llm_gateway 0.1.0

checksums.yaml

@@ -0,0 +1,7 @@
+---
+SHA256:
+  metadata.gz: 912c28c117bad46986f56b7571497ff276cb3ac5b9e39b442c7b29aeb50692e6
+  data.tar.gz: 2a087a8344992f50e0a4258d6aecbbeb7b5670cba62cfefed7c8fe054ed98c40
+SHA512:
+  metadata.gz: 541fb18a405c58505913a7b5dc57fe109dacec72644224cdfb9746e5da290ca93cab2670731ae41f865cc79a20f28d6ee5dab3ea6690ab2161f82fe82b3c81f2
+  data.tar.gz: 2d6b3294fb5140e7e8b3754ae4c16bd983a133db5a3dcf61fd1adc0f07eed18003f4be724e2b90d98ca98f0aa78678d87648d434893836b52934e7da4fb7a979

data/.rubocop.yml

@@ -0,0 +1,5 @@
+inherit_gem:
+  rubocop-rails-omakase: rubocop.yml
+
+AllCops:
+  TargetRubyVersion: 3.1

data/CHANGELOG.md

@@ -0,0 +1,5 @@
+## [Unreleased]
+
+## [0.1.0] - 2025-08-04
+
+- Initial release

data/CODE_OF_CONDUCT.md

@@ -0,0 +1,132 @@
+# Contributor Covenant Code of Conduct
+
+## Our Pledge
+
+We as members, contributors, and leaders pledge to make participation in our
+community a harassment-free experience for everyone, regardless of age, body
+size, visible or invisible disability, ethnicity, sex characteristics, gender
+identity and expression, level of experience, education, socio-economic status,
+nationality, personal appearance, race, caste, color, religion, or sexual
+identity and orientation.
+
+We pledge to act and interact in ways that contribute to an open, welcoming,
+diverse, inclusive, and healthy community.
+
+## Our Standards
+
+Examples of behavior that contributes to a positive environment for our
+community include:
+
+* Demonstrating empathy and kindness toward other people
+* Being respectful of differing opinions, viewpoints, and experiences
+* Giving and gracefully accepting constructive feedback
+* Accepting responsibility and apologizing to those affected by our mistakes,
+  and learning from the experience
+* Focusing on what is best not just for us as individuals, but for the overall
+  community
+
+Examples of unacceptable behavior include:
+
+* The use of sexualized language or imagery, and sexual attention or advances of
+  any kind
+* Trolling, insulting or derogatory comments, and personal or political attacks
+* Public or private harassment
+* Publishing others' private information, such as a physical or email address,
+  without their explicit permission
+* Other conduct which could reasonably be considered inappropriate in a
+  professional setting
+
+## Enforcement Responsibilities
+
+Community leaders are responsible for clarifying and enforcing our standards of
+acceptable behavior and will take appropriate and fair corrective action in
+response to any behavior that they deem inappropriate, threatening, offensive,
+or harmful.
+
+Community leaders have the right and responsibility to remove, edit, or reject
+comments, commits, code, wiki edits, issues, and other contributions that are
+not aligned to this Code of Conduct, and will communicate reasons for moderation
+decisions when appropriate.
+
+## Scope
+
+This Code of Conduct applies within all community spaces, and also applies when
+an individual is officially representing the community in public spaces.
+Examples of representing our community include using an official email address,
+posting via an official social media account, or acting as an appointed
+representative at an online or offline event.
+
+## Enforcement
+
+Instances of abusive, harassing, or otherwise unacceptable behavior may be
+reported to the community leaders responsible for enforcement at
+[INSERT CONTACT METHOD].
+All complaints will be reviewed and investigated promptly and fairly.
+
+All community leaders are obligated to respect the privacy and security of the
+reporter of any incident.
+
+## Enforcement Guidelines
+
+Community leaders will follow these Community Impact Guidelines in determining
+the consequences for any action they deem in violation of this Code of Conduct:
+
+### 1. Correction
+
+**Community Impact**: Use of inappropriate language or other behavior deemed
+unprofessional or unwelcome in the community.
+
+**Consequence**: A private, written warning from community leaders, providing
+clarity around the nature of the violation and an explanation of why the
+behavior was inappropriate. A public apology may be requested.
+
+### 2. Warning
+
+**Community Impact**: A violation through a single incident or series of
+actions.
+
+**Consequence**: A warning with consequences for continued behavior. No
+interaction with the people involved, including unsolicited interaction with
+those enforcing the Code of Conduct, for a specified period of time. This
+includes avoiding interactions in community spaces as well as external channels
+like social media. Violating these terms may lead to a temporary or permanent
+ban.
+
+### 3. Temporary Ban
+
+**Community Impact**: A serious violation of community standards, including
+sustained inappropriate behavior.
+
+**Consequence**: A temporary ban from any sort of interaction or public
+communication with the community for a specified period of time. No public or
+private interaction with the people involved, including unsolicited interaction
+with those enforcing the Code of Conduct, is allowed during this period.
+Violating these terms may lead to a permanent ban.
+
+### 4. Permanent Ban
+
+**Community Impact**: Demonstrating a pattern of violation of community
+standards, including sustained inappropriate behavior, harassment of an
+individual, or aggression toward or disparagement of classes of individuals.
+
+**Consequence**: A permanent ban from any sort of public interaction within the
+community.
+
+## Attribution
+
+This Code of Conduct is adapted from the [Contributor Covenant][homepage],
+version 2.1, available at
+[https://www.contributor-covenant.org/version/2/1/code_of_conduct.html][v2.1].
+
+Community Impact Guidelines were inspired by
+[Mozilla's code of conduct enforcement ladder][Mozilla CoC].
+
+For answers to common questions about this code of conduct, see the FAQ at
+[https://www.contributor-covenant.org/faq][FAQ]. Translations are available at
+[https://www.contributor-covenant.org/translations][translations].
+
+[homepage]: https://www.contributor-covenant.org
+[v2.1]: https://www.contributor-covenant.org/version/2/1/code_of_conduct.html
+[Mozilla CoC]: https://github.com/mozilla/diversity
+[FAQ]: https://www.contributor-covenant.org/faq
+[translations]: https://www.contributor-covenant.org/translations

