Bedrock Gateway Client

A Python client for AWS Bedrock Gateway with IAM authentication. Works with any private API Gateway setup for AWS Bedrock, featuring both Python API and CLI interface.

Features

• ✅ Configurable - Works with any Bedrock Gateway deployment
• ✅ No secrets required - Uses your AWS IAM credentials
• ✅ Simple API - Just client.chat("message")
• ✅ Multiple configuration methods - Constructor, environment variables, or config file
• ✅ CLI included - Command-line interface for quick testing

Installation

pip install bedrock-gateway-client==1.0.3

Requirements:

• Python 3.8+
• boto3>=1.26.0
• requests>=2.28.0
• pyyaml>=6.0
• click>=8.0.0

Quick Start

Python API

from bedrock_gateway_client import BedrockClient

client = BedrockClient(
    gateway_url="https://your-api-gateway.execute-api.region.amazonaws.com/prod/invoke",
    api_id="your-api-id",  # Your API Gateway ID
    region="your-region"
)

response = client.chat("Tell me about koalas")
print(response.text)
print(f"Tokens used: {response.tokens}")

CLI Interface

# Configure once
bedrock-gateway configure --gateway-url "https://your-api-gateway.execute-api.region.amazonaws.com/prod/invoke" --api-id "your-api-id" --region your-region

# Chat with Claude
bedrock-gateway chat "Tell me about koalas"

Configuration Options

Direct Configuration

from bedrock_gateway_client import BedrockClient

client = BedrockClient(
    gateway_url="https://your-api-gateway.execute-api.region.amazonaws.com/prod/invoke",
    api_id="your-api-id",  # Your API Gateway ID
    region="your-region"
)

Environment Variables

export BEDROCK_GATEWAY_URL="https://your-api-gateway.execute-api.region.amazonaws.com/prod/invoke"
export BEDROCK_GATEWAY_API_ID="your-api-id"
export BEDROCK_GATEWAY_REGION="your-region"

from bedrock_gateway_client import BedrockClient

client = BedrockClient()  # Uses environment variables

CLI Configuration

# Save configuration to file
bedrock-gateway configure \
  --gateway-url "https://your-api-gateway.execute-api.region.amazonaws.com/prod/invoke" \
  --api-id "your-api-id" \
  --region your-region

Global Configuration

from bedrock_gateway_client import configure, chat

configure(
    gateway_url="https://your-api-gateway.execute-api.region.amazonaws.com/prod/invoke",
    api_id="your-api-id",
    region="your-region",
    save=True
)

# Now use simple chat function
response = chat("Tell me about koalas")
print(response)

Usage Examples

Basic Chat

response = client.chat(
    "What is AWS Bedrock?",
    model="sonnet-4.5",
    max_tokens=500
)

print(response.text)
print(f"Tokens: {response.tokens}")
print(f"Latency: {response.latency_ms}ms")

With System Prompt

response = client.chat(
    "What is 2+2?",
    system="You are a helpful math tutor.",
    model="sonnet-4.5"
)

Multi-turn Conversation

# First turn
response1 = client.chat("My name is Alice")

# Build history
history = [
    {"role": "user", "content": [{"text": "My name is Alice"}]},
    {"role": "assistant", "content": [{"text": response1.text}]}
]

# Second turn with context
response2 = client.chat(
    "What's my name?",
    conversation_history=history
)

Command-Line Interface

# Show help
bedrock-gateway --help

# Configure your Bedrock Gateway
bedrock-gateway configure \
  --gateway-url "https://your-api-gateway.execute-api.region.amazonaws.com/prod/invoke" \
  --api-id "your-api-id" \
  --region your-region

# Chat with Claude
bedrock-gateway chat "Tell me about koalas"

# Advanced chat with options
bedrock-gateway chat "Explain quantum computing" \
  --model sonnet-4.5 \
  --max-tokens 1000

# Check your AWS identity
bedrock-gateway whoami

Available Models

• sonnet-4.5 - Best for detailed, complex conversations
• haiku-4.5 - Fast and efficient for simple tasks

Configuration File

Configuration is saved to ~/.bedrock-gateway/config.yaml:

api_id: your-api-id
gateway_url: https://your-api-gateway.execute-api.region.amazonaws.com/prod/invoke
region: your-region
model_map:
  sonnet-4.5: anthropic.claude-sonnet-4-5-v1:0
  haiku-4.5: anthropic.claude-haiku-4-5-v1:0

Response Object

response.text           # The generated text
response.tokens         # Total tokens used
response.input_tokens   # Input tokens
response.output_tokens  # Output tokens
response.latency_ms     # Response time in milliseconds
response.request_id     # Unique request ID
response.model          # Model used
response.stop_reason    # Why generation stopped
response.raw_response   # Full raw response

Troubleshooting

Common Issues

1. 403 Forbidden Error If you get a 403 Forbidden error, check:

• Your AWS credentials have the required permissions
• The API Gateway resource policy allows your role
• Both gateway_url and api_id are configured correctly

# Verify configuration
bedrock-gateway whoami

# Test with verbose output
bedrock-gateway chat "Hello" --verbose

2. Authentication Errors Ensure your AWS credentials are properly configured:

# Check AWS credentials
aws sts get-caller-identity

# Configure AWS credentials if needed
aws configure

3. Gateway Not Found Make sure both gateway_url and api_id are set correctly in your configuration.

Authentication

Uses your AWS credentials automatically:

• Default AWS credentials (~/.aws/credentials)
• AWS profile: BedrockClient(profile="my-profile")
• Environment variables (AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY)
• IAM role (when running on EC2/Lambda/ECS)

Required AWS Permissions

Your IAM role needs these permissions:

{
  "Version": "2012-10-17",
  "Statement": [
    {
      "Effect": "Allow",
      "Action": [
        "execute-api:Invoke"
      ],
      "Resource": "arn:aws:execute-api:region:account-id:your-api-id/*/*"
    }
  ]
}

License

MIT License - see LICENSE file for details

Support

For issues, open a GitHub issue at: https://github.com/parv3sh/bedrock-gateway-client/issues