ruby_llm 1.1.0rc1 → 1.1.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 7f1ca42c438f99a40f6ada6ab87e23377403d9d6c8db6049d54960308b6dd42f
-  data.tar.gz: 0467734c40a0a64505b8f4de2eea4b929e2b0c615fedb3fe88d75f8cf20040de
+  metadata.gz: b157007e50a6d43f11591847306e9e950904ad71dde849dded2f4364f376fa0f
+  data.tar.gz: 9f2024f254134590b971a98b6fe1cae161bfd25378f747cd6c192237add0d80c
 SHA512:
-  metadata.gz: ed886fb15081ce27309c0e583223e62adf171907f0f047344763e9eae5ede4dbb96657bb1b69dde9e68bc1ed56acedb93717273067236e40104dddcc09886f33
-  data.tar.gz: 7f2aa9c50cfd1a525273256ff014712944c976229f6b90fe338decaf2525bcb82f19dec187d14795ce0891a7fceadb5c048905c8afa8066b0ddbe37d54c48edd
+  metadata.gz: d804afb295a9b9d174f44ac110ce60e9d6b3ab6a4e96d003532bbdf5c95bc6db34b7affcec44d6025eff98d71a38a356417dee3912776b44feb735b803c49ed6
+  data.tar.gz: ef562bad49590fc86fb78311f861688175a55b9ab2b6d68c23d74cbaeabc1ecf9703b420bed5f8bc689fb1542cbe2fe484b1aa52e8df53d8b6fedb6737b19f18

data/README.md

@@ -135,7 +135,7 @@ chat.ask "Tell me a story about a Ruby programmer" do |chunk|
   print chunk.content
 end
 
-# Set personality or behavior with instructions (aka system prompts) - available from 1.1.0
+# Set personality or behavior with instructions (aka system prompts)
 chat.with_instructions "You are a friendly Ruby expert who loves to help beginners"
 
 # Understand content in multiple forms
@@ -171,7 +171,7 @@ end
 # In a background job
 chat = Chat.create! model_id: "gpt-4o-mini"
 
-# Set personality or behavior with instructions (aka system prompts) - they're persisted too! - available from 1.1.0
+# Set personality or behavior with instructions (aka system prompts) - they're persisted too!
 chat.with_instructions "You are a friendly Ruby expert who loves to help beginners"
 
 chat.ask("What's your favorite Ruby gem?") do |chunk|

data/lib/ruby_llm/chat.rb

@@ -32,7 +32,9 @@ module RubyLLM
 
     alias say ask
 
-    def with_instructions(instructions)
+    def with_instructions(instructions, replace: false)
+      @messages = @messages.reject! { |msg| msg.role == :system } if replace
+
       add_message role: :system, content: instructions
       self
     end

data/lib/ruby_llm/configuration.rb

@@ -10,6 +10,7 @@ module RubyLLM
   #     config.anthropic_api_key = ENV['ANTHROPIC_API_KEY']
   #   end
   class Configuration
+    # Provider-specific configuration
     attr_accessor :openai_api_key,
                   :anthropic_api_key,
                   :gemini_api_key,
@@ -18,15 +19,26 @@ module RubyLLM
                   :bedrock_secret_key,
                   :bedrock_region,
                   :bedrock_session_token,
+                  # Default models
                   :default_model,
                   :default_embedding_model,
                   :default_image_model,
+                  # Connection configuration
                   :request_timeout,
-                  :max_retries
+                  :max_retries,
+                  :retry_interval,
+                  :retry_backoff_factor,
+                  :retry_interval_randomness
 
     def initialize
+      # Connection configuration
       @request_timeout = 120
       @max_retries = 3
+      @retry_interval = 0.1
+      @retry_backoff_factor = 2
+      @retry_interval_randomness = 0.5
+
+      # Default models
       @default_model = 'gpt-4o-mini'
       @default_embedding_model = 'text-embedding-3-small'
       @default_image_model = 'dall-e-3'

data/lib/ruby_llm/provider.rb

@@ -7,7 +7,7 @@ module RubyLLM
   module Provider
     # Common functionality for all LLM providers. Implements the core provider
     # interface so specific providers only need to implement a few key methods.
-    module Methods
+    module Methods # rubocop:disable Metrics/ModuleLength
       extend Streaming
 
       def complete(messages, tools:, temperature:, model:, &block) # rubocop:disable Metrics/MethodLength
@@ -108,9 +108,9 @@ module RubyLLM
 
