### Install Dependencies

Source: https://github.com/jparkerweb/bedrock-wrapper/blob/main/AGENTS.md

Installs project dependencies. Run this command after cloning the repository.

```bash
npm install
```

--------------------------------

### Install bedrock-wrapper

Source: https://github.com/jparkerweb/bedrock-wrapper/blob/main/_autodocs/README.md

Install the bedrock-wrapper package using npm.

```bash
npm install bedrock-wrapper
```

--------------------------------

### Converse API Request Example (Kimi-K2)

Source: https://github.com/jparkerweb/bedrock-wrapper/blob/main/test-final-output.txt

This is an example of a request sent to the Converse API for Kimi-K2.

```json
{
  "modelId": "moonshot.kimi-k2-thinking",
  "messages": [
    {
      "role": "user",
      "content": [
        {
          "text": "Respond with exactly one word: What is 1+1?"
        }
      ]
    }
  ],
  "inferenceConfig": {
    "maxTokens": 800,
    "temperature": 0.2
  }
}
```

--------------------------------

### Example AWS Credentials Configuration

Source: https://github.com/jparkerweb/bedrock-wrapper/blob/main/_autodocs/types.md

Provides an example of how to configure AWS credentials using environment variables. This object is used as the first parameter to bedrockWrapper().

```javascript
const awsCreds = {
  region: "us-west-2",
  accessKeyId: process.env.AWS_ACCESS_KEY_ID,
  secretAccessKey: process.env.AWS_SECRET_ACCESS_KEY,
};
```

--------------------------------

### Complete Chat Application Example

Source: https://github.com/jparkerweb/bedrock-wrapper/blob/main/_autodocs/README.md

This example demonstrates how to build a complete chat application using the bedrock-wrapper. It includes setting up AWS credentials, defining a chat function that handles user messages, streaming responses, and querying available models.

```javascript
import dotenv from 'dotenv';
import { bedrockWrapper, listBedrockWrapperSupportedModels } from 'bedrock-wrapper';

dotenv.config();

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

async function chat(userMessage) {
  const request = {
    messages: [
      { role: "system", content: "You are a helpful assistant." },
      { role: "user", content: userMessage }
    ],
    model: "Claude-3-5-Sonnet",
    max_tokens: 500,
    stream: true,
    temperature: 0.7,
  };

  console.log("Assistant: ");
  let response = "";
  try {
    for await (const chunk of bedrockWrapper(awsCreds, request, { 
      useConverseAPI: true 
    })) {
      response += chunk;
      process.stdout.write(chunk);
    }
    console.log("\n");
    return response;
  } catch (error) {
    console.error("Error:", error.message);
    throw error;
  }
}

// Usage
const response = await chat("What is artificial intelligence?");
console.log("Complete response:", response);

// Query models
const models = await listBedrockWrapperSupportedModels();
console.log(`Available: ${models.length} models`);
```

--------------------------------

### Environment Setup

Source: https://github.com/jparkerweb/bedrock-wrapper/blob/main/AGENTS.md

Configuration for AWS region, credentials, and LLM parameters. Create a .env file with these variables.

```dotenv
AWS_REGION=us-west-2
AWS_ACCESS_KEY_ID=your_key
AWS_SECRET_ACCESS_KEY=your_secret
LLM_MAX_GEN_TOKENS=1024
LLM_TEMPERATURE=0.2
```

--------------------------------

### Clean Reinstall

Source: https://github.com/jparkerweb/bedrock-wrapper/blob/main/AGENTS.md

Cleans the project by removing node_modules and package-lock.json, followed by a fresh install.

```bash
npm run clean
```

--------------------------------

### Invoke API Response Example (DeepSeek V3.1)

Source: https://github.com/jparkerweb/bedrock-wrapper/blob/main/test-run-output.txt

