AWS Bedrock Provider

@flowrag/provider-bedrock provides embeddings, entity extraction, and reranking via AWS Bedrock.

Installation

npm install @flowrag/provider-bedrock @aws-sdk/client-bedrock-runtime

Configuration

Uses standard AWS credentials (environment variables, IAM role, SSO profile, etc.):

export AWS_REGION=eu-central-1

Embedder

import { BedrockEmbedder } from '@flowrag/provider-bedrock';

const embedder = new BedrockEmbedder({
  model: 'amazon.titan-embed-text-v2:0', // default
});

embedder.dimensions; // 1024

Extractor

import { BedrockExtractor } from '@flowrag/provider-bedrock';

const extractor = new BedrockExtractor({
  model: 'anthropic.claude-haiku-4-5-20251001-v1:0', // default
});

Reranker

import { BedrockReranker } from '@flowrag/provider-bedrock';

const reranker = new BedrockReranker();
// Uses amazon.rerank-v1:0

Model Constants

Well-known model IDs are exported as constants:

import { BedrockEmbeddingModels, BedrockLLMModels, BedrockRerankerModels } from '@flowrag/provider-bedrock';

BedrockEmbeddingModels.TITAN_EMBED_V2;    // 'amazon.titan-embed-text-v2:0'
BedrockEmbeddingModels.COHERE_EMBED_V4;   // 'cohere.embed-v4:0'
BedrockLLMModels.CLAUDE_HAIKU_4_5;        // 'anthropic.claude-haiku-4-5-20251001-v1:0'
BedrockLLMModels.CLAUDE_SONNET_4_6;       // 'anthropic.claude-sonnet-4-6'
BedrockRerankerModels.COHERE_RERANK_V3_5; // 'cohere.rerank-v3-5:0'

All model parameters accept any string, so you can use newer models as they become available.

Enterprise Recommendations

Embedding

Pick	Model	Why
🏆 Best value	Titan Embed V2	Cheap, 1024 dims, wide regional availability
👑 Best quality	Cohere Embed V4	Latest gen, text+image, superior benchmarks
🌍 Multilingual	Cohere Embed Multilingual V3	100+ languages, 1024 dims

LLM (Entity Extraction)

Pick	Model	Why
🏆 Best value	Claude Haiku 4.5	Fast, $1/$5 per 1M tokens, reliable structured output
👑 Best quality	Claude Sonnet 4.6	Latest Sonnet, excellent extraction, wide cross-region
💎 Top tier	Claude Opus 4.6	Absolute best, use for critical extraction
💰 Budget	Nova Lite	20x cheaper than Haiku, sufficient for simple schemas

Reranker

Pick	Model	Why
🏆 Best value	Amazon Rerank V1	Dedicated reranker, $1/1000 queries
👑 Best quality	Cohere Rerank 3.5	SOTA on BEIR, multilingual, reasoning

TIP

Cohere models (Embed, Rerank) use a different body format than Amazon/Anthropic models. FlowRAG handles this automatically — just pass the model ID.

Full Example (Lambda)

import { BedrockEmbedder, BedrockExtractor, BedrockReranker } from '@flowrag/provider-bedrock';
import { S3KVStorage } from '@flowrag/storage-s3';
import { OpenSearchVectorStorage, OpenSearchGraphStorage } from '@flowrag/storage-opensearch';

const rag = createFlowRAG({
  schema,
  storage: {
    kv: new S3KVStorage({ client: s3Client, bucket: 'my-rag-bucket', prefix: 'kv/' }),
    vector: new OpenSearchVectorStorage({ client: osClient, dimensions: 1024 }),
    graph: new OpenSearchGraphStorage({ client: osClient }),
  },
  embedder: new BedrockEmbedder(),
  extractor: new BedrockExtractor(),
  reranker: new BedrockReranker(), // optional
});

See AWS Lambda Deployment for a complete example.