data/LICENSE.txt

@@ -0,0 +1,21 @@
+The MIT License (MIT)
+
+Copyright (c) 2025 billybonks
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in
+all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+THE SOFTWARE.

data/README.md

@@ -0,0 +1,128 @@
+# 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:
+
+```bash
+bundle add llm_gateway
+```
+
+Or install it yourself:
+
+```bash
+gem install llm_gateway
+```
+
+## Usage
+
+### Basic Chat
+
+```ruby
+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.'
+)
+```
+
+### Tool Usage (Function Calling)
+
+```ruby
+# 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.'
+)
+```
+
+### Response Format
+
+All providers return responses in a consistent format:
+
+```ruby
+{
+  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:
+
+```ruby
+begin
+  result = LlmGateway::Client.chat('invalid-model', 'Hello')
+rescue LlmGateway::Errors::UnsupportedModel => e
+  puts "Unsupported model: #{e.message}"
+rescue LlmGateway::Errors::AuthenticationError => e
+  puts "Authentication failed: #{e.message}"
+rescue LlmGateway::Errors::RateLimitError => e
+  puts "Rate limit exceeded: #{e.message}"
+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](https://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](https://github.com/Hyper-Unearthing/llm_gateway/blob/master/CODE_OF_CONDUCT.md).
+
+## License
+
+The gem is available as open source under the terms of the [MIT License](https://opensource.org/licenses/MIT).
+
+## 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](https://github.com/Hyper-Unearthing/llm_gateway/blob/master/CODE_OF_CONDUCT.md).

data/Rakefile

@@ -0,0 +1,29 @@
+# frozen_string_literal: true
+
+require "bundler/gem_tasks"
+require "minitest/test_task"
+
+Minitest::TestTask.create
+
+require "rubocop/rake_task"
+
+RuboCop::RakeTask.new
+
+begin
+  require "gem-release"
+
+  desc "Release with changelog"
+  task :release do
+    # Generate changelog first
+    sh "github_changelog_generator" # or your preferred changelog tool
+    sh "git add CHANGELOG.md"
+    sh "git commit -m 'Update changelog' || echo 'No changelog changes'"
+
+    # Release
+    sh "gem bump --version patch --tag --push --release"
+  end
+rescue LoadError
+  # gem-release not available in this environment
+end
+
+task default: %i[test rubocop]

