Gemini Provider

@flowrag/provider-gemini provides embeddings, entity extraction, and reranking via Google's Gemini API.

Installation

npm install @flowrag/provider-gemini

Configuration

Set your API key:

export GEMINI_API_KEY=your-key

Embedder

import { GeminiEmbedder } from '@flowrag/provider-gemini';

const embedder = new GeminiEmbedder({
  model: 'gemini-embedding-001', // default
});

embedder.dimensions; // 3072

Extractor

import { GeminiExtractor } from '@flowrag/provider-gemini';

const extractor = new GeminiExtractor({
  model: 'gemini-3-flash-preview', // default
});

The extractor sends your schema (entity types, relation types, custom fields) to Gemini and receives structured entity/relation data.

Reranker

import { GeminiReranker } from '@flowrag/provider-gemini';

const reranker = new GeminiReranker();

Uses LLM-based relevance scoring — each result is evaluated against the query.

Full Example

import { GeminiEmbedder, GeminiExtractor, GeminiReranker } from '@flowrag/provider-gemini';

const rag = createFlowRAG({
  schema,
  storage: { kv, vector, graph },
  embedder: new GeminiEmbedder(),
  extractor: new GeminiExtractor(),
  reranker: new GeminiReranker(), // optional
});