ruby_llm_community 0.0.5 → 1.0.0

data/lib/ruby_llm/active_record/acts_as.rb

@@ -3,17 +3,19 @@
 module RubyLLM
   module ActiveRecord
     # Adds chat and message persistence capabilities to ActiveRecord models.
-    # Provides a clean interface for storing chat history, message metadata,
-    # and attachments in your database.
     module ActsAs
       extend ActiveSupport::Concern
 
       class_methods do # rubocop:disable Metrics/BlockLength
-        def acts_as_chat(message_class: 'Message', tool_call_class: 'ToolCall')
+        def acts_as_chat(message_class: 'Message', tool_call_class: 'ToolCall',
+                         model_class: 'Model', model_foreign_key: nil, model_primary_key: nil)
           include ChatMethods
 
           @message_class = message_class.to_s
           @tool_call_class = tool_call_class.to_s
+          @model_class = model_class.to_s
+          @model_foreign_key = model_foreign_key || ActiveSupport::Inflector.foreign_key(@model_class)
+          @model_primary_key = model_primary_key || 'model_id'
 
           has_many :messages,
                    -> { order(created_at: :asc) },
@@ -21,13 +23,43 @@ module RubyLLM
                    inverse_of: :chat,
                    dependent: :destroy
 
+          # Set up model association if model registry is configured
+          if RubyLLM.config.model_registry_class
+            belongs_to :model,
+                       class_name: @model_class,
+                       foreign_key: @model_foreign_key,
+                       primary_key: @model_primary_key,
+                       optional: true
+          end
+
           delegate :add_message, to: :to_llm
         end
 
-        def acts_as_message(chat_class: 'Chat',
+        def acts_as_model(chat_class: 'Chat')
+          include ModelMethods
+
+          @chat_class = chat_class.to_s
+
+          validates :model_id, presence: true, uniqueness: { scope: :provider }
+          validates :name, presence: true
+          validates :provider, presence: true
+
+          # Set up chat association if configured
+          return unless RubyLLM.config.model_registry_class
+
+          has_many :chats,
+                   class_name: @chat_class,
+                   foreign_key: 'model_id',
+                   primary_key: 'model_id'
+        end
+
+        def acts_as_message(chat_class: 'Chat', # rubocop:disable Metrics/ParameterLists
                             chat_foreign_key: nil,
                             tool_call_class: 'ToolCall',
                             tool_call_foreign_key: nil,
+                            model_class: 'Model',
+                            model_foreign_key: nil,
+                            model_primary_key: nil,
                             touch_chat: false)
           include MessageMethods
 
@@ -37,6 +69,10 @@ module RubyLLM
           @tool_call_class = tool_call_class.to_s
           @tool_call_foreign_key = tool_call_foreign_key || ActiveSupport::Inflector.foreign_key(@tool_call_class)
 
+          @model_class = model_class.to_s
+          @model_foreign_key = model_foreign_key || ActiveSupport::Inflector.foreign_key(@model_class)
+          @model_primary_key = model_primary_key || 'model_id'
+
           belongs_to :chat,
                      class_name: @chat_class,
                      foreign_key: @chat_foreign_key,
@@ -53,7 +89,21 @@ module RubyLLM
                      optional: true,
                      inverse_of: :result
 
-          delegate :tool_call?, :tool_result?, :tool_results, to: :to_llm
+          has_many :tool_results,
+                   through: :tool_calls,
+                   source: :result,
+                   class_name: @message_class
+
+          # Set up model association if model registry is configured
+          if RubyLLM.config.model_registry_class
+            belongs_to :model,
+                       class_name: @model_class,
+                       foreign_key: @model_foreign_key,
+                       primary_key: @model_primary_key,
+                       optional: true
+          end
+
+          delegate :tool_call?, :tool_result?, to: :to_llm
         end
 
         def acts_as_tool_call(message_class: 'Message', message_foreign_key: nil, result_foreign_key: nil)
@@ -75,8 +125,7 @@ module RubyLLM
       end
     end
 
