ruby_llm 0.1.0.pre25 → 0.1.0.pre27

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 9985c28a42c2ba63f82067dd277e87ad85f5c270f0973bfe30f1a6f2f0cbd86f
-  data.tar.gz: d344c6f85776b8a11bbb59d89a29280bc4bcd7bcc30747b95a4f86b5a53c016e
+  metadata.gz: cd52ddd301a631215257c48fa3c1bc460203a228128b7d8c4c2f4837beaf2f63
+  data.tar.gz: a0cfae0a59ff41ffeef2a1c64ff660a61839147d2587769e8e7760555fd7f8a2
 SHA512:
-  metadata.gz: a409b992af493499ab7e55f5126f19e9968f445804cc0f5a8a56a3de9c8d99b0238079bfcab169d9633516dd7096124d0401ae6df31e1c254afc5e12acc82ff0
-  data.tar.gz: 4eb477e4941f5e229acf8870b6623e1942ff30932b226261dee9597e3cb0494ae828d5ce401932f396d8232be456f8c59e8c0ab302f71aba9814c630691f4b62
+  metadata.gz: 854ea7639ed9e01aafe2ff987bebc47023529544525ec86dbfd1e216119ad6affae78dcadbf4d565f061970ae85454f00bcb7a3f42f6eebd876f6fa3287b8616
+  data.tar.gz: 99bf687ffca1d6ba80745a9406bb96f47148a9bb5a0ed499cddda5e10461675ecb5c434e12e5e3a208f4877fca5bd57b59ab15dd8e9f0f1473c66b86c44fdeb9

data/README.md

@@ -1,6 +1,16 @@
 # RubyLLM
 
-A delightful Ruby interface to the latest large language models. Stop wrestling with multiple APIs and inconsistent interfaces. RubyLLM gives you a clean, unified way to work with models from OpenAI, Anthropic, and more.
+A delightful Ruby way to work with AI language models. Provides a unified interface to OpenAI, Anthropic, Google, and DeepSeek models with automatic token counting, proper streaming support, and a focus on developer happiness. No wrapping your head around multiple APIs - just clean Ruby code that works.
+
+<p align="center">
+  <img src="https://upload.wikimedia.org/wikipedia/commons/4/4d/OpenAI_Logo.svg" alt="OpenAI" height="40" width="120">
+  &nbsp;&nbsp;&nbsp;&nbsp;
+  <img src="https://upload.wikimedia.org/wikipedia/commons/7/78/Anthropic_logo.svg" alt="Anthropic" height="40" width="120">
+  &nbsp;&nbsp;&nbsp;&nbsp;
+<img src="https://upload.wikimedia.org/wikipedia/commons/8/8a/Google_Gemini_logo.svg" alt="Google" height="40" width="120">
+  &nbsp;&nbsp;&nbsp;&nbsp;
+  <img src="https://upload.wikimedia.org/wikipedia/commons/e/ec/DeepSeek_logo.svg" alt="DeepSeek" height="40" width="120"]>
+</p>
 
 [![Gem Version](https://badge.fury.io/rb/ruby_llm.svg)](https://badge.fury.io/rb/ruby_llm)
 [![Ruby Style Guide](https://img.shields.io/badge/code_style-standard-brightgreen.svg)](https://github.com/testdouble/standard)
@@ -28,6 +38,8 @@ require 'ruby_llm'
 RubyLLM.configure do |config|
   config.openai_api_key = ENV['OPENAI_API_KEY']
   config.anthropic_api_key = ENV['ANTHROPIC_API_KEY']
+  config.gemini_api_key = ENV['GEMINI_API_KEY']
+  config.deepseek_api_key = ENV['DEEPSEEK_API_KEY']
 end
 ```
 
@@ -61,7 +73,7 @@ image_models = RubyLLM.models.image_models
 Conversations are simple and natural:
 
 ```ruby
-chat = RubyLLM.chat model: 'claude-3-5-sonnet-20241022'
+chat = RubyLLM.chat model: 'gemini-2.0-flash'
 
 # Ask questions
 response = chat.ask "What's your favorite Ruby feature?"
@@ -89,7 +101,7 @@ Need vector embeddings for your text? RubyLLM makes it simple:
 RubyLLM.embed "Hello, world!"
 
