ruby_llm_community 0.0.6 → 1.1.0

data/lib/ruby_llm/models.rb

@@ -1,14 +1,7 @@
 # frozen_string_literal: true
 
 module RubyLLM
-  # Registry of available AI models and their capabilities. Provides a clean interface
-  # to discover and work with models from different providers.
-  #
-  # Example:
-  #   RubyLLM.models.all                                  # All available models
-  #   RubyLLM.models.chat_models                          # Models that support chat
-  #   RubyLLM.models.by_provider('openai').chat_models    # OpenAI chat models
-  #   RubyLLM.models.find('claude-3')                     # Get info about a specific model
+  # Registry of available AI models and their capabilities.
   class Models
     include Enumerable
 
@@ -25,20 +18,24 @@ module RubyLLM
         File.expand_path('models.json', __dir__)
       end
 
-      def refresh!
-        # Collect models from both sources
-        provider_models = fetch_from_providers
-        parsera_models = fetch_from_parsera
+      def schema_file
+        File.expand_path('models_schema.json', __dir__)
+      end
 
-        # Merge with parsera data taking precedence
+      def refresh!(remote_only: false)
+        provider_models = fetch_from_providers(remote_only: remote_only)
+        parsera_models = fetch_from_parsera
         merged_models = merge_models(provider_models, parsera_models)
-
         @instance = new(merged_models)
       end
 
-      def fetch_from_providers
+      def fetch_from_providers(remote_only: true)
         config = RubyLLM.config
-        configured_classes = Provider.configured_remote_providers(config)
+        configured_classes = if remote_only
+                               Provider.configured_remote_providers(config)
+                             else
+                               Provider.configured_providers(config)
+                             end
         configured = configured_classes.map { |klass| klass.new(config) }
 
         RubyLLM.logger.info "Fetching models from providers: #{configured.map(&:name).join(', ')}"
@@ -50,7 +47,6 @@ module RubyLLM
         config ||= RubyLLM.config
         provider_class = provider ? Provider.providers[provider.to_sym] : nil
 
-        # Check if provider is local
         if provider_class
           temp_instance = provider_class.new(config)
           assume_exists = true if temp_instance.local?
@@ -62,14 +58,15 @@ module RubyLLM
           provider_class ||= raise(Error, "Unknown provider: #{provider.to_sym}")
           provider_instance = provider_class.new(config)
 
-          model = Model::Info.new(
-            id: model_id,
-            name: model_id.tr('-', ' ').capitalize,
-            provider: provider_instance.slug,
-            capabilities: %w[function_calling streaming],
-            modalities: { input: %w[text image], output: %w[text] },
-            metadata: { warning: 'Assuming model exists, capabilities may not be accurate' }
-          )
+          model = if provider_instance.local?
+                    begin
+                      Models.find(model_id, provider)
+                    rescue ModelNotFoundError
+                      nil
+                    end
+                  end
+
+          model ||= Model::Info.default(model_id, provider_instance.slug)
         else
           model = Models.find model_id, provider
           provider_class = Provider.providers[model.provider.to_sym] || raise(Error,
@@ -110,20 +107,36 @@ module RubyLLM
         all_keys = parsera_by_key.keys | provider_by_key.keys
 
         models = all_keys.map do |key|
-          if (parsera_model = parsera_by_key[key])
-            if (provider_model = provider_by_key[key])
-              add_provider_metadata(parsera_model, provider_model)
-            else
-              parsera_model
-            end
+          parsera_model = find_parsera_model(key, parsera_by_key)
+          provider_model = provider_by_key[key]
+
+          if parsera_model && provider_model
+            add_provider_metadata(parsera_model, provider_model)
+          elsif parsera_model
+            parsera_model
           else
-            provider_by_key[key]
+            provider_model
           end
         end
 