This is an example of a non-streaming response from the Invoke API for the DeepSeek V3.1 model, showing the assistant's text output.

```json
{
  "output": {
    "message": {
      "role": "assistant",
      "content": [
        {
          "text": "two"
        }
      ]
    }
  },
  "stopReason": "end_turn",
  "usage": {
    "inputTokens": 17,
    "outputTokens": 2,
    "totalTokens": 19
  },
  "metrics": {
    "latencyMs": 306
  },
  "$metadata": {
    "httpStatusCode": 200,
    "requestId": "b7addef3-7683-46ca-b69d-568d7cdbff8e",
    "attempts": 1,
    "totalRetryDelay": 0
  }
}
```

--------------------------------

### With Thinking Mode (Claude)

Source: https://github.com/jparkerweb/bedrock-wrapper/blob/main/_autodocs/api-reference/bedrockWrapper.md

Illustrates how to enable and utilize the 'thinking' mode for Claude models to get reasoning steps.

```APIDOC
## With Thinking Mode (Claude)

### Description
This example enables the 'thinking' mode for Claude models, which includes reasoning steps in the output, typically enclosed in `<think>` tags.

### Method
`bedrockWrapper(awsCreds, request)`

### Parameters
- **awsCreds** (object) - AWS credentials and region.
- **request** (object) - Configuration for the model request, including `model` set to a thinking-enabled model, `temperature` (required), and `include_thinking_data: true`.

### Request Example
```javascript
const request = {
  messages: [
    { role: "user", content: "Solve: What's 2^100 in scientific notation?" }
  ],
  model: "Claude-4-5-Sonnet-Thinking",
  max_tokens: 4000,
  stream: true,
  temperature: 1.0,  // Required for thinking
  include_thinking_data: true
};

for await (const chunk of bedrockWrapper(awsCreds, request)) {
  process.stdout.write(chunk);
}
// Output: "<think>...reasoning...</think>\n\n<response>"
```

### Response
- **chunk** (string) - Streaming response chunks, potentially including reasoning tags and the final response.
```

--------------------------------

### Invoke API Request for GPT-OSS-20B

Source: https://github.com/jparkerweb/bedrock-wrapper/blob/main/test-run-output.txt

Example of a request sent to the Invoke API for the GPT-OSS-20B model, including formatted prompts.

```javascript
Final formatted prompt: [
  {
    role: 'user',
    content: 'Respond with exactly one word: What is 1+1?'
  },
  {
    role: 'assistant',
    content: ''
  }
]

Final request: {
  "messages": [
    {
      "role": "user",
      "content": "Respond with exactly one word: What is 1+1?"
    },
    {
      "role": "assistant",
      "content": ""
    }
  ],
  "max_completion_tokens": 800,
  "temperature": 0.2
}
```

--------------------------------

### Converse API Response Example (Claude-3-Haiku)

Source: https://github.com/jparkerweb/bedrock-wrapper/blob/main/test-converse-output.txt

This is an example of a successful Converse API response for Claude-3-Haiku.

```json
{
  "output": {
    "message": {
      "role": "assistant",
      "content": [
        {
          "text": "Two."
        }
      ]
    }
  },
  "stopReason": "end_turn",
  "usage": {
    "inputTokens": 21,
    "outputTokens": 4,
    "totalTokens": 25
  },
  "metrics": {
    "latencyMs": 654
  },
  "$metadata": {
    "httpStatusCode": 200,
    "requestId": "1ad9346f-baa6-4ade-9f33-3245795f4cb8",
    "attempts": 1,
    "totalRetryDelay": 0
  }
}
```

--------------------------------

### Converse API Response Example (Claude-3-5-Haiku)

Source: https://github.com/jparkerweb/bedrock-wrapper/blob/main/test-converse-output.txt

This is an example of a successful Converse API response for Claude-3-5-Haiku.

```json
{
  "output": {
    "message": {
      "role": "assistant",
      "content": [
        {
          "text": "Two"
        }
      ]
    }
  },
  "stopReason": "end_turn",
  "usage": {
    "inputTokens": 21,
    "outputTokens": 4,
    "totalTokens": 25,
    "cacheReadInputTokens": 0,
    "cacheWriteInputTokens": 0
  },
  "metrics": {
    "latencyMs": 654
  },
  "$metadata": {
    "httpStatusCode": 200,
    "requestId": "1ad9346f-baa6-4ade-9f33-3245795f4cb8",
    "attempts": 1,
    "totalRetryDelay": 0
  }
}
```

