Description

n8n-nodes-aws-bedrock-assumerole

An n8n community node for AWS Bedrock with AssumeRole authentication support.

!n8n.io – Workflow Automation

Features

• 🔐 AssumeRole Authentication: Secure cross-account access using AWS STS AssumeRole
• 🤖 Multiple Claude Models: Support for Claude 3.5 Sonnet, Claude 3 Opus, Sonnet, Haiku, and more
• 🎨 Image Generation: Support for Amazon Nova Canvas and Titan Image Generator models
• 🤝 AI Agent Compatible: Includes Chat Model sub-node for use with n8n AI Agent
• ⚡ Credential Caching: Automatic caching of temporary credentials with expiration handling
• 🛡️ Error Handling: Comprehensive error handling and logging
• 🔄 Batch Processing: Process multiple items in a single workflow execution
• 📊 Usage Tracking: Detailed usage information and response metadata

Available Nodes

This package includes two nodes:

1. AWS Bedrock (AssumeRole) – Standalone node for direct AWS Bedrock API calls
2. AWS Bedrock Chat Model – Chat Model sub-node for use with n8n AI Agent

Supported Models

This node uses AWS Bedrock inference profiles for optimal performance and availability:

Text/Chat Models (Claude)

• Claude 3.5 Sonnet v2 – us.anthropic.claude-3-5-sonnet-20241022-v2:0 (default)
• Claude 3.5 Sonnet v1 – us.anthropic.claude-3-5-sonnet-20240620-v1:0
• Claude 3.5 Haiku – us.anthropic.claude-3-5-haiku-20241022-v1:0
• Claude 3.7 Sonnet – us.anthropic.claude-3-7-sonnet-20250219-v1:0
• Claude Sonnet 4 – us.anthropic.claude-sonnet-4-20250514-v1:0
• Claude Sonnet 4.5 – us.anthropic.claude-sonnet-4-5-20250929-v1:0
• Claude Haiku 4.5 – us.anthropic.claude-haiku-4-5-20251001-v1:0
• Claude Opus 4 – us.anthropic.claude-opus-4-20250514-v1:0
• Claude Opus 4.1 – us.anthropic.claude-opus-4-1-20250805-v1:0

Image Generation Models

• Amazon Nova Canvas v1 – amazon.nova-canvas-v1:0 – State-of-the-art image generation
• Amazon Titan Image Generator v2 – amazon.titan-image-generator-v2:0 – High-quality image generation with advanced controls

Installation

Option 1: Install from npm (Recommended)

Install globally for n8n
npm install -g n8n-nodes-aws-bedrock-assumerole
Or install locally in your n8n custom nodes directory
cd ~/.n8n/custom/
npm install n8n-nodes-aws-bedrock-assumerole

Option 2: Install from source

Clone the repository
git clone https://github.com/cabify/n8n-nodes-aws-bedrock-assumerole.git
cd n8n-nodes-aws-bedrock-assumerole
Install dependencies
npm install
Build the project
npm run build
Link for local development
npm link
In your n8n installation directory
npm link n8n-nodes-aws-bedrock-assumerole

Configuration

1. AWS Credentials Setup

You have two options for providing AWS credentials:

#### Option A: Environment Variables (Recommended)
Set these environment variables on your n8n server:

export AWSACCESSKEY_ID="your-access-key-id"
export AWSSECRETACCESS_KEY="your-secret-access-key"
export AWS_REGION="us-east-1"

#### Option B: Credential Fields
Fill in the credential fields directly in the n8n UI (less secure).

2. Create AWS AssumeRole Credential

1. Go to Credentials in your n8n instance
2. Click Add Credential
3. Search for “AWS Assume Role”
4. Configure the following:
– Access Key ID: Leave empty to use environment variable (recommended)
– Secret Access Key: Leave empty to use environment variable (recommended)
– Role ARN to Assume: arn:aws:iam:::role/
– AWS Region: us-east-1 (or your preferred region)
– Session Duration: 3600 (1 hour, adjust as needed)

3. AWS IAM Setup

#### Base Account Role/User Permissions
The base AWS credentials need the following permission:

{
    "Version": "2012-10-17",
    "Statement": [
        {
            "Effect": "Allow",
            "Action": "sts:AssumeRole",
            "Resource": "arn:aws:iam:::role/"
        }
    ]
}

#### Target Account Role
The role to be assumed needs:

{
    "Version": "2012-10-17",
    "Statement": [
        {
            "Effect": "Allow",
            "Action": [
                "bedrock:InvokeModel"
            ],
            "Resource": [
                "arn:aws:bedrock:::foundation-model/anthropic."
            ]
        }
    ]
}

And the trust relationship:

{
    "Version": "2012-10-17",
    "Statement": [
        {
            "Effect": "Allow",
            "Principal": {
                "AWS": "arn:aws:iam:::role/"
            },
            "Action": "sts:AssumeRole"
        }
    ]
}

4. Application Inference Profiles

This node supports AWS Bedrock Application Inference Profiles, allowing you to route traffic through specific profiles for cost and usage tracking.