-    # Methods mixed into chat models to handle message persistence and
-    # provide a conversation interface.
+    # Methods mixed into chat models.
     module ChatMethods
       extend ActiveSupport::Concern
 
@@ -85,10 +134,20 @@ module RubyLLM
       end
 
       def to_llm(context: nil)
+        # If we have a model association, use both model_id and provider
+        # Otherwise, model_id is a string that RubyLLM can resolve
+        if respond_to?(:model) && model
+          model_to_use = model.model_id
+          provider_to_use = model.provider.to_sym
+        else
+          model_to_use = model_id
+          provider_to_use = nil
+        end
+
         @chat ||= if context
-                    context.chat(model: model_id)
+                    context.chat(model: model_to_use, provider: provider_to_use)
                   else
-                    RubyLLM.chat(model: model_id)
+                    RubyLLM.chat(model: model_to_use, provider: provider_to_use)
                   end
         @chat.reset_messages!
 
@@ -148,6 +207,11 @@ module RubyLLM
         self
       end
 
+      def cache_prompts(...)
+        to_llm.cache_prompts(...)
+        self
+      end
+
       def on_new_message(&block)
         to_llm
 
@@ -206,26 +270,33 @@ module RubyLLM
       private
 
       def cleanup_failed_messages
-        RubyLLM.logger.debug "RubyLLM: API call failed, destroying message: #{@message.id}"
+        RubyLLM.logger.warn "RubyLLM: API call failed, destroying message: #{@message.id}"
         @message.destroy
       end
 
-      def cleanup_orphaned_tool_results
-        loop do
-          messages.reload
-          last = messages.order(:id).last
+      def cleanup_orphaned_tool_results # rubocop:disable Metrics/PerceivedComplexity
+        messages.reload
+        last = messages.order(:id).last
 
-          break unless last&.tool_call? || last&.tool_result?
+        return unless last&.tool_call? || last&.tool_result?
 
+        if last.tool_call?
           last.destroy
+        elsif last.tool_result?
+          tool_call_message = last.parent_tool_call.message
+          expected_results = tool_call_message.tool_calls.pluck(:id)
+          actual_results = tool_call_message.tool_results.pluck(:tool_call_id)
+
+          if expected_results.sort != actual_results.sort
+            tool_call_message.tool_results.each(&:destroy)
+            tool_call_message.destroy
+          end
         end
       end
 
       def setup_persistence_callbacks
-        # Only set up once per chat instance
         return @chat if @chat.instance_variable_get(:@_persistence_callbacks_setup)
 
-        # Set up persistence callbacks (user callbacks will be chained via on_new_message/on_end_message methods)
         @chat.on_new_message { persist_new_message }
         @chat.on_end_message { |msg| persist_message_completion(msg) }
 
@@ -237,15 +308,21 @@ module RubyLLM
         @message = messages.create!(role: :assistant, content: '')
       end
 
-      def persist_message_completion(message)
+      def persist_message_completion(message) # rubocop:disable Metrics/PerceivedComplexity
         return unless message
 
         tool_call_id = find_tool_call_id(message.tool_call_id) if message.tool_call_id
 
         transaction do
-          # Convert parsed JSON back to JSON string for storage
           content = message.content
-          content = content.to_json if content.is_a?(Hash) || content.is_a?(Array)
+          attachments_to_persist = nil
+
+          if content.is_a?(RubyLLM::Content)
+            attachments_to_persist = content.attachments if content.attachments.any?
+            content = content.text
+          elsif content.is_a?(Hash) || content.is_a?(Array)
+            content = content.to_json
+          end
 
           @message.update!(
             role: message.role,
@@ -256,6 +333,8 @@ module RubyLLM
           )
           @message.write_attribute(@message.class.tool_call_foreign_key, tool_call_id) if tool_call_id
           @message.save!
+
+          persist_content(@message, attachments_to_persist) if attachments_to_persist
           persist_tool_calls(message.tool_calls) if message.tool_calls.present?
         end
       end
@@ -297,8 +376,7 @@ module RubyLLM
       def convert_to_active_storage_format(source)
         return if source.blank?
 
