bedrock-wrapper v2.3.0

🪨 Bedrock Wrapper

Bedrock Wrapper is an npm package that simplifies the integration of existing OpenAI-compatible API objects with AWS Bedrock's serverless inference LLMs. Follow the steps below to integrate into your own application, or alternativly use the 🔀 Bedrock Proxy Endpoint project to spin up your own custom OpenAI server endpoint for even easier inference (using the standard baseUrl, and apiKey params).

Maintained by

Install

• install package: npm install bedrock-wrapper

Usage

1. import bedrockWrapper

   import { bedrockWrapper } from "bedrock-wrapper";

2. create an awsCreds object and fill in your AWS credentials

   const awsCreds = {
       region: AWS_REGION,
       accessKeyId: AWS_ACCESS_KEY_ID,
       secretAccessKey: AWS_SECRET_ACCESS_KEY,
   };

3. clone your openai chat completions object into openaiChatCompletionsCreateObject or create a new one and edit the values

   const openaiChatCompletionsCreateObject = {
       "messages": messages,
       "model": "Llama-3-1-8b",
       "max_tokens": LLM_MAX_GEN_TOKENS,
       "stream": true,
       "temperature": LLM_TEMPERATURE,
       "top_p": LLM_TOP_P,
   };

   the messages variable should be in openai's role/content format

   messages = [
       {
           role: "system",
           content: "You are a helpful AI assistant that follows instructions extremely well. Answer the user questions accurately. Think step by step before answering the question. You will get a $100 tip if you provide the correct answer.",
       },
       {
           role: "user",
           content: "Describe why openai api standard used by lots of serverless LLM api providers is better than aws bedrock invoke api offered by aws bedrock. Limit your response to five sentences.",
       },
       {
           role: "assistant",
           content: "",
       },
   ]

   the model value should be the corresponding modelName value in the bedrock_models section below (see Supported Models below)

4. call the bedrockWrapper function and pass in the previously defined awsCreds and openaiChatCompletionsCreateObject objects

   // create a variable to hold the complete response
   let completeResponse = "";
   // invoke the streamed bedrock api response
   for await (const chunk of bedrockWrapper(awsCreds, openaiChatCompletionsCreateObject)) {
       completeResponse += chunk;
       // ---------------------------------------------------
       // -- each chunk is streamed as it is received here --
       // ---------------------------------------------------
       process.stdout.write(chunk); // ⇠ do stuff with the streamed chunk
   }
   // console.log(`\n\completeResponse:\n${completeResponse}\n`); // ⇠ optional do stuff with the complete response returned from the API reguardless of stream or not

   if calling the unstreamed version you can call bedrockWrapper like this

   // create a variable to hold the complete response
   let completeResponse = "";
   if (!openaiChatCompletionsCreateObject.stream){ // invoke the unstreamed bedrock api response
       const response = await bedrockWrapper(awsCreds, openaiChatCompletionsCreateObject);
       for await (const data of response) {
           completeResponse += data;
       }
       // ----------------------------------------------------
       // -- unstreamed complete response is available here --
       // ----------------------------------------------------
       console.log(`\n\completeResponse:\n${completeResponse}\n`); // ⇠ do stuff with the complete response
   }

Supported Models

To return the list progrmatically you can import and call listBedrockWrapperSupportedModels:

import { listBedrockWrapperSupportedModels } from 'bedrock-wrapper';
console.log(`\nsupported models:\n${JSON.stringify(await listBedrockWrapperSupportedModels())}\n`);

Additional Bedrock model support can be added.
Please modify the bedrock_models.js file and submit a PR 🏆 or create an Issue.

Image Support

For models with image support (Claude 3.5 Sonnet, Claude 3.7 Sonnet, and Claude 3.7 Sonnet Thinking), you can include images in your messages using the following format:

messages = [
    {
        role: "system",
        content: "You are a helpful AI assistant that can analyze images.",
    },
    {
        role: "user",
        content: [
            { type: "text", text: "What's in this image?" },
            { 
                type: "image_url", 
                image_url: {
                    url: "data:image/jpeg;base64,/9j/4AAQSkZJRgABAQEA..." // base64 encoded image
                }
            }
        ]
    }
]

You can also use a direct URL to an image instead of base64 encoding:

messages = [
    {
        role: "user",
        content: [
            { type: "text", text: "Describe this image in detail." },
            { 
                type: "image_url", 
                image_url: {
                    url: "https://example.com/path/to/image.jpg" // direct URL to image
                }
            }
        ]
    }
]

You can include multiple images in a single message by adding more image_url objects to the content array.

📢 P.S.

In case you missed it at the beginning of this doc, for an even easier setup, use the 🔀 Bedrock Proxy Endpoint project to spin up your own custom OpenAI server endpoint (using the standard baseUrl, and apiKey params).

📚 References

• AWS Meta Llama Models User Guide
• AWS Mistral Models User Guide
• OpenAI API
• AWS Bedrock
• AWS SDK for JavaScript

Please consider sending me a tip to support my work 😀

🍵 tip me here