           f.request :retry, {
             max: RubyLLM.config.max_retries,
-            interval: 0.05,
-            interval_randomness: 0.5,
-            backoff_factor: 2,
+            interval: RubyLLM.config.retry_interval,
+            interval_randomness: RubyLLM.config.retry_interval_randomness,
+            backoff_factor: RubyLLM.config.retry_backoff_factor,
             exceptions: [
               Errno::ETIMEDOUT,
               Timeout::Error,
@@ -119,9 +119,10 @@ module RubyLLM
               Faraday::RetriableResponse,
               RubyLLM::RateLimitError,
               RubyLLM::ServerError,
-              RubyLLM::ServiceUnavailableError
+              RubyLLM::ServiceUnavailableError,
+              RubyLLM::OverloadedError
             ],
-            retry_statuses: [429, 500, 502, 503, 504]
+            retry_statuses: [429, 500, 502, 503, 504, 529]
           }
 
           f.request :json

data/lib/ruby_llm/providers/bedrock.rb

@@ -33,6 +33,22 @@ module RubyLLM
         end
       end
 
+      def parse_error(response) # rubocop:disable Metrics/MethodLength
+        return if response.body.empty?
+
+        body = try_parse_json(response.body)
+        case body
+        when Hash
+          body['message']
+        when Array
+          body.map do |part|
+            part['message']
+          end.join('. ')
+        else
+          body
+        end
+      end
+
       def sign_request(url, method: :post, payload: nil)
         signer = create_signer
         request = build_request(url, method:, payload:)

data/lib/ruby_llm/stream_accumulator.rb

@@ -42,12 +42,20 @@ module RubyLLM
 
     private
 
-    def tool_calls_from_stream
+    def tool_calls_from_stream # rubocop:disable Metrics/MethodLength
       tool_calls.transform_values do |tc|
+        arguments = if tc.arguments.is_a?(String) && !tc.arguments.empty?
+                      JSON.parse(tc.arguments)
+                    elsif tc.arguments.is_a?(String)
+                      {} # Return empty hash for empty string arguments
+                    else
+                      tc.arguments
+                    end
+
         ToolCall.new(
           id: tc.id,
           name: tc.name,
-          arguments: tc.arguments.is_a?(String) ? JSON.parse(tc.arguments) : tc.arguments
+          arguments: arguments
         )
       end
     end

data/lib/ruby_llm/tool.rb

@@ -72,9 +72,6 @@ module RubyLLM
       result = execute(**args.transform_keys(&:to_sym))
       RubyLLM.logger.debug "Tool #{name} returned: #{result.inspect}"
       result
-    rescue StandardError => e
-      RubyLLM.logger.error "Tool #{name} failed with error: #{e.message}"
-      { error: e.message }
     end
 
     def execute(...)

data/lib/ruby_llm/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module RubyLLM
-  VERSION = '1.1.0rc1'
+  VERSION = '1.1.0'
 end