-        # Let RubyLLM::Attachment handle the heavy lifting
-        attachment = RubyLLM::Attachment.new(source)
+        attachment = source.is_a?(RubyLLM::Attachment) ? source : RubyLLM::Attachment.new(source)
 
         {
           io: StringIO.new(attachment.content),
@@ -311,8 +389,7 @@ module RubyLLM
       end
     end
 
-    # Methods mixed into message models to handle serialization and
-    # provide a clean interface to the underlying message data.
+    # Methods mixed into message models.
     module MessageMethods
       extend ActiveSupport::Concern
 
@@ -378,5 +455,72 @@ module RubyLLM
         tempfile
       end
     end
+
+    # Methods mixed into model registry models.
+    module ModelMethods
+      extend ActiveSupport::Concern
+
+      class_methods do # rubocop:disable Metrics/BlockLength
+        def refresh!
+          RubyLLM.models.refresh!
+
+          transaction do
+            RubyLLM.models.all.each do |model_info|
+              model = find_or_initialize_by(
+                model_id: model_info.id,
+                provider: model_info.provider
+              )
+              model.update!(from_llm_attributes(model_info))
+            end
+          end
+        end
+
+        def from_llm(model_info)
+          new(from_llm_attributes(model_info))
+        end
+
+        private
+
+        def from_llm_attributes(model_info)
+          {
+            model_id: model_info.id,
+            name: model_info.name,
+            provider: model_info.provider,
+            family: model_info.family,
+            model_created_at: model_info.created_at,
+            context_window: model_info.context_window,
+            max_output_tokens: model_info.max_output_tokens,
+            knowledge_cutoff: model_info.knowledge_cutoff,
+            modalities: model_info.modalities.to_h,
+            capabilities: model_info.capabilities,
+            pricing: model_info.pricing.to_h,
+            metadata: model_info.metadata
+          }
+        end
+      end
+
+      def to_llm
+        RubyLLM::Model::Info.new(
+          id: model_id,
+          name: name,
+          provider: provider,
+          family: family,
+          created_at: model_created_at,
+          context_window: context_window,
+          max_output_tokens: max_output_tokens,
+          knowledge_cutoff: knowledge_cutoff,
+          modalities: modalities&.deep_symbolize_keys || {},
+          capabilities: capabilities,
+          pricing: pricing&.deep_symbolize_keys || {},
+          metadata: metadata&.deep_symbolize_keys || {}
+        )
+      end
+
+      delegate :supports?, :supports_vision?, :supports_functions?, :type,
+               :input_price_per_million, :output_price_per_million,
+               :function_calling?, :structured_output?, :batch?,
+               :reasoning?, :citations?, :streaming?,
+               to: :to_llm
+    end
   end
 end

data/lib/ruby_llm/aliases.json

@@ -50,21 +50,57 @@
     "deepseek": "deepseek-chat",
     "openrouter": "deepseek/deepseek-chat"
   },
+  "gemini-1.5-flash": {
+    "gemini": "gemini-1.5-flash",
+    "vertexai": "gemini-1.5-flash"
+  },
+  "gemini-1.5-flash-002": {
+    "gemini": "gemini-1.5-flash-002",
+    "vertexai": "gemini-1.5-flash-002"
+  },
+  "gemini-1.5-flash-8b": {
+    "gemini": "gemini-1.5-flash-8b",
+    "vertexai": "gemini-1.5-flash-8b"
+  },
+  "gemini-1.5-pro": {
+    "gemini": "gemini-1.5-pro",
+    "vertexai": "gemini-1.5-pro"
+  },
+  "gemini-1.5-pro-002": {
+    "gemini": "gemini-1.5-pro-002",
+    "vertexai": "gemini-1.5-pro-002"
+  },
+  "gemini-2.0-flash": {
+    "gemini": "gemini-2.0-flash",
+    "vertexai": "gemini-2.0-flash"
+  },
   "gemini-2.0-flash-001": {
     "gemini": "gemini-2.0-flash-001",
-    "openrouter": "google/gemini-2.0-flash-001"
+    "openrouter": "google/gemini-2.0-flash-001",
+    "vertexai": "gemini-2.0-flash-001"
+  },
+  "gemini-2.0-flash-exp": {
+    "gemini": "gemini-2.0-flash-exp",
+    "vertexai": "gemini-2.0-flash-exp"
   },
   "gemini-2.0-flash-lite-001": {
     "gemini": "gemini-2.0-flash-lite-001",
-    "openrouter": "google/gemini-2.0-flash-lite-001"
+    "openrouter": "google/gemini-2.0-flash-lite-001",
+    "vertexai": "gemini-2.0-flash-lite-001"
   },
   "gemini-2.5-flash": {
     "gemini": "gemini-2.5-flash",
-    "openrouter": "google/gemini-2.5-flash"
+    "openrouter": "google/gemini-2.5-flash",
+    "vertexai": "gemini-2.5-flash"
+  },
+  "gemini-2.5-flash-image-preview": {
+    "gemini": "gemini-2.5-flash-image-preview",
+    "openrouter": "google/gemini-2.5-flash-image-preview"
   },
   "gemini-2.5-flash-lite": {
     "gemini": "gemini-2.5-flash-lite",
-    "openrouter": "google/gemini-2.5-flash-lite"
+    "openrouter": "google/gemini-2.5-flash-lite",
+    "vertexai": "gemini-2.5-flash-lite"
   },
   "gemini-2.5-flash-lite-preview-06-17": {
     "gemini": "gemini-2.5-flash-lite-preview-06-17",
@@ -72,12 +108,21 @@
   },
   "gemini-2.5-pro": {
     "gemini": "gemini-2.5-pro",
-    "openrouter": "google/gemini-2.5-pro"
+    "openrouter": "google/gemini-2.5-pro",
+    "vertexai": "gemini-2.5-pro"
   },
   "gemini-2.5-pro-preview-05-06": {
     "gemini": "gemini-2.5-pro-preview-05-06",
     "openrouter": "google/gemini-2.5-pro-preview-05-06"
   },
+  "gemini-embedding-001": {
+    "gemini": "gemini-embedding-001",
+    "vertexai": "gemini-embedding-001"
+  },
+  "gemini-exp-1206": {
+    "gemini": "gemini-exp-1206",
+    "vertexai": "gemini-exp-1206"
+  },
   "gemma-3-12b-it": {
     "gemini": "gemma-3-12b-it",
     "openrouter": "google/gemma-3-12b-it"
@@ -150,6 +195,10 @@
     "openai": "gpt-4o-2024-11-20",
     "openrouter": "openai/gpt-4o-2024-11-20"
   },
+  "gpt-4o-audio-preview": {
+    "openai": "gpt-4o-audio-preview",
+    "openrouter": "openai/gpt-4o-audio-preview"
+  },
   "gpt-4o-mini": {
     "openai": "gpt-4o-mini",
     "openrouter": "openai/gpt-4o-mini"
@@ -182,6 +231,10 @@
     "openai": "gpt-oss-120b",
     "openrouter": "openai/gpt-oss-120b"
   },
+  "gpt-oss-20b": {
+    "openai": "gpt-oss-20b",
+    "openrouter": "openai/gpt-oss-20b"
+  },
   "o1": {
     "openai": "o1",
     "openrouter": "openai/o1"
@@ -213,5 +266,9 @@
   "o4-mini": {
     "openai": "o4-mini",
     "openrouter": "openai/o4-mini"
+  },
+  "text-embedding-004": {
+    "gemini": "text-embedding-004",
+    "vertexai": "text-embedding-004"
   }
 }

data/lib/ruby_llm/aliases.rb

@@ -1,53 +1,35 @@
 # frozen_string_literal: true
 
 module RubyLLM
-  # Manages model aliases, allowing users to reference models by simpler names
-  # that map to specific model versions across different providers.
-  #
-  # Aliases are defined in aliases.json and follow the format:
-  #   {
-  #     "simple-name": {
-  #       "provider1": "specific-version-for-provider1",
-  #       "provider2": "specific-version-for-provider2"
-  #     }
-  #   }
+  # Manages model aliases for provider-specific versions
   class Aliases
     class << self
-      # Resolves a model ID to its provider-specific version
-      #
-      # @param model_id [String] the model identifier or alias
-      # @param provider_slug [String, Symbol, nil] optional provider to resolve for
-      # @return [String] the resolved model ID or the original if no alias exists
       def resolve(model_id, provider = nil)
         return model_id unless aliases[model_id]
 
         if provider
           aliases[model_id][provider.to_s] || model_id
         else
-          # Get native provider's version
           aliases[model_id].values.first || model_id
         end
       end
 
-      # Returns the loaded aliases mapping
-      # @return [Hash] the aliases mapping
       def aliases
         @aliases ||= load_aliases
       end
 
-      # Loads aliases from the JSON file
-      # @return [Hash] the loaded aliases
+      def aliases_file
+        File.expand_path('aliases.json', __dir__)
+      end
+
       def load_aliases
-        file_path = File.expand_path('aliases.json', __dir__)
-        if File.exist?(file_path)
-          JSON.parse(File.read(file_path))
+        if File.exist?(aliases_file)
+          JSON.parse(File.read(aliases_file))
         else
           {}
         end
       end
 
-      # Reloads aliases from disk
-      # @return [Hash] the reloaded aliases
       def reload!
         @aliases = load_aliases
       end

data/lib/ruby_llm/chat.rb

@@ -1,13 +1,7 @@
 # frozen_string_literal: true
 
 module RubyLLM
-  # Represents a conversation with an AI model. Handles message history,
-  # streaming responses, and tool integration with a simple, conversational API.
-  #
-  # Example:
-  #   chat = RubyLLM.chat
-  #   chat.ask "What's the best way to learn Ruby?"
-  #   chat.ask "Can you elaborate on that?"
+  # Represents a conversation with an AI model
   class Chat
     include Enumerable
 
@@ -22,7 +16,7 @@ module RubyLLM
       @config = context&.config || RubyLLM.config
       model_id = model || @config.default_model
       with_model(model_id, provider: provider, assume_exists: assume_model_exists)
-      @temperature = 0.7
+      @temperature = nil
       @messages = []
       @tools = {}
       @cache_prompts = { system: false, user: false, tools: false }
@@ -138,8 +132,8 @@ module RubyLLM
         messages,
         tools: @tools,
         temperature: @temperature,
-        model: @model.id,
         cache_prompts: @cache_prompts.dup,
+        model: @model,
         params: @params,
         headers: @headers,
         schema: @schema,
@@ -148,7 +142,6 @@ module RubyLLM
 
       @on[:new_message]&.call unless block_given?
 
-      # Parse JSON if schema was set
       if @schema && response.content.is_a?(String)
         begin
           response.content = JSON.parse(response.content)
@@ -177,6 +170,10 @@ module RubyLLM
       @messages.clear
     end
 
+    def instance_variables
+      super - %i[@connection @config]
+    end
+
     private
 
     def wrap_streaming_block(&block)
@@ -191,12 +188,11 @@ module RubyLLM
           @on[:new_message]&.call
         end
 
-        # Pass chunk to user's block
         block.call chunk
       end
     end
 
-    def handle_tool_calls(response, &)
+    def handle_tool_calls(response, &) # rubocop:disable Metrics/PerceivedComplexity
       halt_result = nil
 
       response.tool_calls.each_value do |tool_call|
@@ -204,7 +200,8 @@ module RubyLLM
         @on[:tool_call]&.call(tool_call)
         result = execute_tool tool_call
         @on[:tool_result]&.call(result)
-        message = add_message role: :tool, content: result.to_s, tool_call_id: tool_call.id
+        content = result.is_a?(Content) ? result : result.to_s
+        message = add_message role: :tool, content:, tool_call_id: tool_call.id
         @on[:end_message]&.call(message)
 
         halt_result = result if result.is_a?(Tool::Halt)
@@ -218,9 +215,5 @@ module RubyLLM
       args = tool_call.arguments
       tool.call(args)
     end
-
-    def instance_variables
-      super - %i[@connection @config]
-    end
   end
 end

data/lib/ruby_llm/configuration.rb

@@ -1,16 +1,8 @@
 # frozen_string_literal: true
 
 module RubyLLM
-  # Global configuration for RubyLLM. Manages API keys, default models,
-  # and provider-specific settings.
-  #
-  # Configure via:
-  #   RubyLLM.configure do |config|
-  #     config.openai_api_key = ENV['OPENAI_API_KEY']
-  #     config.anthropic_api_key = ENV['ANTHROPIC_API_KEY']
-  #   end
+  # Global configuration for RubyLLM
   class Configuration
-    # Provider-specific configuration
     attr_accessor :openai_api_key,
                   :openai_api_base,
                   :openai_organization_id,
@@ -18,6 +10,8 @@ module RubyLLM
                   :openai_use_system_role,
                   :anthropic_api_key,
                   :gemini_api_key,
+                  :vertexai_project_id,
+                  :vertexai_location,
                   :deepseek_api_key,
                   :perplexity_api_key,
                   :bedrock_api_key,
@@ -33,6 +27,8 @@ module RubyLLM
                   :default_model,
                   :default_embedding_model,
                   :default_image_model,
+                  # Model registry
+                  :model_registry_class,
                   # Connection configuration
                   :request_timeout,
                   :max_retries,
@@ -47,7 +43,6 @@ module RubyLLM
                   :log_stream_debug
 
     def initialize
-      # Connection configuration
       @request_timeout = 120
       @max_retries = 3
       @retry_interval = 0.1
@@ -55,12 +50,10 @@ module RubyLLM
       @retry_interval_randomness = 0.5
       @http_proxy = nil
 
-      # Default models
       @default_model = 'gpt-4.1-nano'
       @default_embedding_model = 'text-embedding-3-small'
       @default_image_model = 'gpt-image-1'
 
-      # Logging configuration
       @log_file = $stdout
       @log_level = ENV['RUBYLLM_DEBUG'] ? Logger::DEBUG : Logger::INFO
       @log_stream_debug = ENV['RUBYLLM_STREAM_DEBUG'] == 'true'

data/lib/ruby_llm/connection.rb

@@ -48,6 +48,10 @@ module RubyLLM
       end
     end
 
+    def instance_variables
+      super - %i[@config @connection]
+    end
+
     private
 
     def setup_timeout(faraday)
@@ -118,9 +122,5 @@ module RubyLLM
       raise ConfigurationError,
             "#{@provider.name} provider is not configured. Add this to your initialization:\n\n#{config_block}"
     end
-
-    def instance_variables
-      super - %i[@config @connection]
-    end
   end
 end

data/lib/ruby_llm/connection_multipart.rb

@@ -0,0 +1,19 @@
+# frozen_string_literal: true
+
+module RubyLLM
+  # A connection that uses multipart/form-data for file uploads
+  class ConnectionMultipart < Connection
+    def post(url, payload, &)
+      @connection.post url, payload do |req|
+        req.headers.merge! @provider.headers if @provider.respond_to?(:headers)
+        req.headers['Content-Type'] = 'multipart/form-data'
+        yield req if block_given?
+      end
+    end
+
+    def setup_middleware(faraday)
+      super
+      faraday.request :multipart, content_type: 'multipart/form-data'
+    end
+  end
+end

data/lib/ruby_llm/content.rb

@@ -2,7 +2,6 @@
 
 module RubyLLM
   # Represents the content sent to or received from an LLM.
-  # Selects the appropriate attachment class based on the content type.
   class Content
     attr_reader :text, :attachments
 
@@ -19,6 +18,11 @@ module RubyLLM
       self
     end
 
+    def attach(attachment)
+      @attachments << attachment
+      self
+    end
+
     def format
       if @text && @attachments.empty?
         @text
@@ -42,7 +46,6 @@ module RubyLLM
 
     def process_attachments(attachments)
       if attachments.is_a?(Hash)
-        # Ignores types (like :image, :audio, :text, :pdf) since we have robust MIME type detection
         attachments.each_value { |attachment| process_attachments_array_or_string(attachment) }
       else
         process_attachments_array_or_string attachments