--------------------------------

### BedrockWrapper Options Example

Source: https://github.com/jparkerweb/bedrock-wrapper/blob/main/_autodocs/types.md

Illustrates how to instantiate the bedrockWrapper with custom options, such as enabling logging and specifying the use of the Converse API.

```javascript
const options = {
  logging: true,        // See API requests/responses
  useConverseAPI: true  // Use unified Converse API
};

for await (const chunk of bedrockWrapper(awsCreds, request, options)) {
  console.log(chunk);
}
```

--------------------------------

### Invoke API Request Example

Source: https://github.com/jparkerweb/bedrock-wrapper/blob/main/test-converse-output.txt

This is an example of a request sent to the Invoke API for text generation.

```json
{
  "messages": [
    {
      "role": "user",
      "content": "Respond with exactly one word: What is 1+1?"
    },
    {
      "role": "assistant",
      "content": ""
    }
  ],
  "max_tokens": 800,
  "temperature": 0.2,
  "anthropic_version": "bedrock-2023-05-31"
}
```

--------------------------------

### Converse API Response Example (Gemma-3-27b)

Source: https://github.com/jparkerweb/bedrock-wrapper/blob/main/test-final-output.txt

This is an example of a successful non-streaming response from the Converse API for Gemma-3-27b.

```json
{
  "output": {
    "message": {
      "role": "assistant",
      "content": [
        {
          "text": "Two."
        }
      ]
    }
  },
  "stopReason": "end_turn",
  "usage": {
    "inputTokens": 22,
    "outputTokens": 3,
    "totalTokens": 25
  },
  "metrics": {
    "latencyMs": 230
  },
  "$metadata": {
    "httpStatusCode": 200,
    "requestId": "3c9153ea-d63b-4ee3-899f-0514757f6b11",
    "attempts": 1,
    "totalRetryDelay": 0
  }
}
```

--------------------------------

### Complete Bedrock Wrapper Configuration and Execution

Source: https://github.com/jparkerweb/bedrock-wrapper/blob/main/_autodocs/configuration.md

This snippet demonstrates setting up AWS credentials, defining a multi-modal request (text and image), configuring wrapper options for streaming, and executing the request using the bedrockWrapper function. It includes error handling and logs the complete response.

```javascript
import dotenv from 'dotenv';
import { bedrockWrapper } from 'bedrock-wrapper';

dotenv.config();

// AWS credentials
const awsCreds = {
  region: process.env.AWS_REGION,
  accessKeyId: process.env.AWS_ACCESS_KEY_ID,
  secretAccessKey: process.env.AWS_SECRET_ACCESS_KEY,
};

// Request configuration
const request = {
  messages: [
    { role: "system", content: "You are a helpful assistant." },
    {
      role: "user",
      content: [
        { type: "text", text: "Analyze this image:" },
        {
          type: "image_url",
          image_url: { url: "https://example.com/photo.jpg" }
        }
      ]
    }
  ],
  model: "Claude-3-5-Sonnet",
  max_tokens: 1024,
  stream: true,
  temperature: 0.7,
  stop_sequences: ["END"],
};

// Wrapper options
const options = {
  logging: false,        // Disable in production
  useConverseAPI: true   // Use Converse API
};

// Execute request
try {
  let response = "";
  for await (const chunk of bedrockWrapper(awsCreds, request, options)) {
    response += chunk;
    process.stdout.write(chunk);
  }
  console.log("\n✅ Complete response:", response);
} catch (error) {
  console.error("❌ Error:", error.message);
}
```

--------------------------------

### Converse API Request Example (Ministral-3-14b)

Source: https://github.com/jparkerweb/bedrock-wrapper/blob/main/test-converse-output.txt

Example of a request payload for the Converse API for the Ministral-3-14b model.

```json
{
  "modelId": "mistral.ministral-3-14b-instruct",
  "messages": [
    {
      "role": "user",
      "content": [
        {
          "text": "Respond with exactly one word: What is 1+1?"
        }
      ]
    }
  ],
  "inferenceConfig": {
    "maxTokens": 800,
    "temperature": 0.2
  }
}
```

--------------------------------

### Environment Variables Example

Source: https://github.com/jparkerweb/bedrock-wrapper/blob/main/_autodocs/types.md

Lists essential environment variables required for AWS authentication and optional variables for LLM generation parameters. These are typically loaded using a `.env` file.

```bash
AWS_REGION=us-west-2
AWS_ACCESS_KEY_ID=your_access_key
AWS_SECRET_ACCESS_KEY=your_secret_key
LLM_MAX_GEN_TOKENS=1024
LLM_TEMPERATURE=0.7
```

--------------------------------

### Converse API Response Example

Source: https://github.com/jparkerweb/bedrock-wrapper/blob/main/test-converse-output.txt

This is an example of a successful Converse API response for a text generation task.

```json
{
  "output": {
    "message": {
      "role": "assistant",
      "content": [
        {
          "text": "Two"
        }
      ]
    }
  },
  "stopReason": "end_turn",
  "usage": {
    "inputTokens": 21,
    "outputTokens": 4,
    "totalTokens": 25
  },
  "metrics": {
    "latencyMs": 597
  },
  "$metadata": {
    "httpStatusCode": 200,
    "requestId": "2ad19370-0f4c-4c28-93ea-b789541e1220",
    "attempts": 1,
    "totalRetryDelay": 0
  }
}
```

--------------------------------

### Run All Test Suites

Source: https://github.com/jparkerweb/bedrock-wrapper/blob/main/README.md

Execute all available test suites within the package.

```bash
npm run test:all
```

--------------------------------

### Set up Environment Variables for AWS Credentials

Source: https://github.com/jparkerweb/bedrock-wrapper/blob/main/_autodocs/configuration.md

Create a .env file in your project root to store AWS credentials and optional model parameters. Load these variables using the dotenv package.

```bash
# AWS Credentials (required)
AWS_REGION=us-west-2
AWS_ACCESS_KEY_ID=your_access_key_id
AWS_SECRET_ACCESS_KEY=your_access_key_id

# Optional: Common model parameters (not used by library, but useful for examples)
LLM_MAX_GEN_TOKENS=1024
LLM_TEMPERATURE=0.7
```

--------------------------------

### Converse API Response Example (Ministral-3-14b)

Source: https://github.com/jparkerweb/bedrock-wrapper/blob/main/test-converse-output.txt

Example of a successful non-streaming response from the Converse API for the Ministral-3-14b model.

```json
{
  "output": {
    "message": {
      "role": "assistant",
      "content": [
        {
          "text": "Two."
        }
      ]
    }
  },
  "stopReason": "end_turn",
  "usage": {
    "inputTokens": 17,
    "outputTokens": 3,
    "totalTokens": 20
  },
  "metrics": {
    "latencyMs": 224
  },
  "$metadata": {
    "httpStatusCode": 200,
    "requestId": "dc28cf9c-8bfd-4749-8900-f7dc1f05a710",
    "attempts": 1,
    "totalRetryDelay": 0
  }
}
```

--------------------------------

### CLI Usage for Supported Models

Source: https://github.com/jparkerweb/bedrock-wrapper/blob/main/_autodocs/api-reference/listBedrockWrapperSupportedModels.md

Shows how to use the function in a command-line interface context, outputting the parsed model objects in a pretty-printed JSON format.

```javascript
import { listBedrockWrapperSupportedModels } from "bedrock-wrapper";

const models = await listBedrockWrapperSupportedModels();
console.log(JSON.stringify(models.map(m => JSON.parse(m)), null, 2));
```

--------------------------------

### Converse API Response Example (Mistral-3-8b)

Source: https://github.com/jparkerweb/bedrock-wrapper/blob/main/test-converse-output.txt