         models.sort_by { |m| [m.provider, m.id] }
       end
 
+      def find_parsera_model(key, parsera_by_key)
+        # Direct match
+        return parsera_by_key[key] if parsera_by_key[key]
+
+        # VertexAI uses same models as Gemini
+        provider, model_id = key.split(':', 2)
+        return unless provider == 'vertexai'
+
+        gemini_model = parsera_by_key["gemini:#{model_id}"]
+        return unless gemini_model
+
+        # Return Gemini's Parsera data but with VertexAI as provider
+        Model::Info.new(gemini_model.to_h.merge(provider: 'vertexai'))
+      end
+
       def index_by_key(models)
         models.each_with_object({}) do |model, hash|
           hash["#{model.provider}:#{model.id}"] = model
@@ -142,13 +155,21 @@ module RubyLLM
     end
 
     def load_models
+      read_from_json
+    end
+
+    def load_from_json!
+      @models = read_from_json
+    end
+
+    def read_from_json
       data = File.exist?(self.class.models_file) ? File.read(self.class.models_file) : '[]'
       JSON.parse(data, symbolize_names: true).map { |model| Model::Info.new(model) }
     rescue JSON::ParserError
       []
     end
 
-    def save_models
+    def save_to_json
       File.write(self.class.models_file, JSON.pretty_generate(all.map(&:to_h)))
     end
 
@@ -192,8 +213,8 @@ module RubyLLM
       self.class.new(all.select { |m| m.provider == provider.to_s })
     end
 
-    def refresh!
-      self.class.refresh!
+    def refresh!(remote_only: false)
+      self.class.refresh!(remote_only: remote_only)
     end
 
     private

data/lib/ruby_llm/provider.rb

@@ -1,10 +1,7 @@
 # frozen_string_literal: true
 
 module RubyLLM
-  # Base class for LLM providers like OpenAI and Anthropic.
-  # Handles the complexities of API communication, streaming responses,
-  # and error handling so individual providers can focus on their unique features.
-  # Encapsulates configuration and connection to eliminate parameter threading.
+  # Base class for LLM providers.
   class Provider
     include Streaming
 
@@ -45,7 +42,6 @@ module RubyLLM
       normalized_temperature = maybe_normalize_temperature(temperature, model)
 
       payload = Utils.deep_merge(
-        params,
         render_payload(
           messages,
           tools: tools,
@@ -54,7 +50,8 @@ module RubyLLM
           cache_prompts: cache_prompts,
           stream: block_given?,
           schema: schema
-        )
+        ),
+        params
       )
 
       if block_given?
@@ -65,8 +62,10 @@ module RubyLLM
     end
 
     def list_models
-      response = @connection.get models_url
-      parse_list_models_response response, slug, capabilities
+      Array(models_url).flat_map do |url|
+        response = @connection.get(url)
+        parse_list_models_response(response, slug, capabilities)
+      end
     end
 
     def embed(text, model:, dimensions:)
@@ -75,8 +74,8 @@ module RubyLLM
       parse_embedding_response(response, model:, text:)
     end
 
-    def paint(prompt, model:, size:)
-      payload = render_image_payload(prompt, model:, size:)
+    def paint(prompt, model:, size:, with:, params:)
+      payload = render_image_payload(prompt, model:, size:, with:, params:)
       response = @connection.post images_url, payload
       parse_image_response(response, model:)
     end
@@ -126,6 +125,10 @@ module RubyLLM
       nil
     end
 
+    def connection_multipart(config)
+      @connection_multipart ||= ConnectionMultipart.new(self, config)
+    end
+
     class << self
       def name
         to_s.split('::').last
@@ -206,13 +209,12 @@ module RubyLLM
       raise ConfigurationError, "Missing configuration for #{name}: #{missing.join(', ')}"
     end
 
-    def maybe_normalize_temperature(temperature, _model_id)
+    def maybe_normalize_temperature(temperature, _model)
       temperature
     end
 
     def sync_response(connection, payload, additional_headers = {})
       response = connection.post completion_url, payload do |req|
-        # Merge additional headers, with existing headers taking precedence
         req.headers = additional_headers.merge(req.headers) unless additional_headers.empty?
       end
       parse_completion_response response

data/lib/ruby_llm/providers/anthropic/capabilities.rb

@@ -7,17 +7,10 @@ module RubyLLM
       module Capabilities
         module_function
 