 # Use a specific model
-RubyLLM.embed "Ruby is awesome!", model: "text-embedding-3-large"
+RubyLLM.embed "Ruby is awesome!", model: "text-embedding-004"
 
 # Process multiple texts at once
 RubyLLM.embed([
@@ -185,6 +197,10 @@ rescue RubyLLM::UnauthorizedError
   puts "Check your API credentials"
 rescue RubyLLM::BadRequestError => e
   puts "Something went wrong: #{e.message}"
+rescue RubyLLM::PaymentRequiredError
+  puts "Time to top up your API credits"
+rescue RubyLLM::ServiceUnavailableError
+  puts "API service is temporarily down"
 end
 ```
 
@@ -357,7 +373,7 @@ class WeatherTool < RubyLLM::Tool
 end
 
 # Use tools with your persisted chats
-chat = Chat.create! model_id: "gpt-4"
+chat = Chat.create! model_id: "deepseek-reasoner"
 chat.chat.with_tool WeatherTool.new
 
 # Ask about weather - tool usage is automatically saved

data/Rakefile

@@ -1,6 +1,9 @@
 # frozen_string_literal: true
 
+require 'bundler/setup'
 require 'bundler/gem_tasks'
 require 'rake/clean'
 
+Dir.glob('lib/tasks/**/*.rake').each { |r| load r }
+
 task default: %w[build]

data/bin/console

@@ -10,6 +10,8 @@ require 'irb'
 RubyLLM.configure do |config|
   config.openai_api_key = ENV['OPENAI_API_KEY']
   config.anthropic_api_key = ENV['ANTHROPIC_API_KEY']
+  config.gemini_api_key = ENV['GEMINI_API_KEY']
+  config.deepseek_api_key = ENV['DEEPSEEK_API_KEY']
 end
 
 IRB.start(__FILE__)

data/lib/ruby_llm/configuration.rb

@@ -12,6 +12,8 @@ module RubyLLM
   class Configuration
     attr_accessor :openai_api_key,
                   :anthropic_api_key,
+                  :gemini_api_key,
+                  :deepseek_api_key,
                   :default_model,
                   :default_embedding_model,
                   :request_timeout

data/lib/ruby_llm/error.rb

@@ -23,6 +23,8 @@ module RubyLLM
   class InvalidRoleError < StandardError; end
   class UnsupportedFunctionsError < StandardError; end
   class UnauthorizedError < Error; end
+  class PaymentRequiredError < Error; end
+  class ServiceUnavailableError < Error; end
   class BadRequestError < Error; end
   class RateLimitError < Error; end
   class ServerError < Error; end
@@ -42,18 +44,26 @@ module RubyLLM
     end
 
     class << self
-      def parse_error(provider:, response:) # rubocop:disable Metrics/CyclomaticComplexity,Metrics/MethodLength
+      def parse_error(provider:, response:) # rubocop:disable Metrics/CyclomaticComplexity,Metrics/MethodLength,Metrics/AbcSize,Metrics/PerceivedComplexity
         message = provider&.parse_error(response)
 
         case response.status
+        when 200..399
+          message
         when 400
           raise BadRequestError.new(response, message || 'Invalid request - please check your input')
         when 401
           raise UnauthorizedError.new(response, message || 'Invalid API key - check your credentials')
+        when 402
+          raise PaymentRequiredError.new(response, message || 'Payment required - please top up your account')
         when 429
           raise RateLimitError.new(response, message || 'Rate limit exceeded - please wait a moment')
-        when 500..599
+        when 500
           raise ServerError.new(response, message || 'API server error - please try again')
+        when 503
+          raise ServiceUnavailableError.new(response, message || 'API server unavailable - please try again later')
+        else
+          raise Error.new(response, message || 'An unknown error occurred')
         end
       end
     end

data/lib/ruby_llm/model_capabilities/anthropic.rb

@@ -4,7 +4,7 @@ module RubyLLM
   module ModelCapabilities
     # Determines capabilities and pricing for Anthropic models
     module Anthropic
-      extend self
+      module_function
 
       def determine_context_window(model_id)
         case model_id
@@ -43,8 +43,6 @@ module RubyLLM
         model_id.include?('claude-3')
       end
 
-      private
-
       def model_family(model_id)
         case model_id
