Anthropic Ruby SDK (anthropics/anthropic-sdk-ruby)

Anthropic Ruby SDK

https://github.com/anthropics/anthropic-sdk-ruby
Admin
The Claude SDK for Ruby provides access to the Claude API from Ruby applications, enabling...

Tokens:12,672
Snippets:106
Trust Score:8.8
Update:6 days ago
Show doc for...
Context Summary (auto-generated)
Raw
# Anthropic Ruby SDK

The Anthropic Ruby SDK provides access to the Claude API from Ruby applications, enabling developers to integrate Claude's AI capabilities into their projects. The SDK supports all major Claude features including message creation, streaming responses, tool use, structured outputs, extended thinking, web search, and image analysis. It works with the direct Anthropic API as well as AWS Bedrock and Google Cloud Vertex AI deployments.

The SDK is designed with a clean, idiomatic Ruby interface that supports both synchronous and streaming patterns. It includes helper classes for defining tools and input schemas, automatic retry handling with exponential backoff, and comprehensive type definitions for Sorbet and RBS. The library requires Ruby 3.2.0 or later and handles authentication, request building, and response parsing automatically.

## Client Initialization

Create a client instance to interact with the Claude API. The client handles authentication, retries, and connection management.

```ruby
require "anthropic"

# Initialize with API key from environment variable ANTHROPIC_API_KEY
client = Anthropic::Client.new

# Or provide credentials explicitly
client = Anthropic::Client.new(
  api_key: "your-api-key",
  base_url: "https://api.anthropic.com",  # Optional custom base URL
  max_retries: 2,                          # Max retry attempts (default: 2)
  timeout: 600.0,                          # Request timeout in seconds
  initial_retry_delay: 0.5,                # Initial retry delay
  max_retry_delay: 8.0                     # Max retry delay
)
```

## Messages API - Create Message

Send messages to Claude and receive a complete response. This is the primary method for non-streaming interactions.

```ruby
client = Anthropic::Client.new

# Simple message
response = client.messages.create(
  model: "claude-sonnet-4-5-20250929",
  max_tokens: 1024,
  messages: [{role: :user, content: "Hello, Claude!"}]
)

puts response.content.first.text
# => "Hello! How can I help you today?"

# Multi-turn conversation
response = client.messages.create(
  model: "claude-sonnet-4-5-20250929",
  max_tokens: 1024,
  messages: [
    {role: :user, content: "What is Ruby?"},
    {role: :assistant, content: "Ruby is a dynamic, object-oriented programming language..."},
    {role: :user, content: "Show me a simple example"}
  ],
  system_: "You are a helpful programming tutor.",
  temperature: 0.7
)

puts response.content.first.text
puts "Input tokens: #{response.usage.input_tokens}"
puts "Output tokens: #{response.usage.output_tokens}"
```

## Messages API - Streaming Responses

Stream responses in real-time for a better user experience with long outputs. The stream is an Enumerable that emits events as they arrive.

```ruby
client = Anthropic::Client.new

# Create streaming request
stream = client.messages.stream(
  model: "claude-sonnet-4-5-20250929",
  max_tokens: 1024,
  messages: [{role: :user, content: "Write a haiku about programming"}]
)

# Stream text as it arrives
stream.text.each { |text| print(text) }
puts

# Or process individual events
stream = client.messages.stream(
  model: "claude-sonnet-4-5-20250929",
  max_tokens: 1024,
  messages: [{role: :user, content: "Tell me a story"}]
)

stream.each do |event|
  case event
  when Anthropic::Streaming::TextEvent
    print(event.text)  # Incremental text
  when Anthropic::Streaming::MessageStopEvent
    puts "\n\nMessage complete!"
  end
end

# Get accumulated message after stream completes
final_message = stream.accumulated_message
puts "Total tokens: #{final_message.usage.output_tokens}"
```