-        # Determines the context window size for a given model
-        # @param model_id [String] the model identifier
-        # @return [Integer] the context window size in tokens
         def determine_context_window(_model_id)
-          # All Claude 3 and 3.5 and 3.7 models have 200K token context windows
           200_000
         end
 
-        # Determines the maximum output tokens for a given model
-        # @param model_id [String] the model identifier
-        # @return [Integer] the maximum output tokens
         def determine_max_tokens(model_id)
           case model_id
           when /claude-3-7-sonnet/, /claude-3-5/ then 8_192
@@ -25,52 +18,30 @@ module RubyLLM
           end
         end
 
-        # Gets the input price per million tokens for a given model
-        # @param model_id [String] the model identifier
-        # @return [Float] the price per million tokens for input
         def get_input_price(model_id)
           PRICES.dig(model_family(model_id), :input) || default_input_price
         end
 
-        # Gets the output price per million tokens for a given model
-        # @param model_id [String] the model identifier
-        # @return [Float] the price per million tokens for output
         def get_output_price(model_id)
           PRICES.dig(model_family(model_id), :output) || default_output_price
         end
 
-        # Determines if a model supports vision capabilities
-        # @param model_id [String] the model identifier
-        # @return [Boolean] true if the model supports vision
         def supports_vision?(model_id)
-          # All Claude 3, 3.5, and 3.7 models support vision
           !model_id.match?(/claude-[12]/)
         end
 
-        # Determines if a model supports function calling
-        # @param model_id [String] the model identifier
-        # @return [Boolean] true if the model supports functions
         def supports_functions?(model_id)
           model_id.match?(/claude-3/)
         end
 
-        # Determines if a model supports JSON mode
-        # @param model_id [String] the model identifier
-        # @return [Boolean] true if the model supports JSON mode
         def supports_json_mode?(model_id)
           model_id.match?(/claude-3/)
         end
 
-        # Determines if a model supports extended thinking
-        # @param model_id [String] the model identifier
-        # @return [Boolean] true if the model supports extended thinking
         def supports_extended_thinking?(model_id)
           model_id.match?(/claude-3-7-sonnet/)
         end
 
-        # Determines the model family for a given model ID
-        # @param model_id [String] the model identifier
-        # @return [Symbol] the model family identifier
         def model_family(model_id)
           case model_id
           when /claude-3-7-sonnet/  then 'claude-3-7-sonnet'
@@ -83,14 +54,10 @@ module RubyLLM
           end
         end
 
-        # Returns the model type
-        # @param model_id [String] the model identifier (unused but kept for API consistency)
-        # @return [String] the model type, always 'chat' for Anthropic models
         def model_type(_)
           'chat'
         end
 
-        # Pricing information for Anthropic models (per million tokens)
         PRICES = {
           'claude-3-7-sonnet': { input: 3.0, output: 15.0 },
           'claude-3-5-sonnet': { input: 3.0, output: 15.0 },
@@ -100,14 +67,10 @@ module RubyLLM
           'claude-2': { input: 3.0, output: 15.0 }
         }.freeze
 
-        # Default input price if model not found in PRICES
-        # @return [Float] default price per million tokens for input
         def default_input_price
           3.0
         end
 
-        # Default output price if model not found in PRICES
-        # @return [Float] default price per million tokens for output
         def default_output_price
           15.0
         end
@@ -118,7 +81,6 @@ module RubyLLM
             output: ['text']
           }
 
-          # All Claude 3+ models support vision
           unless model_id.match?(/claude-[12]/)
             modalities[:input] << 'image'
             modalities[:input] << 'pdf'
@@ -130,18 +92,13 @@ module RubyLLM
         def capabilities_for(model_id)
           capabilities = ['streaming']
 
-          # Function calling for Claude 3+
           if model_id.match?(/claude-3/)
             capabilities << 'function_calling'
             capabilities << 'batch'
           end
 
-          # Extended thinking (reasoning) for Claude 3.7
-          capabilities << 'reasoning' if model_id.match?(/claude-3-7/)
-
-          # Citations
+          capabilities << 'reasoning' if model_id.match?(/claude-3-7|-4/)
           capabilities << 'citations' if model_id.match?(/claude-3\.5|claude-3-7/)
-
           capabilities
         end
 
