ruby_llm 1.2.0 → 1.3.0rc1

data/lib/tasks/models.rake

@@ -1,42 +1,78 @@
 # frozen_string_literal: true
 
-require_relative 'model_updater'
-require_relative 'capability_scraper'
-require_relative 'capability_generator'
+require 'dotenv/load'
+require 'ruby_llm'
 
-namespace :models do # rubocop:disable Metrics/BlockLength
+def configure_from_env
+  RubyLLM.configure do |config|
+    config.openai_api_key = ENV.fetch('OPENAI_API_KEY', nil)
+    config.anthropic_api_key = ENV.fetch('ANTHROPIC_API_KEY', nil)
+    config.gemini_api_key = ENV.fetch('GEMINI_API_KEY', nil)
+    config.deepseek_api_key = ENV.fetch('DEEPSEEK_API_KEY', nil)
+    config.openrouter_api_key = ENV.fetch('OPENROUTER_API_KEY', nil)
+    configure_bedrock(config)
+    config.request_timeout = 30
+  end
+end
+
+def configure_bedrock(config)
+  config.bedrock_api_key = ENV.fetch('AWS_ACCESS_KEY_ID', nil)
+  config.bedrock_secret_key = ENV.fetch('AWS_SECRET_ACCESS_KEY', nil)
+  config.bedrock_region = ENV.fetch('AWS_REGION', nil)
+  config.bedrock_session_token = ENV.fetch('AWS_SESSION_TOKEN', nil)
+end
+
+def refresh_models
+  initial_count = RubyLLM.models.all.size
+  puts "Refreshing models (#{initial_count} cached)..."
+
+  models = RubyLLM.models.refresh!
+
+  if models.all.empty? && initial_count.zero?
+    puts 'Error: Failed to fetch models.'
+    exit(1)
+  elsif models.all.size == initial_count && initial_count.positive?
+    puts 'Warning: Model list unchanged.'
+  else
+    puts "Saving models.json (#{models.all.size} models)"
+    models.save_models
+  end
+
+  @models = models
+end
+
+def display_model_stats
+  puts "\nModel count:"
+  provider_counts = @models.all.group_by(&:provider).transform_values(&:count)
+
+  RubyLLM::Provider.providers.each_key do |sym|
+    name = sym.to_s.capitalize
+    count = provider_counts[sym.to_s] || 0
+    status = status(sym)
+    puts "  #{name}: #{count} models #{status}"
+  end
+
+  puts 'Refresh complete.'
+end
+
+def status(provider_sym)
+  if RubyLLM::Provider.providers[provider_sym].local?
+    ' (LOCAL - SKIP)'
+  elsif RubyLLM::Provider.providers[provider_sym].configured?
+    ' (OK)'
+  else
+    ' (NOT CONFIGURED)'
+  end
+end
+
+namespace :models do
   desc 'Update available models from providers (API keys needed)'
   task :update do
-    ModelUpdater.new.run
-  end
+    puts 'Configuring RubyLLM...'
+    configure_from_env
 
-  desc 'Update capabilities modules (GEMINI_API_KEY needed)'
-  task :update_capabilities, [:providers] do |_t, args|
-    gemini_key = ENV.fetch('GEMINI_API_KEY', nil)
-    unless gemini_key && !gemini_key.empty?
-      puts 'Error: GEMINI_API_KEY required'
-      exit(1)
-    end
-
-    RubyLLM.configure do |c|
-      c.gemini_api_key = gemini_key
-      c.request_timeout = 300
-    end
-
-    target_providers = CapabilityScraper.parse_providers(args[:providers])
-
-    begin
-      scraper = CapabilityScraper.new(target_providers)
-      scraper.run do |provider, docs_html|
-        generator = CapabilityGenerator.new(provider, docs_html)
-        generator.generate_and_save
-      end
-    rescue StandardError => e
-      puts "Error: #{e.message}"
-      puts e.backtrace.first(5).join("\n")
-    ensure
-      puts 'Update process complete. Review generated files.'
-    end
+    refresh_models
+    display_model_stats
   end
 end
 

data/lib/tasks/models_docs.rake

@@ -1,168 +1,211 @@
 # frozen_string_literal: true
 
-require 'ruby_llm'
+require 'dotenv/load'
 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'
-  }
+namespace :models do
+  desc 'Generate available models documentation'
+  task :docs do
+    FileUtils.mkdir_p('docs/guides') # ensure output directory exists
 