#### 4.1. Credential configuration

In the AWS AssumeRole credential, you can optionally configure:

• Application Inference Profile Account ID: The AWS account ID where your application inference profiles live.
• Application Inference Profiles JSON: A JSON object mapping Bedrock model IDs to application inference profile IDs.

Example JSON:

{
  "us.anthropic.claude-3-5-sonnet-20240620-v1:0": "hs4uvikaus5b",
  "us.anthropic.claude-3-5-sonnet-20241022-v2:0": "0xumpou8xusv",
  "us.anthropic.claude-3-5-haiku-20241022-v1:0": "abc123haiku"
}

• The key is the Bedrock model ID (for example, us.anthropic.claude-3-5-sonnet-20241022-v2:0).
• The value is the application inference profile ID (for example, 0xumpou8xusv), not the full ARN.

The node then builds the final ARN internally using:

arn:aws:bedrock:{region}:{account-id}:application-inference-profile/{profile-id}

If the JSON is invalid, the node will fail with a clear error message pointing to the Application Inference Profiles JSON field.

#### 4.2. Model dropdown behaviour

The Model ID dropdown in the node behaves as follows:

• If Application Inference Profiles JSON is empty or not set:

– The dropdown shows all supported Claude models (the default static list).

• If Application Inference Profiles JSON is present and valid:

– The dropdown shows only the models present in that JSON.
– Known model IDs are displayed with friendly names (for example, “Claude 3.5 Sonnet v2”), unknown ones are shown as their raw model ID.

This ensures that, when you configure specific models and profiles in the credential, users of the node can only select those models.

#### 4.3. Backwards compatibility

If no application inference profile mapping is found for a selected model ID, the node will:

1. Try the legacy single Application Inference Profile ID field (if configured).
2. Otherwise, fall back to using the raw model ID directly (original behaviour).

Usage

Option 1: Using with AI Agent (Recommended for Conversational AI)

The AWS Bedrock Chat Model node is designed to work with n8n’s AI Agent node, enabling conversational AI workflows with tool calling, memory, and more.

#### Setup Steps:

1. Add an AI Agent node to your workflow
2. Connect the AWS Bedrock Chat Model node to the “Chat Model” input of the AI Agent
3. Select your credential in the Chat Model node (the same AWS AssumeRole credential)
4. Choose your model (e.g., Claude 3.5 Sonnet v2)
5. Add tools (optional): Connect tool nodes like Vector Store, Calculator, HTTP Request, etc.
6. Add memory (optional): Connect a memory node for conversation history

#### Benefits of Using with AI Agent:

• ✅ Tool Calling: The AI can use tools to fetch data, perform calculations, etc.
• ✅ Conversation Memory: Maintain context across multiple interactions
• ✅ Structured Output: Parse responses into structured data
• ✅ Multi-step Reasoning: The agent can plan and execute complex tasks

Option 2: Direct API Calls (Standalone Node)

For simple, direct API calls without AI Agent features, use the AWS Bedrock (AssumeRole) node.

#### Basic Workflow Example

1. Add the AWS Bedrock (AssumeRole) node to your workflow
2. Select your credential (created in step 2 above)
3. Configure the node:
– Model ID: Choose from the dropdown (e.g., Claude 3.5 Sonnet)
– Prompt: Enter your prompt or use an expression to get it from previous nodes
– Max Tokens: Set the maximum response length (default: 1000)
– Temperature: Control randomness (0.0 = deterministic, 1.0 = very random)

#### Example Prompt

Analyze the following customer feedback and provide:
1. Sentiment (positive/negative/neutral)
2. Key themes
3. Suggested actions
Customer feedback: "The service was okay but the wait time was too long."

Image Analysis Workflow (Text and Image input)

Note: Image analysis is currently only available with the standalone AWS Bedrock (AssumeRole) node, not with the Chat Model sub-node.