Example of a successful non-streaming response from the Converse API for the Mistral-3-8b model.

```json
{
  "output": {
    "message": {
      "role": "assistant",
      "content": [
        {
          "text": "Two."
        }
      ]
    }
  },
  "stopReason": "end_turn",
  "usage": {
    "inputTokens": 17,
    "outputTokens": 3,
    "totalTokens": 20
  },
  "metrics": {
    "latencyMs": 197
  },
  "$metadata": {
    "httpStatusCode": 200,
    "requestId": "7eca7cf4-d5a7-4135-b152-8542f933d29a",
    "attempts": 1,
    "totalRetryDelay": 0
  }
}
```

--------------------------------

### Converse API Request Example

Source: https://github.com/jparkerweb/bedrock-wrapper/blob/main/test-converse-output.txt

Example of a request payload for the Converse API, specifying model, messages, and inference configuration.

```json
{
  "modelId": "mistral.ministral-3-8b-instruct",
  "messages": [
    {
      "role": "user",
      "content": [
        {
          "text": "Respond with exactly one word: What is 1+1?"
        }
      ]
    }
  ],
  "inferenceConfig": {
    "maxTokens": 800,
    "temperature": 0.2
  }
}
```

--------------------------------

### Example Message with Images

Source: https://github.com/jparkerweb/bedrock-wrapper/blob/main/_autodocs/types.md

Demonstrates how to include both text and image content within a user message. Supports both base64 encoded images and image URLs.

```javascript
{
  role: "user",
  content: [
    { type: "text", text: "Analyze these images:" },
    { type: "image_url", image_url: { url: "data:image/jpeg;base64,/9j/4AAQSkZJRgA..." } },
    { type: "image_url", image_url: { url: "https://example.com/photo.jpg" } }
  ]
}
```

--------------------------------

### Converse API Request Example (Claude-3-Haiku)

Source: https://github.com/jparkerweb/bedrock-wrapper/blob/main/test-converse-output.txt

This is an example of a request sent to the Converse API for text generation using Claude-3-Haiku.

```json
{
  "modelId": "us.anthropic.claude-3-haiku-20240307-v1:0",
  "messages": [
    {
      "role": "user",
      "content": [
        {
          "text": "Respond with exactly one word: What is 1+1?"
        }
      ]
    }
  ],
  "inferenceConfig": {
    "maxTokens": 800,
    "temperature": 0.2
  }
}
```

--------------------------------

### Converse API Request Example (Claude-3-5-Haiku)

Source: https://github.com/jparkerweb/bedrock-wrapper/blob/main/test-converse-output.txt

This is an example of a request sent to the Converse API for text generation using Claude-3-5-Haiku.

```json
{
  "modelId": "us.anthropic.claude-3-5-haiku-20241022-v1:0",
  "messages": [
    {
      "role": "user",
      "content": [
        {
          "text": "Respond with exactly one word: What is 1+1?"
        }
      ]
    }
  ],
  "inferenceConfig": {
    "maxTokens": 800,
    "temperature": 0.2
  }
}
```

--------------------------------

### Basic Usage of listBedrockWrapperSupportedModels

Source: https://github.com/jparkerweb/bedrock-wrapper/blob/main/_autodocs/api-reference/listBedrockWrapperSupportedModels.md

Demonstrates how to call the function and log the total number of supported models. It also shows how to parse each stringified model entry to access its name and ID.

```javascript
import { listBedrockWrapperSupportedModels } from "bedrock-wrapper";

const models = await listBedrockWrapperSupportedModels();
console.log(`${models.length} models supported`);

// Parse each stringified entry
models.forEach(modelStr => {
  const model = JSON.parse(modelStr);
  console.log(`${model.modelName} → ${model.modelId}`);
});
```

--------------------------------

### Configure AWS Credentials

Source: https://github.com/jparkerweb/bedrock-wrapper/blob/main/_autodocs/README.md

Set up AWS credentials and region using a .env file and load them with dotenv.

```bash
AWS_REGION=us-west-2
AWS_ACCESS_KEY_ID=your_key
AWS_SECRET_ACCESS_KEY=your_secret
```

```javascript
import dotenv from 'dotenv';
dotenv.config();
```

--------------------------------

### Configure OpenAI Chat Completions Object

Source: https://github.com/jparkerweb/bedrock-wrapper/blob/main/README.md

Prepare an object for chat completions, mirroring OpenAI's format. Specify model, max tokens, streaming, and temperature. Optional stop sequences can be provided.

```javascript
const openaiChatCompletionsCreateObject = {
    "messages": messages,
    "model": "Claude-4-5-Sonnet",
    "max_tokens": LLM_MAX_GEN_TOKENS,
    "stream": true,
    "temperature": LLM_TEMPERATURE,
    "stop_sequences": ["STOP", "END"], // Optional: sequences that will stop generation
};
```

--------------------------------

### With Vision (Image Analysis)

Source: https://github.com/jparkerweb/bedrock-wrapper/blob/main/_autodocs/api-reference/bedrockWrapper.md

Demonstrates how to use bedrockWrapper for image analysis by providing an image URL in the request.

```APIDOC
## With Vision (Image Analysis)

### Description
This example shows how to perform image analysis by including an image URL within the `messages` content.

### Method
`bedrockWrapper(awsCreds, request)`

### Parameters
- **awsCreds** (object) - AWS credentials and region.
- **request** (object) - Configuration for the model request, including a `content` array with `type: "image_url"`.

### Request Example
```javascript
const request = {
  messages: [
    {
      role: "user",
      content: [
        { type: "text", text: "Describe this image:" },
        {
          type: "image_url",
          image_url: { url: "https://example.com/image.jpg" }
        }
      ]
    }
  ],
  model: "Claude-3-5-Sonnet",  // Must have vision: true
  max_tokens: 500,
  stream: true,
  temperature: 0.5,
};

for await (const chunk of bedrockWrapper(awsCreds, request)) {
  process.stdout.write(chunk);
}
```

### Response
- **chunk** (string) - Streaming response chunks containing the image description.
```

--------------------------------

### Invoke API Request Example (DeepSeek V3.1)

Source: https://github.com/jparkerweb/bedrock-wrapper/blob/main/test-run-output.txt

This is an example of a request sent to the Invoke API for the DeepSeek V3.1 model, similar to the Converse API request structure.

```json
{
  "modelId": "deepseek.v3-v1:0",
  "messages": [
    {
      "role": "user",
      "content": [
        {
          "text": "Respond with exactly one word: What is 1+1?"
        }
      ]
    }
  ],
  "inferenceConfig": {
    "maxTokens": 800,
    "temperature": 0.2
  }
}
```

--------------------------------

### Basic Streaming with Claude

Source: https://github.com/jparkerweb/bedrock-wrapper/blob/main/_autodocs/api-reference/bedrockWrapper.md

Demonstrates a basic streaming request to a Claude model using bedrockWrapper.

```APIDOC
## Basic Streaming with Claude

### Description
This example shows how to initiate a streaming request to a Claude model.

### Method
`bedrockWrapper(awsCreds, request)`

### Parameters
- **awsCreds** (object) - AWS credentials and region.
- **request** (object) - Configuration for the model request, including messages, model name, max tokens, and stream: true.

### Request Example
```javascript
import { bedrockWrapper } from "bedrock-wrapper";

const awsCreds = {
  region: "us-west-2",
  accessKeyId: process.env.AWS_ACCESS_KEY_ID,
  secretAccessKey: process.env.AWS_SECRET_ACCESS_KEY,
};

const request = {
  messages: [
    { role: "system", content: "You are a helpful assistant." },
    { role: "user", content: "Explain quantum computing in 2 sentences." }
  ],
  model: "Claude-3-5-Sonnet",
  max_tokens: 200,
  stream: true,
  temperature: 0.7,
};

let response = "";
for await (const chunk of bedrockWrapper(awsCreds, request)) {
  response += chunk;
  process.stdout.write(chunk);
}
console.log("\nComplete:", response);
```

### Response
- **chunk** (string) - Streaming response chunks from the model.
```