LlmGateway
Provide nuts and bolts for LLM APIs. The goal is to provide a unified interface for multiple LLM provider API's; And Enable developers to have as much control as they want.
You can use the clients directly, Or you can use the gateway to have interop between clients.
Supported Providers
Anthropic, OpenAi, Groq
Installation
Add the gem to your application's Gemfile:
bundle add llm_gateway
Or install it yourself:
gem install llm_gateway
Usage
Basic Chat
require 'llm_gateway'
# Simple text completion
result = LlmGateway::Client.chat(
'claude-sonnet-4-20250514',
'What is the capital of France?'
)
# With system message
result = LlmGateway::Client.chat(
'gpt-4',
'What is the capital of France?',
system: 'You are a helpful geography teacher.'
)
Prompt Class
You can also create reusable prompt classes by subclassing LlmGateway::Prompt:
# Simple text completion with prompt class
class GeographyQuestionPrompt < LlmGateway::Prompt
def initialize(model, question)
super(model)
@question = question
end
def prompt
@question
end
end
# Usage
geography_prompt = GeographyQuestionPrompt.new('claude-sonnet-4-20250514', 'What is the capital of France?')
result = geography_prompt.run
# With system message
class GeographyTeacherPrompt < LlmGateway::Prompt
def initialize(model, question)
super(model)
@question = question
end
def prompt
@question
end
def system_prompt
'You are a helpful geography teacher.'
end
end
# Usage
teacher_prompt = GeographyTeacherPrompt.new('gpt-4', 'What is the capital of France?')
result = teacher_prompt.run
Using Prompt with Tools
You can combine the Prompt class with tools for more complex interactions:
class WeatherAssistantPrompt < LlmGateway::Prompt
def initialize(model, location)
super(model)
@location = location
end
def prompt
"What's the weather like in #{@location}?"
end
def system_prompt
'You are a helpful weather assistant.'
end
def tools
[{
name: 'get_weather',
description: 'Get current weather for a location',
input_schema: {
type: 'object',
properties: {
location: { type: 'string', description: 'City name' }
},
required: ['location']
}
}]
end
end
# Usage
weather_prompt = WeatherAssistantPrompt.new('claude-sonnet-4-20250514', 'Singapore')
result = weather_prompt.run
Tool Usage (Function Calling)
# Define a tool
weather_tool = {
name: 'get_weather',
description: 'Get current weather for a location',
input_schema: {
type: 'object',
properties: {
location: { type: 'string', description: 'City name' }
},
required: ['location']
}
}
# Use the tool
result = LlmGateway::Client.chat(
'claude-sonnet-4-20250514',
'What\'s the weather in Singapore?',
tools: [weather_tool],
system: 'You are a helpful weather assistant.'
)
# Note: Tools are not automatically executed. The LLM will indicate when a tool should be called,
# but it's up to you to find the appropriate tool and execute it based on the response.
# Example of handling tool execution with conversation transcript:
class WeatherAssistant
def initialize
@transcript = []
end
def (content)
# Add user message to transcript
@transcript << { role: 'user', content: [{ type: 'text', text: content }] }
result = LlmGateway::Client.chat(
'claude-sonnet-4-20250514',
@transcript,
tools: [weather_tool],
system: 'You are a helpful weather assistant.'
)
process_response(result[:choices][0][:content])
end
private
def process_response(response)
# Add assistant response to transcript
@transcript << { role: 'assistant', content: response }
response.each do ||
if [:type] == 'text'
puts [:text]
elsif [:type] == 'tool_use'
result = handle_tool_use()
# Add tool result to transcript
tool_result = {
type: 'tool_result',
tool_use_id: [:id],
content: result
}
@transcript << { role: 'user', content: [tool_result] }
# Continue conversation with full transcript context
follow_up = LlmGateway::Client.chat(
'claude-sonnet-4-20250514',
@transcript,
tools: [weather_tool],
system: 'You are a helpful weather assistant.'
)
process_response(follow_up[:choices][0][:content])
end
end
end
def handle_tool_use()
tool_class = WeatherAssistantPrompt.find_tool([:name])
raise "Unknown tool: #{[:name]}" unless tool_class
# Execute the tool with the provided input
tool_instance = tool_class.new
tool_instance.execute([:input])
rescue StandardError => e
"Error executing tool: #{e.}"
end
end
# Usage
assistant = WeatherAssistant.new
assistant.("What's the weather in Singapore?")
Response Format
All providers return responses in a consistent format:
{
choices: [
{
content: [
{ type: 'text', text: 'The capital of France is Paris.' }
],
finish_reason: 'end_turn',
role: 'assistant'
}
],
usage: {
input_tokens: 15,
output_tokens: 8,
total_tokens: 23
},
model: 'claude-sonnet-4-20250514',
id: 'msg_abc123'
}
Error Handling
LlmGateway provides consistent error handling across all providers:
begin
result = LlmGateway::Client.chat('invalid-model', 'Hello')
rescue LlmGateway::Errors::UnsupportedModel => e
puts "Unsupported model: #{e.}"
rescue LlmGateway::Errors::AuthenticationError => e
puts "Authentication failed: #{e.}"
rescue LlmGateway::Errors::RateLimitError => e
puts "Rate limit exceeded: #{e.}"
end
Development
After checking out the repo, run bin/setup to install dependencies. Then, run rake test to run the tests. You can also run bin/console for an interactive prompt that will allow you to experiment.
To install this gem onto your local machine, run bundle exec rake install. To release a new version, update the version number in version.rb, and then run bundle exec rake release, which will create a git tag for the version, push git commits and the created tag, and push the .gem file to rubygems.org.
Contributing
Bug reports and pull requests are welcome on GitHub at https://github.com/Hyper-Unearthing/llm_gateway. This project is intended to be a safe, welcoming space for collaboration, and contributors are expected to adhere to the code of conduct.
License
The gem is available as open source under the terms of the MIT License.
Code of Conduct
Everyone interacting in the LlmGateway project's codebases, issue trackers, chat rooms and mailing lists is expected to follow the code of conduct.