To analyze an image together with a text prompt using Claude models that support vision capabilities:

1. Add a Form Trigger (or any node that outputs binary data) with a file field, for example labeled imagetoanalize.
2. Connect that node to AWS Bedrock (AssumeRole).
3. Configure the Bedrock node:
– Model ID: Select any Claude model that supports image input (for example, Claude Sonnet 4).
– Input Type: Set to Text and Image.
– Image Binary Property: Set to the name of the binary field that contains the uploaded image. For a Form Trigger file field labeled imagetoanalize, the binary key is also imagetoanalize.
– Prompt: Provide the instruction you want to send together with the image, for example: Describe what is written in this image.
4. Execute the workflow by submitting the form with an image file.

You can import the ready-to-use example workflow from examples/image-analysis-workflow.json.

Image Generation Workflow (Nova Canvas / Titan Image)

Generate images from text prompts using Amazon Nova Canvas or Titan Image Generator models:

1. Add the AWS Bedrock (AssumeRole) node to your workflow.
2. Configure the Bedrock node:
– Model ID: Select Amazon Nova Canvas v1 or Amazon Titan Image Generator v2.
– Prompt: Describe the image you want to generate (e.g., “A futuristic city at sunset with flying cars”).
– Negative Prompt (optional): Describe what NOT to include (e.g., “blurry, low quality, text”).
– Image Width/Height: Choose the dimensions (512, 768, 1024, or 1280 pixels).
– Image Quality: Select standard or premium.
– Number of Images: Generate 1-4 images at once.
– Seed (optional): Set a specific seed for reproducible results (0 = random).
– CFG Scale (Titan Image only): Controls how closely the image follows the prompt (1-15).

3. The node outputs binary image data that can be:
– Saved to disk using the Write Binary File node
– Uploaded to cloud storage (S3, Google Drive, etc.)
– Sent via email or messaging platforms
– Further processed in your workflow

#### Image Generation Response Format

For image generation models, the node returns:

{
  "modelId": "arn:aws:bedrock:us-east-1:123456789:application-inference-profile/abc123",
  "configuredModelId": "amazon.nova-canvas-v1:0",
  "prompt": "A futuristic city at sunset",
  "imageIndex": 0,
  "totalImages": 1,
  "imageWidth": 1024,
  "imageHeight": 1024,
  "imageQuality": "standard",
  "timestamp": "2026-01-08T10:00:00.000Z"
}

The generated image is available in the binary.data property as a PNG file.

Image Editing Workflow (Inpainting, Outpainting, Variations, Background Removal)

Both Nova Canvas and Titan Image Generator support advanced image editing capabilities:

#### Image Task Types

| Task Type | Description | Required Fields |
|———–|————-|—————–|
| Text to Image | Generate a new image from a text prompt | Prompt |
| Inpainting | Modify areas inside a masked region | Source Image, Mask (prompt or image), Prompt |
| Outpainting | Extend or modify areas outside a masked region | Source Image, Mask (prompt or image), Prompt |
| Image Variation | Create variations of an existing image | Source Image, Prompt (optional) |
| Background Removal | Remove the background (outputs transparent PNG) | Source Image |

#### Inpainting Example

Replace part of an image based on a text description of the area to modify:

1. Add a node that provides an image (e.g., Read Binary File, HTTP Request, or Form Trigger).
2. Add the AWS Bedrock (AssumeRole) node.
3. Configure:
– Model ID: Select Amazon Nova Canvas v1 or Amazon Titan Image Generator v2
– Image Task Type: Select Inpainting (Edit Inside Mask)
– Source Image Binary Property: data (or the name of your binary property)
– Mask Prompt: Describe the area to modify (e.g., “the sky”, “the person’s shirt”)
– Prompt: Describe what to put in that area (e.g., “a beautiful sunset sky”)
– Negative Prompt (optional): What to avoid

#### Outpainting Example

Extend an image beyond its original boundaries:

1. Provide a source image.
2. Configure:
– Image Task Type: Select Outpainting (Edit Outside Mask)
– Mask Prompt: Describe the area to preserve (e.g., “the main subject”)
– Prompt: Describe what to generate in the extended area
– Outpainting Mode: Default (allows blending) or Precise (strict boundary)

#### Image Variation Example

Create variations of an existing image:

1. Provide a source image.
2. Configure:
– Image Task Type: Select Image Variation
– Similarity Strength: 0.2 (more variation) to 1.0 (more similar to original)
– Prompt (optional): Guide the variation direction

#### Background Removal Example

Remove the background from an image (outputs transparent PNG):

1. Provide a source image.
2. Configure:
– Image Task Type: Select Background Removal
– No prompt needed – the model automatically detects and removes the background

#### Mask Options

For Inpainting and Outpainting, you can specify the mask in two ways:

1. Mask Prompt (recommended): A text description of the area to mask (e.g., “the sky”, “the person’s face”)
2. Mask Image: A binary black/white image where:
– Black pixels = area to modify
– White pixels = area to preserve

If both are provided, the Mask Image takes precedence.

#### AWS Documentation

For more details on image editing capabilities, see:

• Amazon Nova Canvas User Guide
• Amazon Titan Image Generator User Guide

#### Application Inference Profiles for Image Models

Configure image generation models in your credentials JSON just like Claude models:

{
  "us.anthropic.claude-3-5-sonnet-20241022-v2:0": "0xumpou8xusv",
  "amazon.nova-canvas-v1:0": "b3tcu2bezmae",
  "amazon.titan-image-generator-v2:0": "12fut6sh2vgi"
}

Response Format (Standalone Node – Text Models)

The AWS Bedrock (AssumeRole) standalone node returns a JSON object with:

{
  "modelId": "us.anthropic.claude-3-5-sonnet-20241022-v2:0",
  "prompt": "Your original prompt",
  "response": {
    "content": [
      {
        "text": "The AI response text",
        "type": "text"
      }
    ],
    "usage": {
      "input_tokens": 25,
      "output_tokens": 150
    }
  },
  "usage": {
    "input_tokens": 25,
    "output_tokens": 150
  },
  "content": "The AI response text",
  "timestamp": "2024-11-12T17:46:00.000Z"
}

Comparison: Chat Model vs Standalone Node

| Feature | AWS Bedrock Chat Model | AWS Bedrock (AssumeRole) |
|———|————————|————————–|
| Use Case | AI Agent workflows | Direct API calls |
| Tool Calling | ✅ Yes (via AI Agent) | ❌ No |
| Conversation Memory | ✅ Yes (via AI Agent) | ❌ No |
| Image Analysis | ❌ Not yet supported | ✅ Yes |
| Image Generation | ❌ No | ✅ Yes (Nova Canvas, Titan Image) |
| Batch Processing | ❌ No | ✅ Yes |
| Structured Output | ✅ Yes (via AI Agent) | ⚠️ Manual parsing |
| Best For | Conversational AI, agents with tools | Simple prompts, image analysis/generation, batch jobs |

Development

Prerequisites

• Node.js 18+
• npm or yarn
• TypeScript

Setup

Clone the repository
git clone https://github.com/cabify/n8n-nodes-aws-bedrock-assumerole.git
cd n8n-nodes-aws-bedrock-assumerole
Install dependencies
npm install
Build the project
npm run build
Run linting
npm run lint
Run tests
npm test

Project Structure

n8n-nodes-aws-bedrock-assumerole/
├── credentials/
│   └── AwsAssumeRole.credentials.ts    # AWS AssumeRole credential definition
├── nodes/
│   └── AwsBedrockAssumeRole.node.ts    # Main node implementation
├── icons/
│   ├── aws.svg                         # AWS credential icon
│   └── bedrock.svg                     # Bedrock node icon
├── dist/                               # Compiled JavaScript (generated)
├── package.json                        # Package configuration
├── tsconfig.json                       # TypeScript configuration
├── .eslintrc.js                        # ESLint configuration
├── .prettierrc                         # Prettier configuration
└── README.md                           # This file

Troubleshooting

Common Issues

#### 1. “Missing AWS base credentials”

• Ensure AWSACCESSKEYID and AWSSECRETACCESSKEY are set as environment variables
• Or fill in the credential fields in the n8n UI

#### 2. “AssumeRole failed”

• Verify the Role ARN is correct
• Check that the base credentials have sts:AssumeRole permission
• Ensure the target role trusts the base account/role

#### 3. “Access Denied” when invoking Bedrock

• Verify the assumed role has bedrock:InvokeModel permission
• Check that the model ID is available in your AWS region
• Ensure Bedrock is enabled in your AWS account

#### 4. Node not appearing in n8n

• Restart n8n after installation
• Check that the package is installed in the correct location
• Verify the package.json n8n configuration is correct