-  # 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: '--:'
-  }
+    # Generate markdown content
+    output = generate_models_markdown
 
-  # Build the table
-  lines = []
+    # Write the output
+    File.write('docs/guides/available-models.md', output)
+    puts 'Generated docs/guides/available-models.md'
+  end
+end
 
-  # Header row
-  lines << "| #{MODEL_KEYS_TO_DISPLAY.map { |key| headers[key] }.join(' | ')} |"
+def generate_models_markdown
+  <<~MARKDOWN
+    ---
+    layout: default
+    title: Available Models
+    parent: Guides
+    nav_order: 10
+    permalink: /guides/available-models
+    ---
 
-  # Alignment row
-  lines << "| #{MODEL_KEYS_TO_DISPLAY.map { |key| alignments[key] }.join(' | ')} |"
+    # Available Models
+    {: .no_toc }
 
-  # 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
+    This guide lists all models available in RubyLLM, automatically generated from the current model registry.
+    {: .fs-6 .fw-300 }
 
-    lines << "| #{values.join(' | ')} |"
-  end
+    ## Table of contents
+    {: .no_toc .text-delta }
+
+    1. TOC
+    {:toc}
+
+    ---
+
+    ## 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.
+
+    ## Last Updated
+    {: .d-inline-block }
+
+    #{Time.now.utc.strftime('%Y-%m-%d')}
+    {: .label .label-green }
+
+    ## Models by Provider
+
+    #{generate_provider_sections}
+
+    ## Models by Capability
 
-  lines.join("\n")
+    #{generate_capability_sections}
+
+    ## Models by Modality
+
+    #{generate_modality_sections}
+  MARKDOWN
 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
+def generate_provider_sections
+  RubyLLM::Provider.providers.keys.map do |provider|
+    models = RubyLLM.models.by_provider(provider)
+    next if models.none?
+
+    <<~PROVIDER
+      ### #{provider.to_s.capitalize} (#{models.count})
 
-    output = <<~MARKDOWN
-      ---
-      layout: default
-      title: Available Models
-      parent: Guides
-      nav_order: 10
-      permalink: /guides/available-models
-      ---
+      #{models_table(models)}
+    PROVIDER
+  end.compact.join("\n\n")
+end
 
-      # Available Models
-      {: .no_toc }
+def generate_capability_sections
+  capabilities = {
+    'Function Calling' => RubyLLM.models.select(&:function_calling?),
+    'Structured Output' => RubyLLM.models.select(&:structured_output?),
+    'Streaming' => RubyLLM.models.select { |m| m.capabilities.include?('streaming') },
+    # 'Reasoning' => RubyLLM.models.select { |m| m.capabilities.include?('reasoning') },
+    'Batch Processing' => RubyLLM.models.select { |m| m.capabilities.include?('batch') }
+  }
 
-      This guide lists all models available in RubyLLM, automatically generated from the current model registry.
-      {: .fs-6 .fw-300 }
+  capabilities.map do |capability, models|
+    next if models.none?
 
-      ## Table of contents
-      {: .no_toc .text-delta }
+    <<~CAPABILITY
+      ### #{capability} (#{models.count})
 
-      1. TOC
-      {:toc}
+      #{models_table(models)}
+    CAPABILITY
+  end.compact.join("\n\n")
+end
 
-      ---
+def generate_modality_sections # rubocop:disable Metrics/PerceivedComplexity
+  sections = []
 
-      ## Contributing
+  # Models that support vision/images
+  vision_models = RubyLLM.models.select { |m| (m.modalities.input || []).include?('image') }
+  if vision_models.any?
+    sections << <<~SECTION
+      ### Vision Models (#{vision_models.count})
 
-      The model list is automatically generated from the model registry. To add or update models:
+      Models that can process images:
 
-      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`
+      #{models_table(vision_models)}
+    SECTION
+  end
 
-      See [Contributing Guide](/CONTRIBUTING.md) for more details.
+  # Models that support audio
+  audio_models = RubyLLM.models.select { |m| (m.modalities.input || []).include?('audio') }
+  if audio_models.any?
+    sections << <<~SECTION
+      ### Audio Input Models (#{audio_models.count})
 
-      ## Additional Model Information
+      Models that can process audio:
 
-      The tables below show basic model information including context windows, token limits, and pricing. Models also have additional capabilities not shown in the tables:
+      #{models_table(audio_models)}
+    SECTION
+  end
 
-      - **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
+  # Models that support PDFs
+  pdf_models = RubyLLM.models.select { |m| (m.modalities.input || []).include?('pdf') }
+  if pdf_models.any?
+    sections << <<~SECTION
+      ### PDF Models (#{pdf_models.count})
 
-      For complete model information, you can check the `models.json` file in the RubyLLM source code.
+      Models that can process PDF documents:
 
-      For more information about working with models, see the [Working with Models](/guides/models) guide.
+      #{models_table(pdf_models)}
+    SECTION
+  end
 
-      ## Models by Type
-      {: .d-inline-block }
+  # Models for embeddings
+  embedding_models = RubyLLM.models.select { |m| (m.modalities.output || []).include?('embeddings') }
+  if embedding_models.any?
+    sections << <<~SECTION
+      ### Embedding Models (#{embedding_models.count})
 
-      Last updated: #{Time.now.utc.strftime('%Y-%m-%d')}
-      {: .label .label-green }
+      Models that generate embeddings:
 
-      ### Chat Models (#{RubyLLM.models.chat_models.count})
+      #{models_table(embedding_models)}
+    SECTION
+  end
 
-      #{to_markdown_table(RubyLLM.models.chat_models)}
+  sections.join("\n\n")
+end
 
-      ### Image Models (#{RubyLLM.models.image_models.count})
+def models_table(models)
+  return '*No models found*' if models.none?
 
-      #{to_markdown_table(RubyLLM.models.image_models)}
+  headers = ['Model', 'ID', 'Provider', 'Context', 'Max Output', 'Standard Pricing (per 1M tokens)']
+  alignment = [':--', ':--', ':--', '--:', '--:', ':--']
 
-      ### Audio Models (#{RubyLLM.models.audio_models.count})
+  rows = models.sort_by { |m| [m.provider, m.name] }.map do |model|
+    # Format pricing information
+    pricing = standard_pricing_display(model)
 
-      #{to_markdown_table(RubyLLM.models.audio_models)}
+    [
+      model.name,
+      model.id,
+      model.provider,
+      model.context_window || '-',
+      model.max_output_tokens || '-',
+      pricing
+    ]
+  end
 
-      ### Embedding Models (#{RubyLLM.models.embedding_models.count})
+  table = []
+  table << "| #{headers.join(' | ')} |"
+  table << "| #{alignment.join(' | ')} |"
 
-      #{to_markdown_table(RubyLLM.models.embedding_models)}
+  rows.each do |row|
+    table << "| #{row.join(' | ')} |"
+  end
 
-      ### Moderation Models (#{RubyLLM.models.select { |m| m.type == 'moderation' }.count})
+  table.join("\n")
+end
 
-      #{to_markdown_table(RubyLLM.models.select { |m| m.type == 'moderation' })}
+def standard_pricing_display(model)
+  # Access pricing data using to_h to get the raw hash
+  pricing_data = model.pricing.to_h[:text_tokens]&.dig(:standard) || {}
 
-      ## Models by Provider
+  if pricing_data.any?
+    parts = []
 
-      #{RubyLLM::Provider.providers.keys.map do |provider|
-        models = RubyLLM.models.by_provider(provider)
-        next if models.none?
+    parts << "In: $#{format('%.2f', pricing_data[:input_per_million])}" if pricing_data[:input_per_million]
 
-        <<~PROVIDER
-          ### #{provider.to_s.capitalize} Models (#{models.count})
+    parts << "Out: $#{format('%.2f', pricing_data[:output_per_million])}" if pricing_data[:output_per_million]
 
-            #{to_markdown_table(models)}
-        PROVIDER
-      end.compact.join("\n")}
-    MARKDOWN
+    if pricing_data[:cached_input_per_million]
+      parts << "Cache: $#{format('%.2f', pricing_data[:cached_input_per_million])}"
+    end
 
-    File.write('docs/guides/available-models.md', output)
-    puts 'Generated docs/guides/available-models.md'
+    return parts.join(', ') if parts.any?
   end
+
+  '-'
 end

data/lib/tasks/vcr.rake

@@ -1,5 +1,7 @@
 # frozen_string_literal: true
 
+require 'dotenv/load'
+
 # Helper functions at the top level
 def record_all_cassettes(cassette_dir)
   # Re-record all cassettes
@@ -11,7 +13,7 @@ def record_all_cassettes(cassette_dir)
   puts 'Done recording. Please review the new cassettes.'
 end
 
-def record_for_providers(providers, cassette_dir) # rubocop:disable Metrics/AbcSize,Metrics/MethodLength
+def record_for_providers(providers, cassette_dir)
   # Get the list of available providers from RubyLLM itself
   all_providers = RubyLLM::Provider.providers.keys.map(&:to_s)
 
@@ -46,7 +48,7 @@ def record_for_providers(providers, cassette_dir) # rubocop:disable Metrics/AbcS
   puts 'Please review the updated cassettes for sensitive information.'
 end
 
-def find_matching_cassettes(dir, providers) # rubocop:disable Metrics/MethodLength
+def find_matching_cassettes(dir, providers)
   cassettes = []
 
   Dir.glob("#{dir}/**/*.yml").each do |file|

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: ruby_llm
 version: !ruby/object:Gem::Version
-  version: 1.2.0
+  version: 1.3.0rc1
 platform: ruby
 authors:
 - Carmine Paolino
 autorequire:
 bindir: bin
 cert_chain: []
-date: 2025-04-17 00:00:00.000000000 Z
+date: 2025-05-12 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: base64
@@ -108,11 +108,12 @@ dependencies:
     - - "~>"
       - !ruby/object:Gem::Version
         version: '2'
-description: A delightful Ruby way to work with AI. Chat in text, analyze and generate
-  images, understand audio, and use tools through a unified interface to OpenAI, Anthropic,
-  Google, AWS Bedrock Anthropic, and DeepSeek. Built for developer happiness with
-  automatic token counting, proper streaming, and Rails integration. No wrapping your
-  head around multiple APIs - just clean Ruby code that works.
+description: A delightful Ruby way to work with AI. Tired of juggling different SDKs?
+  RubyLLM provides one beautiful, Ruby-like interface for OpenAI, Anthropic, Gemini,
+  Bedrock, OpenRouter, DeepSeek, Ollama, and any OpenAI-compatible API. Chat (with
+  text, images, audio, PDFs), generate images, create embeddings, use tools (function
+  calling), stream responses, and integrate with Rails effortlessly. Minimal dependencies,
+  maximum developer happiness - just clean Ruby code that works.
 email:
 - carmine@paolino.me
 executables: []
@@ -125,14 +126,21 @@ files:
 - lib/ruby_llm/active_record/acts_as.rb
 - lib/ruby_llm/aliases.json
 - lib/ruby_llm/aliases.rb
+- lib/ruby_llm/attachments.rb
+- lib/ruby_llm/attachments/audio.rb
+- lib/ruby_llm/attachments/image.rb
+- lib/ruby_llm/attachments/pdf.rb
 - lib/ruby_llm/chat.rb
 - lib/ruby_llm/chunk.rb
 - lib/ruby_llm/configuration.rb
+- lib/ruby_llm/connection.rb
 - lib/ruby_llm/content.rb
+- lib/ruby_llm/context.rb
 - lib/ruby_llm/embedding.rb
 - lib/ruby_llm/error.rb
 - lib/ruby_llm/image.rb
 - lib/ruby_llm/message.rb
+- lib/ruby_llm/mime_types.rb
 - lib/ruby_llm/model_info.rb
 - lib/ruby_llm/models.json
 - lib/ruby_llm/models.rb
@@ -148,6 +156,7 @@ files:
 - lib/ruby_llm/providers/bedrock.rb
 - lib/ruby_llm/providers/bedrock/capabilities.rb
 - lib/ruby_llm/providers/bedrock/chat.rb
+- lib/ruby_llm/providers/bedrock/media.rb
 - lib/ruby_llm/providers/bedrock/models.rb
 - lib/ruby_llm/providers/bedrock/signing.rb
 - lib/ruby_llm/providers/bedrock/streaming.rb
@@ -168,6 +177,9 @@ files:
 - lib/ruby_llm/providers/gemini/models.rb
 - lib/ruby_llm/providers/gemini/streaming.rb
 - lib/ruby_llm/providers/gemini/tools.rb
+- lib/ruby_llm/providers/ollama.rb
+- lib/ruby_llm/providers/ollama/chat.rb
+- lib/ruby_llm/providers/ollama/media.rb
 - lib/ruby_llm/providers/openai.rb
 - lib/ruby_llm/providers/openai/capabilities.rb
 - lib/ruby_llm/providers/openai/chat.rb
@@ -177,18 +189,15 @@ files:
 - lib/ruby_llm/providers/openai/models.rb
 - lib/ruby_llm/providers/openai/streaming.rb
 - lib/ruby_llm/providers/openai/tools.rb
+- lib/ruby_llm/providers/openrouter.rb
+- lib/ruby_llm/providers/openrouter/models.rb
 - lib/ruby_llm/railtie.rb
 - lib/ruby_llm/stream_accumulator.rb
 - lib/ruby_llm/streaming.rb
 - lib/ruby_llm/tool.rb
 - lib/ruby_llm/tool_call.rb
+- lib/ruby_llm/utils.rb
 - lib/ruby_llm/version.rb
-- lib/tasks/browser_helper.rb
-- lib/tasks/capability_generator.rb
-- lib/tasks/capability_scraper.rb
-- lib/tasks/cli_helper.rb
-- lib/tasks/code_validator.rb
-- lib/tasks/model_updater.rb
 - lib/tasks/models.rake
 - lib/tasks/models_docs.rake
 - lib/tasks/vcr.rake
@@ -220,5 +229,5 @@ requirements: []
 rubygems_version: 3.5.22
 signing_key:
 specification_version: 4
-summary: Beautiful Ruby interface to modern AI
+summary: A single delightful Ruby way to work with AI.
 test_files: []