         when /claude-3-5-sonnet/  then :claude35_sonnet

data/lib/ruby_llm/model_capabilities/deepseek.rb

@@ -0,0 +1,93 @@
+# frozen_string_literal: true
+
+module RubyLLM
+  module ModelCapabilities
+    # Determines capabilities and pricing for DeepSeek models
+    module DeepSeek
+      module_function
+
+      def context_window_for(model_id)
+        case model_id
+        when /deepseek-(?:chat|reasoner)/ then 64_000
+        else 32_768 # Sensible default
+        end
+      end
+
+      def max_tokens_for(_model_id)
+        8_192
+      end
+
+      def input_price_for(model_id)
+        PRICES.dig(model_family(model_id), :input_miss) || default_input_price
+      end
+
+      def output_price_for(model_id)
+        PRICES.dig(model_family(model_id), :output) || default_output_price
+      end
+
+      def cache_hit_price_for(model_id)
+        PRICES.dig(model_family(model_id), :input_hit) || default_cache_hit_price
+      end
+
+      def supports_vision?(_model_id)
+        true # Both deepseek-chat and deepseek-reasoner support vision
+      end
+
+      def supports_functions?(_model_id)
+        true # Both models support function calling
+      end
+
+      def supports_json_mode?(_model_id)
+        true # Both models support JSON mode
+      end
+
+      def format_display_name(model_id)
+        case model_id
+        when 'deepseek-chat' then 'DeepSeek V3'
+        when 'deepseek-reasoner' then 'DeepSeek R1'
+        else
+          model_id.split('-')
+                  .map(&:capitalize)
+                  .join(' ')
+        end
+      end
+
+      def model_type(_model_id)
+        'chat' # Both models are chat models
+      end
+
+      def model_family(model_id)
+        case model_id
+        when /deepseek-chat/ then 'deepseek'
+        when /deepseek-reasoner/ then 'deepseek_reasoner'
+        else 'deepseek' # Default to base deepseek family
+        end
+      end
+
+      PRICES = {
+        chat: {
+          input_hit: 0.07,   # $0.07 per million tokens on cache hit
+          input_miss: 0.27,  # $0.27 per million tokens on cache miss
+          output: 1.10       # $1.10 per million tokens output
+        },
+        reasoner: {
+          input_hit: 0.14,   # $0.14 per million tokens on cache hit
+          input_miss: 0.55,  # $0.55 per million tokens on cache miss
+          output: 2.19       # $2.19 per million tokens output
+        }
+      }.freeze
+
+      def default_input_price
+        0.27 # Default to chat cache miss price
+      end
+
+      def default_output_price
+        1.10 # Default to chat output price
+      end
+
+      def default_cache_hit_price
+        0.07 # Default to chat cache hit price
+      end
+    end
+  end
+end

data/lib/ruby_llm/model_capabilities/gemini.rb

@@ -0,0 +1,132 @@
+# frozen_string_literal: true
+
+module RubyLLM
+  module ModelCapabilities
+    # Determines capabilities and pricing for Google Gemini models
+    module Gemini # rubocop:disable Metrics/ModuleLength
+      module_function
+
+      def context_window_for(model_id)
+        case model_id
+        when /gemini-2\.0-flash/ then 1_048_576
+        when /gemini-1\.5-pro/ then 2_097_152
+        when /gemini-1\.5/ then 1_048_576
+        when /text-embedding/, /embedding-001/ then 2_048
+        else 32_768 # Sensible default for unknown models
+        end
+      end
+
+      def max_tokens_for(model_id)
+        case model_id
+        when /gemini-2\.0-flash/, /gemini-1\.5/ then 8_192
+        when /text-embedding/, /embedding-001/ then 768 # Output dimension size for embeddings
+        when /aqa/ then 1_024
+        else 4_096 # Sensible default
+        end
+      end
+
+      def input_price_for(model_id)
+        PRICES.dig(pricing_family(model_id), :input) || default_input_price
+      end
+
+      def output_price_for(model_id)
+        PRICES.dig(pricing_family(model_id), :output) || default_output_price
+      end