data/lib/tasks/models_docs.rake

@@ -0,0 +1,156 @@
+# frozen_string_literal: true
+
+require 'ruby_llm'
+require 'fileutils'
+
+MODEL_KEYS_TO_DISPLAY = %i[
+  id
+  type
+  display_name
+  provider
+  context_window
+  max_tokens
+  family
+  input_price_per_million
+  output_price_per_million
+].freeze
+
+def to_markdown_table(models) # rubocop:disable Metrics/AbcSize,Metrics/MethodLength
+  to_display_hash = ->(model) { model.to_h.slice(*MODEL_KEYS_TO_DISPLAY) }
+  model_hashes = Array(models).map { |model| to_display_hash.call(model) }
+
+  # Create abbreviated headers
+  headers = {
+    id: 'ID',
+    type: 'Type',
+    display_name: 'Name',
+    provider: 'Provider',
+    context_window: 'Context',
+    max_tokens: 'MaxTok',
+    family: 'Family',
+    input_price_per_million: 'In$/M',
+    output_price_per_million: 'Out$/M'
+  }
+
+  # Create header row with alignment markers
+  # Right-align numbers, left-align text
+  alignments = {
+    id: ':--',
+    type: ':--',
+    display_name: ':--',
+    provider: ':--',
+    context_window: '--:',
+    max_tokens: '--:',
+    family: ':--',
+    input_price_per_million: '--:',
+    output_price_per_million: '--:'
+  }
+
+  # Build the table
+  lines = []
+
+  # Header row
+  lines << "| #{MODEL_KEYS_TO_DISPLAY.map { |key| headers[key] }.join(' | ')} |"
+
+  # Alignment row
+  lines << "| #{MODEL_KEYS_TO_DISPLAY.map { |key| alignments[key] }.join(' | ')} |"
+
+  # Data rows
+  model_hashes.each do |model_hash|
+    values = MODEL_KEYS_TO_DISPLAY.map do |key|
+      if model_hash[key].is_a?(Float)
+        format('%.2f', model_hash[key])
+      else
+        model_hash[key]
+      end
+    end
+
+    lines << "| #{values.join(' | ')} |"
+  end
+
+  lines.join("\n")
+end
+
+namespace :models do # rubocop:disable Metrics/BlockLength
+  desc 'Generate available models documentation'
+  task :docs do # rubocop:disable Metrics/BlockLength
+    FileUtils.mkdir_p('docs/guides') # ensure output directory exists
+
+    output = <<~MARKDOWN
+      ---
+      layout: default
+      title: Available Models
+      parent: Guides
+      nav_order: 10
+      permalink: /guides/available-models
+      ---
+
+      # Available Models
+
+      This guide lists all models available in RubyLLM, automatically generated from the current model registry.
+
+      _Last updated: #{Time.now.utc.strftime('%Y-%m-%d')}_
+
+      ## Contributing
+
+      The model list is automatically generated from the model registry. To add or update models:
+
+      1. Edit the appropriate `capabilities.rb` file in `lib/ruby_llm/providers/<provider>/`
+      2. Run `rake models:update` to refresh the model registry
+      3. Submit a pull request with the updated `models.json`
+
+      See [Contributing Guide](/CONTRIBUTING.md) for more details.
+
+      ## Additional Model Information
+
+      The tables below show basic model information including context windows, token limits, and pricing. Models also have additional capabilities not shown in the tables:
+
+      - **Vision Support**: Whether the model can process images
+      - **Function Calling**: Whether the model supports function calling
+      - **JSON Mode**: Whether the model can be constrained to output valid JSON
+      - **Structured Output**: Whether the model supports structured output formats
+
+      For complete model information, you can check the `models.json` file in the RubyLLM source code.
+
+      For more information about working with models, see the [Working with Models](/guides/models) guide.
+
+      ## Models by Type
+
+      ### Chat Models (#{RubyLLM.models.chat_models.count})
+
+      #{to_markdown_table(RubyLLM.models.chat_models)}
+
+      ### Image Models (#{RubyLLM.models.image_models.count})
+
+      #{to_markdown_table(RubyLLM.models.image_models)}
+
+      ### Audio Models (#{RubyLLM.models.audio_models.count})
+
+      #{to_markdown_table(RubyLLM.models.audio_models)}
+
+      ### Embedding Models (#{RubyLLM.models.embedding_models.count})
+
+      #{to_markdown_table(RubyLLM.models.embedding_models)}
+
+      ### Moderation Models (#{RubyLLM.models.select { |m| m.type == 'moderation' }.count})
+
+      #{to_markdown_table(RubyLLM.models.select { |m| m.type == 'moderation' })}
+
+      ## Models by Provider
+
+      #{RubyLLM::Provider.providers.keys.map do |provider|
+        models = RubyLLM.models.by_provider(provider)
+        next if models.none?
+
+        <<~PROVIDER
+          ### #{provider.to_s.capitalize} Models (#{models.count})
+
+            #{to_markdown_table(models)}
+        PROVIDER
+      end.compact.join("\n")}
+    MARKDOWN
+
+    File.write('docs/guides/available-models.md', output)
+    puts 'Generated docs/guides/available-models.md'
+  end
+end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: ruby_llm
 version: !ruby/object:Gem::Version
-  version: 1.1.0rc1
+  version: 1.1.0
 platform: ruby
 authors:
 - Carmine Paolino
 autorequire:
 bindir: bin
 cert_chain: []
-date: 2025-04-02 00:00:00.000000000 Z
+date: 2025-04-08 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: base64
@@ -176,6 +176,7 @@ files:
 - lib/tasks/code_validator.rb
 - lib/tasks/model_updater.rb
 - lib/tasks/models.rake
+- lib/tasks/models_docs.rake
 - lib/tasks/vcr.rake
 homepage: https://rubyllm.com
 licenses: