feat(community): Support AWS Bedrock invoke model

[AllenFang]

Context

In short, AWS Bedrock not only support Converse but also Invoke model which is suitable for single-turn inference, content generation tasks, embeddings. This MR is inspired from #7697 discussion. Today AWS supports the cache point but Converse hasn't support yet from aws SDK but Invoke Model supported.

About PR

This is first draft for supporting AWS Bedrock Invoke Model and I want to let team and @jacoblee93 to decide whether we should continue to support this feature. If no, feel free to close. If true, I will then add some testing codes, code comments and prepare document.

Notes

• this PR doesn't support streaming, i want streaming to be supported in next time for minimizing the code changes and review efforts
• cache checkpoints will be implemented in this PR once team decide to support Invoke Model in langchain.

Let me know your thoughts. Thx 🙏 🙏

Local Testing

• Basic Message
  Request:

const model = new ChatBedrockInvokeModel({
    region: "us-east-1",
    model: "anthropic.claude-3-5-sonnet-20240620-v1:0",
    credentials: {
      accessKeyId: "...",
      secretAccessKey: "...",
    },
});

const response = await model.invoke([
 [new HumanMessage("Hello, how are you?")]
], {
  body: {
      max_tokens: 120,
      anthropic_version: "bedrock-2023-05-31"
      // temperature: 0.5,
      // top_p: 2,
      // stop_sequences: [],
      // ... others
    },
});

Response:

AIMessage {
  "id": "msg_bdrk_01Ci5ueXLCNPduGzbYUyJGkR",
  "content": "Hello! As an AI language model, I don't have feelings or emotions, but I'm functioning properly and ready to assist you with any questions or tasks you may have. How can I help you today?",
  "additional_kwargs": {},
  "response_metadata": {
    "$metadata": {
      "httpStatusCode": 200,
      "requestId": "7497471b-664c-43c5-b374-083603ead322",
      "attempts": 1,
      "totalRetryDelay": 0
    },
    "contentType": "application/json",
    "type": "message",
    "model": "claude-3-haiku-20240307",
    "stop_reason": "end_turn",
    "stop_sequence": null
  },
  "tool_calls": [],
  "invalid_tool_calls": [],
  "usage_metadata": {
    "input_tokens": 13,
    "output_tokens": 45,
    "total_tokens": 58
  }
}

• With tool
  Request:

const model = new ChatBedrockInvokeModel({
    region: "us-east-1",
    model: "anthropic.claude-3-5-sonnet-20240620-v1:0",
    credentials: {
      accessKeyId: "...",
      secretAccessKey: "....",
    },
});

const response = await model.invoke([
 [new HumanMessage("Hello, how are you?")]
], {
  body: {
      max_tokens: 120,
      anthropic_version: "bedrock-2023-05-31"
      // temperature: 0.5,
      // top_p: 2,
      // stop_sequences: [],
      // ... others
    },
    tools: [....]
});

Response:

AIMessage {
  "id": "msg_bdrk_017An62jigbN7gShHUpXHJ3x",
  "content": [
    {
      "type": "text",
      "text": "Here is the calculation:"
    }
  ],
  "additional_kwargs": {},
  "response_metadata": {
    "$metadata": {
      "httpStatusCode": 200,
      "requestId": "a9a70436-4cd2-463c-9ad9-e771cf32f804",
      "attempts": 1,
      "totalRetryDelay": 0
    },
    "contentType": "application/json",
    "type": "message",
    "model": "claude-3-haiku-20240307",
    "stop_reason": "tool_use",
    "stop_sequence": null
  },
  "tool_calls": [
    {
      "id": "toolu_bdrk_01DiDiMN2WVzpeab5x9FPueb",
      "name": "multiply",
      "args": {
        "a": 2,
        "b": 3
      },
      "type": "tool_call"
    }
  ],
  "invalid_tool_calls": [],
  "usage_metadata": {
    "input_tokens": 372,
    "output_tokens": 68,
    "total_tokens": 440
  }
}

• Images
• Documents

[vercel]

The latest updates on your projects. Learn more about Vercel for Git ↗︎

Name	Status	Preview	Comments	Updated (UTC)
langchainjs-docs	✅ Ready (Inspect)	Visit Preview		Mar 7, 2025 8:28am

1 Skipped Deployment

Name	Status	Preview	Comments	Updated (UTC)
langchainjs-api-refs	⬜️ Ignored (Inspect)			Mar 7, 2025 8:28am

[jacoblee93]

Hey @AllenFang, thanks for this PR!

@3coins are you all planning to gradually migrate everything over to the Converse API?

If so, I'd rather just leave the parallel implementation in community. If not, this makes sense.