data/lib/tasks/browser_helper.rb

@@ -1,97 +0,0 @@
-# frozen_string_literal: true
-
-require 'ferrum'
-require_relative 'cli_helper'
-
-class BrowserHelper # rubocop:disable Style/Documentation
-  REALISTIC_USER_AGENT = 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/123.0.0.0 Safari/537.36' # rubocop:disable Layout/LineLength
-
-  def initialize
-    @browser = create_browser
-  end
-
-  def goto(url)
-    @browser.goto(url)
-  end
-
-  def current_url
-    @browser.page.url
-  rescue StandardError
-    'N/A'
-  end
-
-  def get_page_content(context = 'current page') # rubocop:disable Metrics/MethodLength
-    puts "    Extracting HTML for #{context}..."
-
-    begin
-      sleep(1.0) # Small delay for page stability
-      html = @browser.body
-
-      if html && !html.empty?
-        puts "      Extracted ~#{html.length} chars of HTML"
-        puts '      WARNING: Challenge page detected' if html.match?(/challenge-platform|Checking site/)
-        html
-      else
-        puts '      Warning: Empty content returned'
-        ''
-      end
-    rescue StandardError => e
-      puts "      Error getting HTML: #{e.class} - #{e.message}"
-      ''
-    end
-  end
-
-  def wait_for_page_load
-    handle_cloudflare_challenge
-  end
-
-  def close
-    puts "\nClosing browser..."
-    @browser.quit
-  rescue StandardError => e
-    puts "  Warning: Error closing browser: #{e.message}"
-  end
-
-  private
-
-  def create_browser
-    puts '  Initializing browser for manual interaction...'
-
-    Ferrum::Browser.new(
-      window_size: [1366, 768],
-      headless: false,
-      browser_options: browser_options,
-      timeout: 120,
-      process_timeout: 120,
-      pending_connection_errors: false
-    )
-  end
-
-  def browser_options
-    {
-      'user-agent' => REALISTIC_USER_AGENT,
-      'disable-gpu' => nil,
-      'no-sandbox' => nil,
-      'disable-blink-features' => 'AutomationControlled',
-      'disable-infobars' => nil,
-      'start-maximized' => nil
-    }
-  end
-
-  def handle_cloudflare_challenge # rubocop:disable Metrics/MethodLength
-    puts "\nWaiting for Cloudflare challenge resolution..."
-    puts 'c: Challenge solved'
-    puts 'q: Quit/Skip'
-
-    choice = CliHelper.get_user_choice('Confirm when ready', %w[c q])
-    return false if choice == 'q'
-
-    begin
-      @browser.page.target_id
-      true
-    rescue StandardError
-      puts 'Browser check failed after challenge'
-      false
-    end
-  end
-end