@@ -154,13 +111,11 @@ module RubyLLM
             output_per_million: prices[:output]
           }
 
-          # Batch is typically half the price
           batch_pricing = {
             input_per_million: prices[:input] * 0.5,
             output_per_million: prices[:output] * 0.5
           }
 
-          # Add reasoning output pricing for 3.7 models
           if model_id.match?(/claude-3-7/)
             standard_pricing[:reasoning_output_per_million] = prices[:output] * 2.5
             batch_pricing[:reasoning_output_per_million] = prices[:output] * 1.25

data/lib/ruby_llm/providers/anthropic/chat.rb

@@ -40,10 +40,10 @@ module RubyLLM
           end
 
           {
-            model: model,
+            model: model.id,
             messages:,
             stream: stream,
-            max_tokens: RubyLLM.models.find(model)&.max_tokens || 4096
+            max_tokens: model.max_tokens || 4096
           }
         end
 

data/lib/ruby_llm/providers/anthropic/media.rb

@@ -8,7 +8,6 @@ module RubyLLM
         module_function
 
         def format_content(content, cache: false)
-          # Convert Hash/Array back to JSON string for API
           return [format_text(content.to_json, cache:)] if content.is_a?(Hash) || content.is_a?(Array)
           return [format_text(content, cache:)] unless content.is_a?(Content)
 
@@ -99,7 +98,7 @@ module RubyLLM
           with_cache_control(
             {
               type: 'text',
-              text: Utils.format_text_file_for_llm(text_file)
+              text: text_file.for_llm
             },
             cache:
           )

data/lib/ruby_llm/providers/anthropic/tools.rb

@@ -46,7 +46,7 @@ module RubyLLM
           {
             type: 'tool_result',
             tool_use_id: msg.tool_call_id,
-            content: msg.content
+            content: Media.format_content(msg.content)
           }
         end
 
@@ -73,7 +73,6 @@ module RubyLLM
         def parse_tool_calls(content_blocks)
           return nil if content_blocks.nil?
 
-          # Handle single content block (backward compatibility)
           content_blocks = [content_blocks] unless content_blocks.is_a?(Array)
 
           tool_calls = {}

data/lib/ruby_llm/providers/anthropic.rb

@@ -2,8 +2,7 @@
 
 module RubyLLM
   module Providers
-    # Anthropic Claude API integration. Handles the complexities of
-    # Claude's unique message format and tool calling conventions.
+    # Anthropic Claude API integration.
     class Anthropic < Provider
       include Anthropic::Chat
       include Anthropic::Embeddings

data/lib/ruby_llm/providers/bedrock/chat.rb

@@ -11,7 +11,6 @@ module RubyLLM
           signature = sign_request("#{connection.connection.url_prefix}#{completion_url}", payload:)
           response = connection.post completion_url, payload do |req|
             req.headers.merge! build_headers(signature.headers, streaming: block_given?)
-            # Merge additional headers, with existing headers taking precedence
             req.headers = additional_headers.merge(req.headers) unless additional_headers.empty?
           end
           Anthropic::Chat.parse_completion_response response
@@ -42,8 +41,7 @@ module RubyLLM
 
         def render_payload(messages, tools:, temperature:, model:, stream: false, schema: nil, # rubocop:disable Lint/UnusedMethodArgument,Metrics/ParameterLists
                            cache_prompts: { system: false, user: false, tools: false })
-          # Hold model_id in instance variable for use in completion_url and stream_url
-          @model_id = model
+          @model_id = model.id
 
           system_messages, chat_messages = Anthropic::Chat.separate_messages(messages)
           system_content = Anthropic::Chat.build_system_content(system_messages, cache: cache_prompts[:system])
@@ -67,7 +65,7 @@ module RubyLLM
           {
             anthropic_version: 'bedrock-2023-05-31',
             messages: messages,
-            max_tokens: RubyLLM.models.find(model)&.max_tokens || 4096
+            max_tokens: model.max_tokens || 4096
           }
         end
       end