data/lib/llm_gateway/adapters/claude/client.rb

@@ -0,0 +1,56 @@
+# frozen_string_literal: true
+
+require_relative "../../base_client"
+
+module LlmGateway
+  module Adapters
+    module Claude
+      class Client < BaseClient
+        def initialize(model_key: "claude-3-7-sonnet-20250219", api_key: ENV["ANTHROPIC_API_KEY"])
+          @base_endpoint = "https://api.anthropic.com/v1"
+          super(model_key: model_key, api_key: api_key)
+        end
+
+        def chat(messages, response_format: { type: "text" }, tools: nil, system: [], max_completion_tokens: 4096)
+          body = {
+            model: model_key,
+            max_tokens: max_completion_tokens,
+            messages: messages
+          }
+
+          body.merge!(tools: tools) if LlmGateway::Utils.present?(tools)
+          body.merge!(system: system) if LlmGateway::Utils.present?(system)
+
+          post("messages", body)
+        end
+
+        def download_file(file_id)
+          get("files/#{file_id}/content")
+        end
+
+        private
+
+        def build_headers
+          {
+            "anthropic-version" => "2023-06-01",
+            "content-type" => "application/json",
+            "x-api-key" => api_key,
+            "anthropic-beta" => "code-execution-2025-05-22,files-api-2025-04-14"
+          }
+        end
+
+        def handle_client_specific_errors(response, error)
+          case response.code.to_i
+          when 400
+            if error["message"]&.start_with?("prompt is too long")
+              raise Errors::PromptTooLong.new(error["message"], error["type"])
+            end
+          end
+
+          # If we get here, we didn't handle it specifically
+          raise Errors::APIStatusError.new(error["message"], error["type"])
+        end
+      end
+    end
+  end
+end

data/lib/llm_gateway/adapters/claude/input_mapper.rb

@@ -0,0 +1,34 @@
+# frozen_string_literal: true
+
+module LlmGateway
+  module Adapters
+    module Claude
+      class InputMapper
+        extend LlmGateway::FluentMapper
+
+        map :messages do |_, value|
+          value.map do |msg|
+            msg = msg.merge(role: "user") if msg[:role] == "developer"
+            msg.slice(:role, :content)
+          end
+        end
+
+        map :system do |_, value|
+          if value.empty?
+            nil
+          elsif value.length == 1 && value.first[:role] == "system"
+            # If we have a single system message, convert to Claude format
+            [ { type: "text", text: value.first[:content] } ]
+          else
+            # For multiple messages or non-standard format, pass through
+            value
+          end
+        end
+
+        map :tools do |_, value|
+          value
+        end
+      end
+    end
+  end
+end

data/lib/llm_gateway/adapters/claude/output_mapper.rb

@@ -0,0 +1,24 @@
+# frozen_string_literal: true
+
+module LlmGateway
+  module Adapters
+    module Claude
+      class OutputMapper
+        extend LlmGateway::FluentMapper
+
+        map :id
+        map :model
+        map :usage
+        map :choices do |_, _|
+          # Claude returns content directly at root level, not in a choices array
+          # We need to construct the choices array from the full response data
+          [ {
+            content: @data[:content] || [], # Use content directly from Claude response
+            finish_reason: @data[:stop_reason],
+            role: "assistant"
+          } ]
+        end
+      end
+    end
+  end
+end

data/lib/llm_gateway/adapters/groq/client.rb

@@ -0,0 +1,58 @@
+# frozen_string_literal: true
+
+require_relative "../../base_client"
+
+module LlmGateway
+  module Adapters
+    module Groq
+      class Client < BaseClient
+        def initialize(model_key: "gemma2-9b-it", api_key: ENV["GROQ_API_KEY"])
+          @base_endpoint = "https://api.groq.com/openai/v1"
+          super(model_key: model_key, api_key: api_key)
+        end
+
+        def chat(messages, response_format: { type: "text" }, tools: nil, system: [], max_completion_tokens: 4096)
+          body = {
+            model: model_key,
+            messages: system + messages,
+            temperature: 0,
+            max_completion_tokens: max_completion_tokens,
+            response_format: response_format,
+            tools: tools
+          }
+
+          post("chat/completions", body)
+        end
+
+        private
+
+        def build_headers
+          {
+            "content-type" => "application/json",
+            "Authorization" => "Bearer #{api_key}"
+          }
+        end
+
+        def handle_client_specific_errors(response, error)
+          # Groq likely uses 'code' like OpenAI since it's OpenAI-compatible
+          error_code = error["code"]
+
+          case response.code.to_i
+
+          when 413
+            if error["message"]&.start_with?("Request too large")
+              raise Errors::PromptTooLong.new(error["message"], error["type"])
+            end
+          when 429
+            raise Errors::RateLimitError.new(error["type"], error_code) if error_code == "rate_limit_exceeded"
+
+            raise Errors::OverloadError.new(error["message"], error_code)
+          end
+
+          # If we get here, we didn't handle it specifically
+          raise Errors::APIStatusError.new(error["message"], error_code)
+        end
+      end
+    end
+  end
+end

