgvector dimensions) and hope provider defaults don’t change.
  • Risk of silent breaking changes. If Google increases default dims, our pipelines fail (DB rejects vectors of wrong length).
  • Cost & performance. Many tasks don’t need 3k-dim vectors; smaller dims reduce storage and retrieval latency with minimal quality loss (per Google docs).

Today’s behavior

const emb = new GoogleGenerativeAIEmbeddings({
  model: "models/gemini-embedding-001",
  // no way to set outputDimensionality
});

const v = await emb.embedQuery("What is the meaning of life?");
// Returns default size (currently 3072)

Behavior from offitial google doc (https://ai.google.dev/gemini-api/docs/embeddings#javascript_3):

import { GoogleGenAI } from "@google/genai";

async function main() {
    const ai = new GoogleGenAI({});

    const response = await ai.models.embedContent({
        model: 'gemini-embedding-001',
        content: 'What is the meaning of life?',
        outputDimensionality: 768,
    });

    const embeddingLength = response.embedding.values.length;
    console.log(`Length of embedding: ${embeddingLength}`);
}

main();