+
+      def supports_vision?(model_id)
+        return false if model_id.match?(/text-embedding|embedding-001|aqa/)
+        return false if model_id.match?(/flash-lite/)
+        return false if model_id.match?(/imagen/)
+
+        # Only pro and regular flash models support vision
+        model_id.match?(/gemini-[12]\.(?:5|0)-(?:pro|flash)(?!-lite)/)
+      end
+
+      def supports_functions?(model_id)
+        return false if model_id.match?(/text-embedding|embedding-001|aqa/)
+        return false if model_id.match?(/imagen/)
+        return false if model_id.match?(/flash-lite/)
+        return false if model_id.match?(/bison|gecko|evergreen/)
+
+        # Currently only full models support function calling
+        model_id.match?(/gemini-[12]\.(?:5|0)-(?:pro|flash)(?!-lite)/)
+      end
+
+      def supports_json_mode?(model_id)
+        return false if model_id.match?(/text-embedding|embedding-001|aqa/)
+        return false if model_id.match?(/imagen/)
+        return false if model_id.match?(/flash-lite/)
+        return false if model_id.match?(/bison|gecko|evergreen/)
+
+        # Gemini 1.5+ models support JSON mode
+        model_id.match?(/gemini-[12]\.(?:5|0)-(?:pro|flash)(?!-lite)/)
+      end
+
+      def format_display_name(model_id)
+        return model_id unless model_id.start_with?('models/')
+
+        model_id
+          .delete_prefix('models/')
+          .split('-')
+          .map(&:capitalize)
+          .join(' ')
+          .gsub(/(\d+\.\d+)/, ' \1') # Add space before version numbers
+          .gsub(/\s+/, ' ')          # Clean up multiple spaces
+          .strip
+      end
+
+      def model_type(model_id)
+        case model_id
+        when /text-embedding|embedding/ then 'embedding'
+        when /imagen/ then 'image'
+        when /bison|text-bison/ then 'legacy'
+        else 'chat'
+        end
+      end
+
+      def model_family(model_id) # rubocop:disable Metrics/CyclomaticComplexity,Metrics/MethodLength
+        case model_id
+        when /gemini-2\.0-flash-lite/ then 'gemini20_flash_lite'
+        when /gemini-2\.0-flash/ then 'gemini20_flash'
+        when /gemini-1\.5-flash-8b/ then 'gemini15_flash_8b'
+        when /gemini-1\.5-flash/ then 'gemini15_flash'
+        when /gemini-1\.5-pro/ then 'gemini15_pro'
+        when /text-embedding-004/ then 'embedding4'
+        when /embedding-001/ then 'embedding1'
+        when /bison|text-bison/ then 'bison'
+        when /imagen/ then 'imagen3'
+        else 'other'
+        end
+      end
+
+      def pricing_family(model_id)
+        case model_id
+        when /gemini-2\.0-flash-lite/ then :flash_lite_2 # rubocop:disable Naming/VariableNumber
+        when /gemini-2\.0-flash/ then :flash_2 # rubocop:disable Naming/VariableNumber
+        when /gemini-1\.5-flash-8b/ then :flash_8b
+        when /gemini-1\.5-flash/ then :flash
+        when /gemini-1\.5-pro/ then :pro
+        when /text-embedding|embedding/ then :embedding
+        else :base
+        end
+      end
+
+      PRICES = {
+        flash_2: { input: 0.10, output: 0.40 },       # Gemini 2.0 Flash # rubocop:disable Naming/VariableNumber
+        flash_lite_2: { input: 0.075, output: 0.30 }, # Gemini 2.0 Flash Lite # rubocop:disable Naming/VariableNumber
+        flash: { input: 0.075, output: 0.30 },        # Gemini 1.5 Flash basic pricing
+        flash_8b: { input: 0.0375, output: 0.15 },    # Gemini 1.5 Flash 8B
+        pro: { input: 1.25, output: 5.0 },            # Gemini 1.5 Pro
+        embedding: { input: 0.00, output: 0.00 }      # Text Embedding models are free
+      }.freeze
+
+      def default_input_price
+        0.075 # Default to Flash pricing
+      end
+
+      def default_output_price
+        0.30  # Default to Flash pricing
+      end
+    end
+  end
+end

data/lib/ruby_llm/model_capabilities/openai.rb

@@ -3,7 +3,7 @@
 module RubyLLM
   module ModelCapabilities
     # Determines capabilities and pricing for OpenAI models
-    module OpenAI
+    module OpenAI # rubocop:disable Metrics/ModuleLength
       extend self
 
       def context_window_for(model_id)
@@ -55,21 +55,45 @@ module RubyLLM
                 .then { |name| apply_special_formatting(name) }
       end
 
-      private
+      def model_type(model_id)
+        case model_id
+        when /text-embedding|embedding/ then 'embedding'
+        when /dall-e/ then 'image'
+        when /tts|whisper/ then 'audio'
+        when /omni-moderation/ then 'moderation'
+        else 'chat'
+        end
+      end
 
-      def model_family(model_id) # rubocop:disable Metrics/CyclomaticComplexity
+      def model_family(model_id) # rubocop:disable Metrics/AbcSize,Metrics/CyclomaticComplexity,Metrics/MethodLength
         case model_id
-        when /o1-2024/                then :o1
-        when /o1-mini/                then :o1_mini
-        when /gpt-4o-realtime/        then :gpt4o_realtime
-        when /gpt-4o-mini-realtime/   then :gpt4o_mini_realtime
-        when /gpt-4o-mini/            then :gpt4o_mini
-        when /gpt-4o/                 then :gpt4o
-        when /gpt-4-turbo/            then :gpt4_turbo
-        else :gpt35
+        when /o1-2024|o1-mini-2024/ then 'o1'
+        when /o1-mini/ then 'o1_mini'
+        when /o1/ then 'o1'
+        when /gpt-4o-realtime/ then 'gpt4o_realtime'
+        when /gpt-4o-mini-realtime/ then 'gpt4o_mini_realtime'
+        when /gpt-4o-mini/ then 'gpt4o_mini'
+        when /gpt-4o/ then 'gpt4o'
+        when /gpt-4-turbo/ then 'gpt4_turbo'
+        when /gpt-4/ then 'gpt4'
+        when /gpt-3.5/ then 'gpt35'
+        when /dall-e-3/ then 'dalle3'
+        when /dall-e-2/ then 'dalle2'
+        when /text-embedding-3-large/ then 'embedding3_large'
+        when /text-embedding-3-small/ then 'embedding3_small'
+        when /text-embedding-ada/ then 'embedding2'
+        when /tts-1-hd/ then 'tts1_hd'
+        when /tts-1/ then 'tts1'
+        when /whisper/ then 'whisper1'
+        when /omni-moderation/ then 'moderation'
+        when /babbage/ then 'babbage'
+        when /davinci/ then 'davinci'
+        else 'other'
         end
       end
 
+      private
+
       PRICES = {
         o1: { input: 15.0, output: 60.0 },
         o1_mini: { input: 3.0, output: 12.0 },

data/lib/ruby_llm/model_info.rb

@@ -15,7 +15,7 @@ module RubyLLM
   class ModelInfo
     attr_reader :id, :created_at, :display_name, :provider, :metadata,
                 :context_window, :max_tokens, :supports_vision, :supports_functions,
-                :supports_json_mode, :input_price_per_million, :output_price_per_million
+                :supports_json_mode, :input_price_per_million, :output_price_per_million, :type, :family
 
     def initialize(data) # rubocop:disable Metrics/AbcSize,Metrics/MethodLength
       @id = data[:id]
@@ -24,6 +24,8 @@ module RubyLLM
       @provider = data[:provider]
       @context_window = data[:context_window]
       @max_tokens = data[:max_tokens]
+      @type = data[:type]
+      @family = data[:family]
       @supports_vision = data[:supports_vision]
       @supports_functions = data[:supports_functions]
       @supports_json_mode = data[:supports_json_mode]
@@ -35,11 +37,13 @@ module RubyLLM
     def to_h # rubocop:disable Metrics/MethodLength
       {
         id: id,
-        created_at: created_at.iso8601,
+        created_at: created_at&.iso8601,
         display_name: display_name,
         provider: provider,
         context_window: context_window,
         max_tokens: max_tokens,
+        type: type,
+        family: family,
         supports_vision: supports_vision,
         supports_functions: supports_functions,
         supports_json_mode: supports_json_mode,
@@ -48,13 +52,5 @@ module RubyLLM
         metadata: metadata
       }
     end
-
-    def type
-      metadata['type']
-    end
-
-    def family
-      metadata['family']
-    end
   end
 end