data/lib/llm_gateway/adapters/groq/input_mapper.rb

@@ -0,0 +1,78 @@
+# frozen_string_literal: true
+
+module LlmGateway
+  module Adapters
+    module Groq
+      class InputMapper
+        extend LlmGateway::FluentMapper
+
+        map :system
+        map :response_format
+
+        mapper :tool_usage do
+          map :role, default: "assistant"
+          map :content do
+            nil
+          end
+          map :tool_calls, from: :content do |_, value|
+            value.map do |content|
+              {
+                'id': content[:id],
+                'type': "function",
+                'function': {
+                  'name': content[:name],
+                  'arguments': content[:input].to_json
+                }
+              }
+            end
+          end
+        end
+
+        mapper :tool_result_message do
+          map :role, default: "tool"
+          map :tool_call_id, from: "tool_use_id"
+          map :content
+        end
+
+        map :messages do |_, value|
+          value.map do |msg|
+            if msg[:role] == "user"
+              msg
+            elsif msg[:content].is_a?(Array)
+              results = []
+              # Handle tool_use messages
+              tool_uses = msg[:content].select { |c| c[:type] == "tool_use" }
+              results << map_single(msg, with: :tool_usage) if tool_uses.any?
+              # Handle tool_result messages
+              tool_results = msg[:content].select { |c| c[:type] == "tool_result" }
+              tool_results.each do |content|
+                results << map_single(content, with: :tool_result_message)
+              end
+
+              results
+            else
+              msg
+            end
+          end.flatten
+        end
+
+        map :tools do |_, value|
+          if value
+            value.map do |tool|
+              {
+                type: "function",
+                function: {
+                  name: tool[:name],
+                  description: tool[:description],
+                  parameters: tool[:input_schema]
+                }
+              }
+            end
+          else
+            value
+          end
+        end
+      end
+    end
+  end
+end

data/lib/llm_gateway/adapters/groq/output_mapper.rb

@@ -0,0 +1,46 @@
+# frozen_string_literal: true
+
+module LlmGateway
+  module Adapters
+    module Groq
+      class OutputMapper
+        extend LlmGateway::FluentMapper
+
+        mapper :tool_call do
+          map :id
+          map :type do
+            "tool_use" # Always return 'tool_use' regardless of input
+          end
+          map :name, from: "function.name"
+          map :input, from: "function.arguments" do |_, value|
+            parsed = value.is_a?(String) ? JSON.parse(value) : value
+            parsed
+          end
+        end
+
+        mapper :content_item do
+          map :text, from: "content"
+          map :type, default: "text"
+        end
+
+        map :id
+        map :model
+        map :usage
+        map :choices, from: "choices" do |_, value|
+          value.map do |choice|
+            message = choice[:message] || {}
+            content_item = map_single(message, with: :content_item, default: {})
+            tool_calls = map_collection(message[:tool_calls], with: :tool_call, default: [])
+
+            # Only include content_item if it has actual text content
+            content_array = []
+            content_array << content_item if LlmGateway::Utils.present?(content_item["text"])
+            content_array += tool_calls
+
+            { content: content_array }
+          end
+        end
+      end
+    end
+  end
+end

data/lib/llm_gateway/adapters/open_ai/client.rb

@@ -0,0 +1,59 @@
+# frozen_string_literal: true
+
+require_relative "../../base_client"
+
+module LlmGateway
+  module Adapters
+    module OpenAi
+      class Client < BaseClient
+        def initialize(model_key: "gpt-4o", api_key: ENV["OPENAI_API_KEY"])
+          @base_endpoint = "https://api.openai.com/v1"
+          super(model_key: model_key, api_key: api_key)
+        end
+
+        def chat(messages, response_format: { type: "text" }, tools: nil, system: [], max_completion_tokens: 4096)
+          body = {
+            model: model_key,
+            messages: system + messages,
+            max_completion_tokens: max_completion_tokens
+          }
+          body[:tools] = tools if tools
+
+          post("chat/completions", body)
+        end
+
+        def generate_embeddings(input)
+          body = {
+            input:,
+            model: model_key
+          }
+          post("embeddings", body)
+        end
+
+        private
+
+        def build_headers
+          {
+            "content-type" => "application/json",
+            "Authorization" => "Bearer #{api_key}"
+          }
+        end
+
+        def handle_client_specific_errors(response, error)
+          # OpenAI uses 'code' instead of 'type' for error codes
+          error_code = error["code"]
+
+          case response.code.to_i
+          when 429
+            raise Errors::RateLimitError.new(error["message"], error_code)
+          when 503
+            raise Errors::OverloadError.new(error["message"], error_code)
+          end
+
+          # If we get here, we didn't handle it specifically
+          raise Errors::APIStatusError.new(error["message"], error_code)
+        end
+      end
+    end
+  end
+end