data/lib/ruby_llm/providers/bedrock/media.rb

@@ -11,7 +11,6 @@ module RubyLLM
         module_function
 
         def format_content(content, cache: false)
-          # Convert Hash/Array back to JSON string for API
           return [Anthropic::Media.format_text(content.to_json, cache:)] if content.is_a?(Hash) || content.is_a?(Array)
           return [Anthropic::Media.format_text(content, cache:)] unless content.is_a?(Content)
 

data/lib/ruby_llm/providers/bedrock/models.rb

@@ -25,7 +25,6 @@ module RubyLLM
         def parse_list_models_response(response, slug, capabilities)
           models = Array(response.body['modelSummaries'])
 
-          # Filter to include only models we care about
           models.select { |m| m['modelId'].include?('claude') }.map do |model_data|
             model_id = model_data['modelId']
 
@@ -51,7 +50,6 @@ module RubyLLM
           end
         end
 
-        # Simple test-friendly method that only sets the ID
         def create_model_info(model_data, slug, _capabilities)
           model_id = model_data['modelId']
 
@@ -74,7 +72,25 @@ module RubyLLM
           return model_id unless model_data['inferenceTypesSupported']&.include?('INFERENCE_PROFILE')
           return model_id if model_data['inferenceTypesSupported']&.include?('ON_DEMAND')
 
-          "us.#{model_id}"
+          desired_region_prefix = inference_profile_region_prefix
+
+          # Return unchanged if model already has the correct region prefix
+          return model_id if model_id.start_with?("#{desired_region_prefix}.")
+
+          # Remove any existing region prefix (e.g., "us.", "eu.", "ap.")
+          clean_model_id = model_id.sub(/^[a-z]{2}\./, '')
+
+          # Apply the desired region prefix
+          "#{desired_region_prefix}.#{clean_model_id}"
+        end
+
+        def inference_profile_region_prefix
+          # Extract region prefix from bedrock_region (e.g., "eu-west-3" -> "eu")
+          region = @config.bedrock_region.to_s
+          return 'us' if region.empty? # Default fallback
+
+          # Take first two characters as the region prefix
+          region[0, 2]
         end
       end
     end

data/lib/ruby_llm/providers/bedrock/streaming/base.rb

@@ -5,18 +5,6 @@ module RubyLLM
     class Bedrock
       module Streaming
         # Base module for AWS Bedrock streaming functionality.
-        # Serves as the core module that includes all other streaming-related modules
-        # and provides fundamental streaming operations.
-        #
-        # Responsibilities:
-        # - Stream URL management
-        # - Stream handling and error processing
-        # - Coordinating the functionality of other streaming modules
-        #
-        # @example
-        #   module MyStreamingImplementation
-        #     include RubyLLM::Providers::Bedrock::Streaming::Base
-        #   end
         module Base
           def self.included(base)
             base.include ContentExtraction

data/lib/ruby_llm/providers/bedrock/streaming/content_extraction.rb

@@ -5,13 +5,6 @@ module RubyLLM
     class Bedrock
       module Streaming
         # Module for handling content extraction from AWS Bedrock streaming responses.
-        # Provides methods to extract and process various types of content from the response data.
-        #
-        # Responsibilities:
-        # - Extracting content from different response formats
-        # - Processing JSON deltas and content blocks
-        # - Extracting metadata (tokens, model IDs, tool calls)
-        # - Handling different content structures (arrays, blocks, completions)
         module ContentExtraction
           def json_delta?(data)
             data['type'] == 'content_block_delta' && data.dig('delta', 'type') == 'input_json_delta'

data/lib/ruby_llm/providers/bedrock/streaming/message_processing.rb

@@ -5,18 +5,6 @@ module RubyLLM
     class Bedrock
       module Streaming
         # Module for processing streaming messages from AWS Bedrock.
-        # Handles the core message processing logic, including validation and chunking.
-        #
-        # Responsibilities:
-        # - Processing incoming message chunks
-        # - Validating message structure and content
-        # - Managing message offsets and boundaries
-        # - Error handling during message processing
-        #
-        # @example Processing a message chunk
-        #   offset = process_message(chunk, current_offset) do |processed_chunk|
-        #     handle_processed_chunk(processed_chunk)
-        #   end
         module MessageProcessing
           def process_chunk(chunk, &)
             offset = 0