## Tool Use - Manual Handling

Define tools that Claude can use, then handle tool calls manually. Tools let Claude interact with external systems and data.

```ruby
client = Anthropic::Client.new

# Define tool input schema
class GetWeatherInput < Anthropic::BaseModel
  required :location, String, doc: "The city and state, e.g. San Francisco, CA"
  required :unit, Anthropic::EnumOf[:celsius, :fahrenheit], nil?: true,
           doc: "Temperature unit"

  doc "Get the current weather in a given location"
end

# Send message with tools
message = client.messages.create(
  model: "claude-sonnet-4-5-20250929",
  max_tokens: 1024,
  messages: [{role: "user", content: "What's the weather in San Francisco?"}],
  tools: [GetWeatherInput]
)

# Check if Claude wants to use a tool
if message.stop_reason == :tool_use
  tool_use = message.content.grep(Anthropic::Models::ToolUseBlock).first
  puts "Tool: #{tool_use.name}"
  puts "Input: #{tool_use.input}"

  # Execute tool and return result
  response = client.messages.create(
    model: "claude-sonnet-4-5-20250929",
    max_tokens: 1024,
    messages: [
      {role: "user", content: "What's the weather in San Francisco?"},
      {role: message.role, content: message.content},
      {
        role: "user",
        content: [{
          type: "tool_result",
          tool_use_id: tool_use.id,
          content: [{type: "text", text: "Sunny, 72°F"}]
        }]
      }
    ],
    tools: [GetWeatherInput]
  )

  puts response.content.first.text
end
```

## Tool Use - Auto-Looping Tool Runner

Use the tool runner to automatically execute tools and continue the conversation until completion.

```ruby
client = Anthropic::Client.new

# Define input schema
class CalculatorInput < Anthropic::BaseModel
  required :lhs, Float, doc: "Left operand"
  required :rhs, Float, doc: "Right operand"
  required :operator, Anthropic::InputSchema::EnumOf[:+, :-, :*, :/, :**]
end

# Define tool with implementation
class Calculator < Anthropic::BaseTool
  doc "Performs mathematical calculations"
  input_schema CalculatorInput

  def call(expr)
    case expr.operator
    when :+ then expr.lhs + expr.rhs
    when :- then expr.lhs - expr.rhs
    when :* then expr.lhs * expr.rhs
    when :/ then expr.lhs / expr.rhs
    when :** then expr.lhs ** expr.rhs
    end
  end
end

# Auto-looping tool execution
runner = client.beta.messages.tool_runner(
  model: "claude-sonnet-4-5-20250929",
  max_tokens: 1024,
  messages: [{role: "user", content: "What is (15 * 7) + (12 ** 2)?"}],
  tools: [Calculator.new]
)

# Process each message as it completes
runner.each_message do |message|
  text_blocks = message.content.grep_v(Anthropic::Models::Beta::BetaToolUseBlock)
  puts text_blocks.first&.text unless text_blocks.empty?
end

# Or stream responses in real-time
runner.each_streaming do |stream|
  stream.text.each { |text| print(text) }
end
```

## Structured Outputs

Constrain Claude's responses to a specific JSON schema for reliable data extraction.

```ruby
client = Anthropic::Client.new

# Define output schema
class FamousNumber < Anthropic::BaseModel
  required :value, Float
  optional :reason, String, doc: "Why this number is significant"
end

class MathOutput < Anthropic::BaseModel
  doc "A collection of famous mathematical numbers"
  required :numbers, Anthropic::ArrayOf[FamousNumber], min_length: 3, max_length: 5
end

# Request structured output
message = client.messages.create(
  model: "claude-sonnet-4-5-20250929",
  max_tokens: 1024,
  messages: [{role: "user", content: "Give me famous mathematical numbers"}],
  output_config: {format: MathOutput}
)

# Access parsed output directly
result = message.parsed_output
result.numbers.each do |num|
  puts "#{num.value}: #{num.reason}"
end

# Streaming with structured output
stream = client.messages.stream(
  model: "claude-sonnet-4-5-20250929",
  max_tokens: 1024,
  messages: [{role: "user", content: "List prime numbers under 20"}],
  output_config: {format: MathOutput}
)

stream.text.each { |text| print(text) }
puts stream.accumulated_message.parsed_output
```

## Extended Thinking

Enable Claude's extended thinking capability for complex reasoning tasks.

```ruby
client = Anthropic::Client.new

message = client.messages.create(
  model: "claude-sonnet-4-5-20250929",
  max_tokens: 3200,
  thinking: {type: :enabled, budget_tokens: 1600},
  messages: [{role: :user, content: "Solve: If 3x + 7 = 22, what is x?"}]
)

message.content.each do |block|
  case block
  when Anthropic::ThinkingBlock
    puts "Thinking:"
    puts block.thinking
    puts "---"
  when Anthropic::TextBlock
    puts "Response:"
    puts block.text
  end
end
```