data/lib/llm_gateway/adapters/open_ai/input_mapper.rb

@@ -0,0 +1,19 @@
+# frozen_string_literal: true
+
+module LlmGateway
+  module Adapters
+    module OpenAi
+      class InputMapper < LlmGateway::Adapters::Groq::InputMapper
+        map :system do |_, value|
+          if value.empty?
+            []
+          else
+            value.map do |msg|
+              msg[:role] == "system" ? msg.merge(role: "developer") : msg
+            end
+          end
+        end
+      end
+    end
+  end
+end

data/lib/llm_gateway/adapters/open_ai/output_mapper.rb

@@ -0,0 +1,10 @@
+# frozen_string_literal: true
+
+module LlmGateway
+  module Adapters
+    module OpenAi
+      class OutputMapper < LlmGateway::Adapters::Groq::OutputMapper
+      end
+    end
+  end
+end

data/lib/llm_gateway/base_client.rb

@@ -0,0 +1,104 @@
+# frozen_string_literal: true
+
+require "net/http"
+require "json"
+
+module LlmGateway
+  class BaseClient
+    attr_accessor
+    attr_reader :api_key, :model_key, :base_endpoint
+
+    def initialize(model_key:, api_key:)
+      @model_key = model_key
+      @api_key = api_key
+    end
+
+    def get(url_part, extra_headers = {})
+      endpoint = "#{base_endpoint}/#{url_part.sub(%r{^/}, "")}"
+      response = make_request(endpoint, Net::HTTP::Get, nil, extra_headers)
+      process_response(response)
+    end
+
+    def post(url_part, body = nil, extra_headers = {})
+      endpoint = "#{base_endpoint}/#{url_part.sub(%r{^/}, "")}"
+      response = make_request(endpoint, Net::HTTP::Post, body, extra_headers)
+      process_response(response)
+    end
+
+    protected
+
+    def make_request(endpoint, method, params = nil, extra_headers = {})
+      uri = URI(endpoint)
+      http = Net::HTTP.new(uri.host, uri.port)
+      http.use_ssl = true
+      http.read_timeout = 480
+      http.open_timeout = 10
+
+      request = method.new(uri)
+      headers = build_headers.merge(extra_headers)
+      headers.each { |key, value| request[key] = value }
+      request.body = params.to_json if params
+
+      http.request(request)
+    end
+
+    def process_response(response)
+      case response.code.to_i
+      when 200
+        content_type = response["content-type"]
+        if content_type&.include?("application/json")
+          LlmGateway::Utils.deep_symbolize_keys(JSON.parse(response.body))
+        else
+          response.body
+        end
+      else
+        handle_error(response)
+      end
+    end
+
+    def handle_error(response)
+      error_body = begin
+        JSON.parse(response.body)
+      rescue StandardError
+        {}
+      end
+      error = error_body["error"] || {}
+
+      # Try client-specific error handling first
+      begin
+        handle_client_specific_errors(response, error)
+      rescue Errors::APIStatusError => e
+        # If client doesn't handle it, use standard HTTP status codes
+        # Use the message and code that were already passed to APIStatusError
+        case response.code.to_i
+        when 400
+          raise Errors::BadRequestError.new(e.message, e.code)
+        when 401
+          raise Errors::AuthenticationError.new(e.message, e.code)
+        when 403
+          raise Errors::PermissionDeniedError.new(e.message, e.code)
+        when 404
+          raise Errors::NotFoundError.new(e.message, e.code)
+        when 409
+          raise Errors::ConflictError.new(e.message, e.code)
+        when 422
+          raise Errors::UnprocessableEntityError.new(e.message, e.code)
+        when 429
+          raise Errors::RateLimitError.new(e.message, e.code)
+        when 503
+          raise Errors::OverloadError.new(e.message, e.code)
+        when 500..599
+          raise Errors::InternalServerError.new(e.message, e.code)
+        else
+          raise e # Re-raise the original APIStatusError
+        end
+      end
+    end
+
+    def build_headers
+      raise NotImplementedError, "Subclasses must implement build_headers"
+    end
+
+    def handle_client_specific_errors(response, error); end
+  end
+end