Debug Logging

The node provides detailed console logging. Check your n8n logs for:

• [AWS Bedrock] Resolved credentials
• [AWS Bedrock] AssumeRole successful
• [AWS Bedrock] Model response received

Developers

This project is developed and maintained by:

• @fluty84 – Lead Developer & Business Automation
• @cHiv0rz – Infrastructure Support

Development

Quick Start with Make

This project includes a Makefile for easy development and deployment:

Show all available commands
make help
Development
make install      # Install dependencies
make build        # Build the project
make dev          # Build and start Docker for local testing
make clean        # Clean build artifacts
Docker
make docker-up    # Start Docker containers
make docker-down  # Stop Docker containers
make docker-logs  # Show Docker logs
Deployment
make publish      # Publish to npm (interactive)
make sync         # Sync repositories (GitHub + GitLab)
make release      # Full release: build + publish + sync

Publishing a New Version

The make publish command provides an interactive workflow that handles everything:

#### Step 1: Version Bump
Choose the type of version bump:

• patch (1.0.2 → 1.0.3) – Bug fixes
• minor (1.0.2 → 1.1.0) – New features (backwards compatible)
• major (1.0.2 → 2.0.0) – Breaking changes
• custom – Specify version manually

#### Step 2: Changelog Generation
Select the types of changes included:
1. Added – New features
2. Changed – Changes in existing functionality
3. Deprecated – Soon-to-be removed features
4. Removed – Removed features
5. Fixed – Bug fixes
6. Security – Security fixes

#### Step 3: Changelog Entries
Enter detailed changes for each selected section. The script will automatically:

• Update CHANGELOG.md with proper formatting
• Follow Keep a Changelog format
• Add the current date
• Insert the new entry at the top

#### Step 4: Build & Publish
The script will:

• Build the project (npm run build)
• Publish to npm with public access
• Commit changes to package.json, package-lock.json, and CHANGELOG.md
• Create a git tag (e.g., v1.0.2)

#### Example Workflow

Start the publish process
make publish
Follow the prompts:
1. Select version bump: 1 (patch)
2. Select change types: 5 (Fixed)
3. Enter changes:
   - Fixed custom SVG icons not displaying correctly
   - Removed unused code and imports
4. Confirm publish: y
After publishing, sync repositories
make sync
Or do everything in one command:
make release

Repository Sync

The project supports syncing to multiple repositories:

• GitHub: https://github.com/cabify/n8n-nodes-aws-bedrock-assumerole
• GitLab: https://gitlab.otters.xyz/platform/business-automation/n8n-nodes-aws-bedrock-assumerole

The make sync command will:

• Push code to both GitHub and GitLab
• Push all tags to both repositories
• Verify you’re on the main branch
• Show current status before pushing

Manual Development

If you prefer not to use Make:

Install dependencies
npm install
Build
npm run build
Start Docker for testing
docker-compose up -d
View logs
docker-compose logs -f n8n
Publish manually
npm version patch  # or minor, major
npm run build
npm publish --access public
git push && git push --tags

Project Structure

n8n-bedrock-node/
├── credentials/
│   ├── AwsAssumeRole.credentials.ts  # Credential definition
│   └── aws.svg                        # AWS icon
├── nodes/
│   ├── AwsBedrockAssumeRole.node.ts  # Main node implementation
│   └── bedrock.svg                    # Bedrock icon
├── icons/
│   ├── aws.svg                        # Source AWS icon
│   └── bedrock.svg                    # Source Bedrock icon
├── dist/                              # Compiled output
├── docker-compose.yml                 # Local development setup
├── Makefile                           # Development commands
├── publish-npm.sh                     # npm publish script
├── sync-repos.sh                      # Repository sync script
└── package.json                       # Package configuration

Scripts

• npm run build – Compile TypeScript and copy icons
• npm run copy-icons – Copy icons to dist directories
• npm run lint – Run ESLint (requires setup)
• npm test – Run tests (if available)

Contributing

1. Fork the repository
2. Create a feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'Add some amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

License

This project is licensed under the MIT License – see the LICENSE file for details.

Support

• 📧 Email: business-automation@cabify.com
• 🐛 Issues: GitHub Issues
• 📖 n8n Documentation: n8n.io/docs

Acknowledgments

• Built for the n8n workflow automation platform
• Uses AWS SDK v3 for optimal performance
• Inspired by the need for secure cross-account AWS Bedrock access