@nest-langchain/bedrock

English | 한국어

AWS Bedrock chat model provider for NestJS dependency injection.

This package wires a ChatBedrockConverse instance from @langchain/aws into NestJS dependency injection. Connection info (region, optional credentials) lives at the module level; model and temperature are chosen per preset or per call via a factory — the library no longer picks a default model.

Install

pnpm add @nest-langchain/bedrock @langchain/aws

Module

import { Module } from '@nestjs/common';
import { BedrockProviderModule } from '@nest-langchain/bedrock';

@Module({
  imports: [
    BedrockProviderModule.forRoot({
      region: process.env.AWS_REGION,
      presets: [
        {
          name: 'sonnet',
          model: 'anthropic.claude-3-5-sonnet-20240620-v1:0',
          temperature: 0,
        },
        {
          name: 'haiku',
          model: 'anthropic.claude-3-5-haiku-20241022-v1:0',
          temperature: 0.2,
        },
      ],
    }),
  ],
})
export class AiModule {}

Region resolution order:

1. region option
2. AWS_REGION
3. AWS_DEFAULT_REGION
4. active AWS_PROFILE region from the AWS config file

Pass credentials at the module level when the host app does not use the default AWS provider chain.

Injection

Inject a named preset, or inject the factory for per-call model creation. model is always required — the library never assumes one:

import { Injectable } from '@nestjs/common';
import { ChatBedrockConverse } from '@langchain/aws';
import {
  InjectBedrockChatModel,
  InjectBedrockChatModelFactory,
  BedrockChatModelFactory,
} from '@nest-langchain/bedrock';

@Injectable()
export class BedrockWorkflow {
  constructor(
    @InjectBedrockChatModel('sonnet')
    private readonly model: ChatBedrockConverse,
    @InjectBedrockChatModelFactory()
    private readonly factory: BedrockChatModelFactory,
  ) {}

  run(prompt: string) {
    return this.model.invoke(prompt);
  }

  runWith(model: string, prompt: string) {
    return this.factory.create({ model }).invoke(prompt);
  }
}

For dynamic lookup use getBedrockChatModelToken(name).

Async Connection Info

Use forRootAsync when the region or credentials come from ConfigService or a secrets manager. Presets stay static:

BedrockProviderModule.forRootAsync({
  imports: [ConfigModule],
  inject: [ConfigService],
  useFactory: (config: ConfigService) => ({
    region: config.get('AWS_REGION'),
  }),
  presets: [
    { name: 'sonnet', model: 'anthropic.claude-3-5-sonnet-20240620-v1:0' },
  ],
});

Migration (v0.1 → v0.2)

NEST_LANGCHAIN_BEDROCK_CHAT_MODEL and the module-level model / temperature options were removed. Replace @Inject(NEST_LANGCHAIN_BEDROCK_CHAT_MODEL) with a named preset, or inject the factory with @InjectBedrockChatModelFactory() and call .create({ model }).

Demo

AWS_REGION=us-east-1 pnpm --filter @nest-langchain/demo-providers start
curl -X POST "http://localhost:3006/providers/bedrock/invoke" \
  -H "content-type: application/json" \
  -d '{"prompt":"Write one sentence about Bedrock model routing."}'