data/lib/llm_gateway/client.rb

@@ -0,0 +1,77 @@
+# frozen_string_literal: true
+
+module LlmGateway
+  class Client
+    def self.chat(model, message, response_format: "text", tools: nil, system: nil)
+      client_klass = client_class(model)
+      client = client_klass.new(model_key: model)
+
+      input_mapper = input_mapper_for_client(client)
+      normalized_input = input_mapper.map({
+                                            messages: normalize_messages(message),
+                                            response_format: normalize_response_format(response_format),
+                                            tools: tools,
+                                            system: normalize_system(system)
+                                          })
+      result = client.chat(
+        normalized_input[:messages],
+        response_format: normalized_input[:response_format],
+        tools: normalized_input[:tools],
+        system: normalized_input[:system]
+      )
+      result_mapper(client).map(result)
+    end
+
+    def self.client_class(model)
+      return LlmGateway::Adapters::Claude::Client if model.start_with?("claude")
+      return LlmGateway::Adapters::Groq::Client if model.start_with?("llama")
+      return LlmGateway::Adapters::OpenAi::Client if model.start_with?("gpt") ||
+                                                     model.start_with?("o4-") ||
+                                                     model.start_with?("openai")
+
+      raise LlmGateway::Errors::UnsupportedModel, model
+    end
+
+    def self.input_mapper_for_client(client)
+      return LlmGateway::Adapters::Claude::InputMapper if client.is_a?(LlmGateway::Adapters::Claude::Client)
+      return LlmGateway::Adapters::OpenAi::InputMapper if client.is_a?(LlmGateway::Adapters::OpenAi::Client)
+
+      LlmGateway::Adapters::Groq::InputMapper if client.is_a?(LlmGateway::Adapters::Groq::Client)
+    end
+
+    def self.result_mapper(client)
+      return LlmGateway::Adapters::Claude::OutputMapper if client.is_a?(LlmGateway::Adapters::Claude::Client)
+      return LlmGateway::Adapters::OpenAi::OutputMapper if client.is_a?(LlmGateway::Adapters::OpenAi::Client)
+
+      LlmGateway::Adapters::Groq::OutputMapper if client.is_a?(LlmGateway::Adapters::Groq::Client)
+    end
+
+    def self.normalize_system(system)
+      if system.nil?
+        []
+      elsif system.is_a?(String)
+        [ { role: "system", content: system } ]
+      elsif system.is_a?(Array)
+        system
+      else
+        raise ArgumentError, "System parameter must be a string or array, got #{system.class}"
+      end
+    end
+
+    def self.normalize_messages(message)
+      if message.is_a?(String)
+        [ { 'role': "user", 'content': message } ]
+      else
+        message
+      end
+    end
+
+    def self.normalize_response_format(response_format)
+      if response_format.is_a?(String)
+        { type: response_format }
+      else
+        response_format
+      end
+    end
+  end
+end

data/lib/llm_gateway/errors.rb

@@ -0,0 +1,30 @@
+# frozen_string_literal: true
+
+module LlmGateway
+  module Errors
+    class BaseError < StandardError
+      attr_reader :code
+
+      def initialize(message = nil, code = nil)
+        @code = code
+        super(message)
+      end
+    end
+
+    class BadRequestError < BaseError; end
+    class AuthenticationError < BaseError; end
+    class PermissionDeniedError < BaseError; end
+    class NotFoundError < BaseError; end
+    class ConflictError < BaseError; end
+    class UnprocessableEntityError < BaseError; end
+    class RateLimitError < BaseError; end
+    class InternalServerError < BaseError; end
+    class APIStatusError < BaseError; end
+    class APITimeoutError < BaseError; end
+    class APIConnectionError < BaseError; end
+    class OverloadError < BaseError; end
+    class UnknownError < BaseError; end
+    class PromptTooLong < BadRequestError; end
+    class UnsupportedModel < BaseError; end
+  end
+end