## Image Analysis

Send images to Claude for analysis using base64 encoding or URLs.

```ruby
require "base64"

client = Anthropic::Client.new

# From file
image_data = File.read("image.png")
encoded = Base64.strict_encode64(image_data)

response = client.messages.create(
  model: "claude-sonnet-4-5-20250929",
  max_tokens: 1024,
  messages: [{
    role: :user,
    content: [
      {type: :text, text: "What's in this image?"},
      {
        type: :image,
        source: {
          type: "base64",
          media_type: "image/png",
          data: encoded
        }
      }
    ]
  }]
)

puts response.content.first.text

# From URL
response = client.messages.create(
  model: "claude-sonnet-4-5-20250929",
  max_tokens: 1024,
  messages: [{
    role: :user,
    content: [
      {type: :text, text: "Describe this image"},
      {
        type: :image,
        source: {
          type: "url",
          url: "https://example.com/image.jpg"
        }
      }
    ]
  }]
)
```

## Web Search Tool

Enable Claude to search the web for current information.

```ruby
client = Anthropic::Client.new

message = client.messages.create(
  model: "claude-sonnet-4-5-20250929",
  max_tokens: 1024,
  messages: [{role: :user, content: "What's the current weather in Tokyo?"}],
  tools: [{
    name: "web_search",
    type: "web_search_20250305"
  }]
)

message.content.each do |block|
  case block
  when Anthropic::ServerToolUseBlock
    puts "Searching: #{block.input}"
  when Anthropic::WebSearchToolResultBlock
    puts "Search results:"
    block.content.each { |result| puts "- #{result}" }
  when Anthropic::TextBlock
    puts "Response: #{block.text}"
  end
end

puts "Tokens used: #{message.usage.input_tokens + message.usage.output_tokens}"
```

## Message Batches API

Process multiple requests in batch for high-throughput workloads. Batches can take up to 24 hours to complete.

```ruby
client = Anthropic::Client.new

# Create a batch
batch = client.messages.batches.create(
  requests: [
    {
      custom_id: "request-1",
      params: {
        model: "claude-sonnet-4-5-20250929",
        max_tokens: 1024,
        messages: [{role: "user", content: "Hello!"}]
      }
    },
    {
      custom_id: "request-2",
      params: {
        model: "claude-sonnet-4-5-20250929",
        max_tokens: 1024,
        messages: [{role: "user", content: "Goodbye!"}]
      }
    }
  ]
)

puts "Batch ID: #{batch.id}"
puts "Status: #{batch.processing_status}"

# Check batch status
batch = client.messages.batches.retrieve(batch.id)
puts "Status: #{batch.processing_status}"

# List all batches
client.messages.batches.list.each do |b|
  puts "#{b.id}: #{b.processing_status}"
end

# Stream results when complete
results = client.messages.batches.results_streaming(batch.id)
results.each do |result|
  puts "#{result.custom_id}: #{result.result}"
end

# Cancel a batch
client.messages.batches.cancel(batch.id)

# Delete a completed batch
client.messages.batches.delete(batch.id)
```

## Token Counting

Count tokens before sending a request to estimate costs and manage context windows.

```ruby
client = Anthropic::Client.new

result = client.messages.count_tokens(
  model: "claude-sonnet-4-5-20250929",
  messages: [
    {role: "user", content: "What is the meaning of life?"}
  ],
  system_: "You are a philosopher."
)

puts "Input tokens: #{result.input_tokens}"

# With tools
class SearchInput < Anthropic::BaseModel
  required :query, String
end

result = client.messages.count_tokens(
  model: "claude-sonnet-4-5-20250929",
  messages: [{role: "user", content: "Search for Ruby tutorials"}],
  tools: [SearchInput]
)

puts "Input tokens with tools: #{result.input_tokens}"
```

## Models API

List and retrieve information about available Claude models.

```ruby
client = Anthropic::Client.new

# List all models
client.models.list.each do |model|
  puts "#{model.id}: #{model.display_name}"
end

# Get specific model info
model = client.models.retrieve("claude-sonnet-4-5-20250929")
puts "Model: #{model.display_name}"
puts "Created: #{model.created_at}"
```

## AWS Bedrock Client

Use Claude models through AWS Bedrock. Requires the `aws-sdk-bedrockruntime` gem.

```ruby
require "anthropic"

# AWS credentials resolved from environment or AWS SDK chain
client = Anthropic::Helpers::Bedrock::Client.new(
  aws_region: "us-east-1"  # Optional, uses AWS_REGION if not specified
)

# Or with explicit credentials
client = Anthropic::Helpers::Bedrock::Client.new(
  aws_region: "us-west-2",
  aws_access_key: "AKIA...",
  aws_secret_key: "secret",
  aws_session_token: "token",  # Optional for temporary credentials
  aws_profile: "my-profile"    # Optional AWS profile
)

# Use the same Messages API
response = client.messages.create(
  model: "anthropic.claude-3-5-sonnet-20241022-v2:0",
  max_tokens: 1024,
  messages: [{role: :user, content: "Hello from Bedrock!"}]
)

puts response.content.first.text
```

## Google Cloud Vertex AI Client

Use Claude models through Google Cloud Vertex AI. Requires the `googleauth` gem.

```ruby
require "anthropic"

# GCP credentials resolved via Application Default Credentials
client = Anthropic::Helpers::Vertex::Client.new(
  region: "us-central1",                    # Required, or set CLOUD_ML_REGION
  project_id: "my-gcp-project"              # Required, or set ANTHROPIC_VERTEX_PROJECT_ID
)

# Use the same Messages API
response = client.messages.create(
  model: "claude-sonnet-4-5-20250929",
  max_tokens: 1024,
  messages: [{role: :user, content: "Hello from Vertex AI!"}]
)

puts response.content.first.text

# Streaming works the same way
stream = client.messages.stream(
  model: "claude-sonnet-4-5-20250929",
  max_tokens: 1024,
  messages: [{role: :user, content: "Tell me about GCP"}]
)

stream.text.each { |text| print(text) }
```

## Input Schema Helpers

Define structured input schemas for tools and structured outputs using `Anthropic::BaseModel`.

```ruby
# Basic types
class UserInput < Anthropic::BaseModel
  required :name, String, doc: "User's full name"
  required :age, Integer, doc: "Age in years"
  optional :email, String, doc: "Email address"
  required :active, Anthropic::Boolean
  required :score, Float
end

# Enums
class OrderInput < Anthropic::BaseModel
  required :status, Anthropic::EnumOf[:pending, :shipped, :delivered]
  required :priority, Anthropic::EnumOf[:low, :medium, :high], nil?: true
end

# Arrays
class TeamInput < Anthropic::BaseModel
  required :members, Anthropic::ArrayOf[String]
  required :scores, Anthropic::ArrayOf[Integer], min_length: 1, max_length: 10
end

# Nested models
class Address < Anthropic::BaseModel
  required :street, String
  required :city, String
  required :country, String
end

class PersonInput < Anthropic::BaseModel
  required :name, String
  required :address, Address
  optional :work_address, Address, nil?: true
end

# Union types
class FlexibleInput < Anthropic::BaseModel
  required :id, Anthropic::UnionOf[String, Integer]
  required :value, Anthropic::UnionOf[Float, String, NilClass]
end
```

## Summary

The Anthropic Ruby SDK provides a comprehensive interface for building AI-powered Ruby applications with Claude. Primary use cases include conversational AI chatbots, content generation systems, document analysis pipelines, code assistance tools, and automated reasoning tasks. The SDK's streaming capabilities make it ideal for real-time applications, while the batch API supports high-throughput processing workloads. Tool use enables Claude to interact with external systems, databases, and APIs.

Integration patterns typically involve initializing a client with API credentials, then calling the Messages API for interactions. For production applications, streaming is recommended for user-facing features to provide responsive feedback, while batching is optimal for background processing tasks. The SDK integrates seamlessly with AWS Bedrock and Google Cloud Vertex AI for enterprise deployments that require cloud provider integration. Error handling is built-in with automatic retries and exponential backoff, making the SDK suitable for reliable production systems.