ruby_llm 1.0.1 → 1.1.0rc2

data/lib/tasks/models.rake

@@ -1,208 +1,43 @@
 # frozen_string_literal: true
 
-require 'English'
-require 'faraday'
-require 'fileutils'
-require 'nokogiri'
-require 'ruby_llm'
-
-# URLs to process
-PROVIDER_DOCS = {
-  openai: {
-    models: 'https://platform.openai.com/docs/models',
-    pricing: 'https://platform.openai.com/docs/pricing'
-  },
-  gemini: {
-    models: 'https://ai.google.dev/gemini-api/docs/models/gemini',
-    pricing: 'https://ai.google.dev/gemini-api/docs/pricing'
-  },
-  deepseek: {
-    models: 'https://api-docs.deepseek.com/quick_start/pricing/'
-  },
-  anthropic: {
-    models: 'https://docs.anthropic.com/en/docs/about-claude/models/all-models'
-  }
-}.freeze
-
-def fetch_page(url) # rubocop:disable Metrics/AbcSize,Metrics/MethodLength
-  if url.include?('openai.com')
-    puts "Please visit #{url} and paste the content below (type 'END' on a new line when done):"
-    original_separator = $INPUT_RECORD_SEPARATOR
-    $/ = 'END'
-    content = $stdin.gets&.chomp
-    $/ = original_separator
-
-    raise "No content provided for #{url}" unless content
-
-    content
-  else
-    response = http_client.get(url)
-    html = Nokogiri::HTML(response.body)
-
-    # Remove script tags and comments
-    html.css('script').remove
-    html.xpath('//comment()').remove
-
-    # Extract text content
-    text = html.css('body').text
-
-    # Clean up whitespace
-    text.gsub!(/\s+/, ' ')
-    text.strip!
-
-    text
-  end
-rescue StandardError => e
-  raise "Failed to fetch #{url}: #{e.message}"
-end
-
-def http_client
-  @http_client ||= Faraday.new do |f|
-    f.response :raise_error
-    f.response :logger, RubyLLM.logger, { headers: false, bodies: true }
-  end
-end
+require_relative 'model_updater'
+require_relative 'capability_scraper'
+require_relative 'capability_generator'
 
 namespace :models do # rubocop:disable Metrics/BlockLength
-  desc 'Update available models from providers'
+  desc 'Update available models from providers (API keys needed)'
   task :update do
-    # Configure API keys
-    RubyLLM.configure do |config|
-      config.openai_api_key = ENV.fetch('OPENAI_API_KEY')
-      config.anthropic_api_key = ENV.fetch('ANTHROPIC_API_KEY')
-      config.gemini_api_key = ENV.fetch('GEMINI_API_KEY')
-      config.deepseek_api_key = ENV.fetch('DEEPSEEK_API_KEY')
-    end
-
-    models = RubyLLM.models.refresh!
-    models.save_models
-
-    puts "Updated models.json with #{models.all.size} models:"
-    RubyLLM::Provider.providers.each do |provider_sym, provider_module|
-      provider_name = provider_module.to_s.split('::').last
-      provider_models = models.all.select { |m| m.provider == provider_sym.to_s }
-      puts "#{provider_name} models: #{provider_models.size}"
-    end
+    ModelUpdater.new.run
   end
 
-  desc 'Update model capabilities modules by scraping provider documentation (use PROVIDER=name to update only one)'
-  task :update_capabilities do # rubocop:disable Metrics/BlockLength
-    # Check if a specific provider was requested
-    target_provider = ENV['PROVIDER']&.to_sym
-
-    # Configure API keys
-    RubyLLM.configure do |config|
-      config.openai_api_key = ENV.fetch('OPENAI_API_KEY')
-      config.anthropic_api_key = ENV.fetch('ANTHROPIC_API_KEY')
-      config.gemini_api_key = ENV.fetch('GEMINI_API_KEY')
+  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
 
-    # Filter providers if a specific one was requested
-    providers_to_process = if target_provider && PROVIDER_DOCS.key?(target_provider)
-                             { target_provider => PROVIDER_DOCS[target_provider] }
-                           else
-                             PROVIDER_DOCS
-                           end
-
-    # Process each provider
-    providers_to_process.each do |provider, urls| # rubocop:disable Metrics/BlockLength
-      puts "Processing #{provider}..."
-
-      # Initialize our AI assistants
-      #
-      gemini = RubyLLM.chat(model: 'gemini-2.0-flash').with_temperature(0)
-      claude = RubyLLM.chat(model: 'claude-3-7-sonnet-20250219').with_temperature(0)
-
-      # Read existing capabilities file if present
-      existing_file = "lib/ruby_llm/providers/#{provider}/capabilities.rb"
-      existing_code = File.read(existing_file) if File.exist?(existing_file)
-
-      begin
-        # Download documentation
-        docs = urls.map do |type, url|
-          puts "  Getting #{type} documentation..."
-          content = fetch_page(url)
-
-          puts "\nHere's what I got:\n\n"
-          puts "#{content.slice(0, 500)}...\n\n"
-
-          loop do
-            print 'Does this content look correct? (y/n): '
-            answer = $stdin.gets&.chomp&.downcase
-            break if answer == 'y'
-            raise "Content verification failed for #{url}" if answer == 'n'
-          end
-
-          "#{type.to_s.upcase} DOCUMENTATION:\n\n#{content}"
-        end.join("\n\n")
-
-        # Extract relevant information with Gemini
-        puts '  Extracting model information...'
-        extraction_prompt = <<~PROMPT
-          Extract relevant model capabilities information from this documentation:
-
-          #{docs}
-
-          Focus on:
-          1. Available models and their IDs
-          2. Context window sizes
-          3. Maximum output tokens
-          4. Pricing information
-          5. Model capabilities (vision, function calling, etc)
-
-          Format the information clearly and concisely, focusing only on facts about the models.
-        PROMPT
-
-        model_info = gemini.ask(extraction_prompt).content
-
-        # Generate Ruby code with Claude
-        puts '  Generating Ruby code...'
-        code_prompt = <<~PROMPT
-          I need you to generate a Ruby module for #{provider}'s model capabilities.
-          Use this extracted model information:
-
-          #{model_info}
-
-          The module should go in lib/ruby_llm/providers/#{provider}/capabilities.rb and follow these conventions:
-
-          1. Include methods for determining context windows, token limits, pricing, and capabilities
-          2. Use consistent naming with other providers
-          3. Include detailed pricing information in a PRICES constant
-          4. Follow the existing structure in the codebase
-          5. Use Ruby idioms and clean code practices
-          6. Include module_function to make methods callable at module level
-          7. Include all necessary method documentation
-
-          Here's the existing implementation for reference (maintain similar structure and same method names):
-
-          #{existing_code}
-
-          Focus on accuracy and maintaining consistency with the existing codebase structure.
-        PROMPT
-
-        response = claude.ask(code_prompt)
-
-        # Extract Ruby code from Claude's response
-        puts "  Extracting Ruby code from Claude's response..."
-        ruby_code = nil
-
-        # Look for Ruby code block
-        ruby_code = Regexp.last_match(1).strip if response.content =~ /```ruby\s*(.*?)```/m
-
-        # Verify we found Ruby code
-        raise "No Ruby code block found in Claude's response" if ruby_code.nil? || ruby_code.empty?
+    RubyLLM.configure do |c|
+      c.gemini_api_key = gemini_key
+      c.request_timeout = 300
+    end
 
-        # Save the file
-        file_path = "lib/ruby_llm/providers/#{provider}/capabilities.rb"
-        puts "  Writing #{file_path}..."
+    target_providers = CapabilityScraper.parse_providers(args[:providers])
 
-        FileUtils.mkdir_p(File.dirname(file_path))
-        File.write(file_path, ruby_code)
-      rescue StandardError => e
-        raise "Failed to process #{provider}: #{e.message}"
+    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
-
-    puts "Done! Don't forget to review the generated code and run the tests."
   end
 end
+
+task default: ['models:update']

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

data/lib/tasks/vcr.rake

@@ -29,13 +29,8 @@ def record_for_providers(providers, cassette_dir) # rubocop:disable Metrics/AbcS
     return
   end
 
-  # Get URL patterns from the providers themselves
-  provider_patterns = get_provider_patterns(providers)
-
-  puts "Finding cassettes for providers: #{providers.join(', ')}"
-
   # Find and delete matching cassettes
-  cassettes_to_delete = find_matching_cassettes(cassette_dir, provider_patterns)
+  cassettes_to_delete = find_matching_cassettes(cassette_dir, providers)
 
   if cassettes_to_delete.empty?
     puts 'No cassettes found for the specified providers.'
@@ -51,33 +46,21 @@ def record_for_providers(providers, cassette_dir) # rubocop:disable Metrics/AbcS
   puts 'Please review the updated cassettes for sensitive information.'
 end
 
-def get_provider_patterns(providers) # rubocop:disable Metrics/MethodLength
-  provider_patterns = {}
-
-  providers.each do |provider_name|
-    provider_module = RubyLLM::Provider.providers[provider_name.to_sym]
-    next unless provider_module
-
-    # Extract the base URL from the provider's api_base method
-    api_base = provider_module.api_base.to_s
-
-    # Create a regex pattern from the domain
-    next unless api_base && !api_base.empty?
-
-    domain = URI.parse(api_base).host
-    pattern = Regexp.new(Regexp.escape(domain))
-    provider_patterns[provider_name] = pattern
-  end
-
-  provider_patterns
-end
-
-def find_matching_cassettes(dir, patterns)
+def find_matching_cassettes(dir, providers) # rubocop:disable Metrics/MethodLength
   cassettes = []
 
   Dir.glob("#{dir}/**/*.yml").each do |file|
-    content = File.read(file)
-    cassettes << file if patterns.values.any? { |pattern| content.match?(pattern) }
+    basename = File.basename(file)
+
+    # Precise matching to avoid cross-provider confusion
+    providers.each do |provider|
+      # Match only exact provider prefixes
+      next unless basename =~ /^[^_]*_#{provider}_/ || # For first section like "chat_openai_"
+                  basename =~ /_#{provider}_[^_]+_/    # For middle sections like "_openai_gpt4_"
+
+      cassettes << file
+      break
+    end
   end
 
   cassettes

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: ruby_llm
 version: !ruby/object:Gem::Version
-  version: 1.0.1
+  version: 1.1.0rc2
 platform: ruby
 authors:
 - Carmine Paolino
 autorequire:
-bindir: exe
+bindir: bin
 cert_chain: []
-date: 2025-03-23 00:00:00.000000000 Z
+date: 2025-04-04 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: base64
@@ -96,31 +96,21 @@ dependencies:
         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, 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.
+  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.
 email:
 - carmine@paolino.me
 executables: []
 extensions: []
 extra_rdoc_files: []
 files:
-- ".github/workflows/cicd.yml"
-- ".github/workflows/docs.yml"
-- ".gitignore"
-- ".overcommit.yml"
-- ".rspec"
-- ".rubocop.yml"
-- ".yardopts"
-- CONTRIBUTING.md
-- Gemfile
 - LICENSE
 - README.md
-- Rakefile
-- bin/console
-- bin/setup
 - lib/ruby_llm.rb
 - lib/ruby_llm/active_record/acts_as.rb
+- lib/ruby_llm/aliases.json
+- lib/ruby_llm/aliases.rb
 - lib/ruby_llm/chat.rb
 - lib/ruby_llm/chunk.rb
 - lib/ruby_llm/configuration.rb
@@ -141,8 +131,20 @@ files:
 - lib/ruby_llm/providers/anthropic/models.rb
 - lib/ruby_llm/providers/anthropic/streaming.rb
 - lib/ruby_llm/providers/anthropic/tools.rb
+- 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/models.rb
+- lib/ruby_llm/providers/bedrock/signing.rb
+- lib/ruby_llm/providers/bedrock/streaming.rb
+- lib/ruby_llm/providers/bedrock/streaming/base.rb
+- lib/ruby_llm/providers/bedrock/streaming/content_extraction.rb
+- lib/ruby_llm/providers/bedrock/streaming/message_processing.rb
+- lib/ruby_llm/providers/bedrock/streaming/payload_processing.rb
+- lib/ruby_llm/providers/bedrock/streaming/prelude_handling.rb
 - lib/ruby_llm/providers/deepseek.rb
 - lib/ruby_llm/providers/deepseek/capabilities.rb
+- lib/ruby_llm/providers/deepseek/chat.rb
 - lib/ruby_llm/providers/gemini.rb
 - lib/ruby_llm/providers/gemini/capabilities.rb
 - lib/ruby_llm/providers/gemini/chat.rb
@@ -163,12 +165,19 @@ files:
 - lib/ruby_llm/providers/openai/tools.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/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
-- ruby_llm.gemspec
 homepage: https://rubyllm.com
 licenses:
 - MIT