data/lib/llm_gateway/fluent_mapper.rb

@@ -0,0 +1,144 @@
+# frozen_string_literal: true
+
+module LlmGateway
+  module FluentMapper
+    def self.extended(base)
+      base.instance_variable_set(:@mappers, {})
+      base.instance_variable_set(:@mappings, [])
+    end
+
+    def inherited(subclass)
+      super
+      # Copy parent's mappers and mappings to the subclass
+      subclass.instance_variable_set(:@mappers, @mappers.dup)
+      subclass.instance_variable_set(:@mappings, @mappings.dup)
+    end
+
+    def mapper(name, &block)
+      @mappers[name] = block
+    end
+
+    def map(field_or_data, options = {}, &block)
+      # If called with a single argument and no block, it's the class method usage
+      if block.nil? && options.empty? && !field_or_data.is_a?(Symbol) && !field_or_data.is_a?(String)
+        return new(field_or_data).call
+      end
+
+      # Otherwise it's the field mapping usage
+      @mappings << { field: field_or_data, options: options, block: block }
+    end
+
+    def new(data)
+      MapperInstance.new(data, @mappers, @mappings)
+    end
+
+    class MapperInstance
+      def initialize(data, mappers, mappings)
+        @data = data.respond_to?(:with_indifferent_access) ? data.with_indifferent_access : data
+        @mappers = mappers
+        @mappings = mappings
+        @mapper_definitions = {}
+
+        # Execute mapper definitions
+        mappers.each do |name, block|
+          @mapper_definitions[name] = MapperDefinition.new
+          @mapper_definitions[name].instance_eval(&block)
+        end
+      end
+
+      def call
+        result = {}
+
+        @mappings.each do |mapping|
+          field = mapping[:field]
+          options = mapping[:options]
+          block = mapping[:block]
+
+          from_path = options[:from] || field.to_s
+          default_value = options[:default]
+
+          value = get_nested_value(@data, from_path)
+          value = default_value if value.nil? && !default_value.nil?
+
+          value = instance_exec(field, value, &block) if block
+
+          result[field] = value
+        end
+
+        LlmGateway::Utils.deep_symbolize_keys(result)
+      end
+
+      def map_single(data, options = {})
+        mapper_name = options[:with]
+        default_value = options[:default]
+        return default_value if data.nil? && !default_value.nil?
+        return data unless mapper_name && @mapper_definitions[mapper_name]
+
+        # Apply with_indifferent_access to data
+        data = data.respond_to?(:with_indifferent_access) ? data.with_indifferent_access : data
+
+        mapper_def = @mapper_definitions[mapper_name]
+        result = {}
+
+        mapper_def.mappings.each do |mapping|
+          field = mapping[:field]
+          map_options = mapping[:options]
+          block = mapping[:block]
+
+          from_path = map_options[:from] || field.to_s
+          field_default_value = map_options[:default]
+
+          value = get_nested_value(data, from_path)
+          value = field_default_value if value.nil? && !field_default_value.nil?
+
+          value = instance_exec(field, value, &block) if block
+
+          result[field.to_s] = value
+        end
+
+        result
+      end
+
+      def map_collection(collection, options = {})
+        default_value = options[:default]
+        return default_value if collection.nil? && !default_value.nil?
+        return [] if collection.nil?
+
+        collection.map { |item| map_single(item, options) }
+      end
+
+      private
+
+      def get_nested_value(data, path)
+        return data[path] if data.respond_to?(:[]) && data.key?(path)
+        return data[path.to_sym] if data.respond_to?(:[]) && data.key?(path.to_sym)
+        return data[path.to_s] if data.respond_to?(:[]) && data.key?(path.to_s)
+
+        keys = path.split(".")
+        current = data
+
+        keys.each do |key|
+          return nil unless current.respond_to?(:[])
+
+          current = current[key] || current[key.to_sym] || current[key.to_s]
+
+          return nil if current.nil?
+        end
+
+        current
+      end
+    end
+
+    class MapperDefinition
+      attr_reader :mappings
+
+      def initialize
+        @mappings = []
+      end
+
+      def map(field, options = {}, &block)
+        @mappings << { field: field, options: options, block: block }
+      end
+    end
+  end
+end

data/lib/llm_gateway/utils.rb

@@ -0,0 +1,39 @@
+# frozen_string_literal: true
+
+module LlmGateway
+  module Utils
+    module_function
+
+    def deep_symbolize_keys(hash)
+      case hash
+      when Hash
+        hash.each_with_object({}) do |(key, value), result|
+          result[key.to_sym] = deep_symbolize_keys(value)
+        end
+      when Array
+        hash.map { |item| deep_symbolize_keys(item) }
+      else
+        hash
+      end
+    end
+
+    def present?(value)
+      !blank?(value)
+    end
+
+    def blank?(value)
+      case value
+      when nil
+        true
+      when String
+        value.strip.empty?
+      when Array, Hash
+        value.empty?
+      when Numeric
+        false
+      else
+        value.respond_to?(:empty?) ? value.empty? : false
+      end
+    end
+  end
+end