data/lib/ruby_llm/providers/bedrock/streaming/payload_processing.rb

@@ -7,18 +7,6 @@ module RubyLLM
     class Bedrock
       module Streaming
         # Module for processing payloads from AWS Bedrock streaming responses.
-        # Handles JSON payload extraction, decoding, and chunk creation.
-        #
-        # Responsibilities:
-        # - Extracting and validating JSON payloads
-        # - Decoding Base64-encoded response data
-        # - Creating response chunks from processed data
-        # - Error handling for JSON parsing and processing
-        #
-        # @example Processing a payload
-        #   process_payload(raw_payload) do |chunk|
-        #     yield_chunk_to_client(chunk)
-        #   end
         module PayloadProcessing
           def process_payload(payload, &)
             json_payload = extract_json_payload(payload)

data/lib/ruby_llm/providers/bedrock/streaming/prelude_handling.rb

@@ -5,19 +5,6 @@ module RubyLLM
     class Bedrock
       module Streaming
         # Module for handling message preludes in AWS Bedrock streaming responses.
-        # Manages the parsing and validation of message headers and prelude data.
-        #
-        # Responsibilities:
-        # - Reading and validating message preludes
-        # - Calculating message positions and boundaries
-        # - Finding and validating prelude positions in chunks
-        # - Ensuring message integrity through length validation
-        #
-        # @example Reading a prelude
-        #   if can_read_prelude?(chunk, offset)
-        #     total_length, headers_length = read_prelude(chunk, offset)
-        #     process_message_with_lengths(total_length, headers_length)
-        #   end
         module PreludeHandling
           def can_read_prelude?(chunk, offset)
             chunk.bytesize - offset >= 12

data/lib/ruby_llm/providers/bedrock/streaming.rb

@@ -10,24 +10,6 @@ module RubyLLM
   module Providers
     class Bedrock
       # Streaming implementation for the AWS Bedrock API.
-      # This module provides functionality for handling streaming responses from AWS Bedrock,
-      # including message processing, content extraction, and error handling.
-      #
-      # The implementation is split into several focused modules:
-      # - Base: Core streaming functionality and module coordination
-      # - ContentExtraction: Extracting content from response data
-      # - MessageProcessing: Processing streaming message chunks
-      # - PayloadProcessing: Handling JSON payloads and chunk creation
-      # - PreludeHandling: Managing message preludes and headers
-      #
-      # @example Using the streaming module
-      #   class BedrockClient
-      #     include RubyLLM::Providers::Bedrock::Streaming
-      #
-      #     def stream_response(&block)
-      #       handle_stream(&block)
-      #     end
-      #   end
       module Streaming
         include Base
       end

data/lib/ruby_llm/providers/bedrock.rb

@@ -5,8 +5,7 @@ require 'time'
 
 module RubyLLM
   module Providers
-    # AWS Bedrock API integration. Handles chat completion and streaming
-    # for Claude models.
+    # AWS Bedrock API integration.
     class Bedrock < Provider
       include Bedrock::Chat
       include Bedrock::Streaming

data/lib/ruby_llm/providers/deepseek/capabilities.rb

@@ -10,7 +10,7 @@ module RubyLLM
         def context_window_for(model_id)
           case model_id
           when /deepseek-(?:chat|reasoner)/ then 64_000
-          else 32_768 # Sensible default
+          else 32_768
           end
         end
 
@@ -67,7 +67,6 @@ module RubyLLM
           end
         end
 
-        # Pricing information for DeepSeek models (USD per 1M tokens)
         PRICES = {
           chat: {
             input_hit: 0.07,

data/lib/ruby_llm/providers/deepseek/chat.rb

@@ -8,7 +8,6 @@ module RubyLLM
         module_function
 
         def format_role(role)
-          # DeepSeek doesn't use the new OpenAI convention for system prompts
           role.to_s
         end
       end