data/lib/llm_gateway/version.rb

@@ -0,0 +1,5 @@
+# frozen_string_literal: true
+
+module LlmGateway
+  VERSION = "0.1.0"
+end

data/lib/llm_gateway.rb

@@ -0,0 +1,23 @@
+# frozen_string_literal: true
+
+require_relative "llm_gateway/utils"
+require_relative "llm_gateway/version"
+require_relative "llm_gateway/errors"
+require_relative "llm_gateway/fluent_mapper"
+require_relative "llm_gateway/base_client"
+require_relative "llm_gateway/client"
+
+# Load adapters - order matters for inheritance
+require_relative "llm_gateway/adapters/claude/client"
+require_relative "llm_gateway/adapters/claude/input_mapper"
+require_relative "llm_gateway/adapters/claude/output_mapper"
+require_relative "llm_gateway/adapters/groq/client"
+require_relative "llm_gateway/adapters/groq/input_mapper"
+require_relative "llm_gateway/adapters/groq/output_mapper"
+require_relative "llm_gateway/adapters/open_ai/client"
+require_relative "llm_gateway/adapters/open_ai/input_mapper"
+require_relative "llm_gateway/adapters/open_ai/output_mapper"
+
+module LlmGateway
+  class Error < StandardError; end
+end

data/sig/llm_gateway.rbs

@@ -0,0 +1,4 @@
+module LlmGateway
+  VERSION: String
+  # See the writing guide of rbs: https://github.com/ruby/rbs#guides
+end

metadata

@@ -0,0 +1,72 @@
+--- !ruby/object:Gem::Specification
+name: llm_gateway
+version: !ruby/object:Gem::Version
+  version: 0.1.0
+platform: ruby
+authors:
+- billybonks
+autorequire:
+bindir: exe
+cert_chain: []
+date: 2025-08-04 00:00:00.000000000 Z
+dependencies: []
+description: LlmGateway provides a consistent Ruby interface for multiple LLM providers
+  including Claude, OpenAI, and Groq. Features include unified response formatting,
+  error handling, and fluent data mapping.
+email:
+- sebastienstettler@gmail.com
+executables: []
+extensions: []
+extra_rdoc_files: []
+files:
+- ".rubocop.yml"
+- CHANGELOG.md
+- CODE_OF_CONDUCT.md
+- LICENSE.txt
+- README.md
+- Rakefile
+- lib/llm_gateway.rb
+- lib/llm_gateway/adapters/claude/client.rb
+- lib/llm_gateway/adapters/claude/input_mapper.rb
+- lib/llm_gateway/adapters/claude/output_mapper.rb
+- lib/llm_gateway/adapters/groq/client.rb
+- lib/llm_gateway/adapters/groq/input_mapper.rb
+- lib/llm_gateway/adapters/groq/output_mapper.rb
+- lib/llm_gateway/adapters/open_ai/client.rb
+- lib/llm_gateway/adapters/open_ai/input_mapper.rb
+- lib/llm_gateway/adapters/open_ai/output_mapper.rb
+- lib/llm_gateway/base_client.rb
+- lib/llm_gateway/client.rb
+- lib/llm_gateway/errors.rb
+- lib/llm_gateway/fluent_mapper.rb
+- lib/llm_gateway/utils.rb
+- lib/llm_gateway/version.rb
+- sig/llm_gateway.rbs
+homepage: https://github.com/Hyper-Unearthing/llm_gateway
+licenses:
+- MIT
+metadata:
+  homepage_uri: https://github.com/Hyper-Unearthing/llm_gateway
+  source_code_uri: https://github.com/Hyper-Unearthing/llm_gateway
+  changelog_uri: https://github.com/Hyper-Unearthing/llm_gateway/blob/main/CHANGELOG.md
+post_install_message:
+rdoc_options: []
+require_paths:
+- lib
+required_ruby_version: !ruby/object:Gem::Requirement
+  requirements:
+  - - ">="
+    - !ruby/object:Gem::Version
+      version: 3.1.0
+required_rubygems_version: !ruby/object:Gem::Requirement
+  requirements:
+  - - ">="
+    - !ruby/object:Gem::Version
+      version: '0'
+requirements: []
+rubygems_version: 3.5.22
+signing_key:
+specification_version: 4
+summary: A Ruby gateway for LLM APIs with unified interface for Claude, OpenAI, and
+  Groq
+test_files: []