ruby_llm 0.1.0.pre15 → 0.1.0.pre17

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: ca489465046f71efd3a80717b46a68cdcbc754527f4b5ccade15442d2a95854e
-  data.tar.gz: b1091dd99c8b96266b767b1be69ade14a88cffc0c3f2cd07c1bf20d9efebe6e2
+  metadata.gz: 899bee99690875afeb0d8bb83b654cf05ccaa4db7006971eb122e7188e91f6b7
+  data.tar.gz: a050cc712573b90bc2d2d1012f40c1866be940e001b6d2aaff6089834bf89584
 SHA512:
-  metadata.gz: 8bd968d14ec94fb254ae21a3d253cbe8b952145d3626540bf92cc5ded13d660e8f45b651319d3591f6a6e6b137c02cb94540c2a00680c25c2d7d3ecf4d799bbf
-  data.tar.gz: d1a5e1ef82b95f5ebe59b27440d703e88d4332f47cb706da8c4f09487608d6de572e78a38756af938a0e0ed0dbf8dc0f4a45aa82ac1d32f444681c3172baf40a
+  metadata.gz: f18a0c1094c407d7690151825f1c0cc557e2621fda75569c052ab9d396a46b7201e8aa3ba643a42da3d1b52e2a8a95216d315f66632cd6cc4de1fee0a372cfb1
+  data.tar.gz: 312856c0e943e93933e33e492d1e599732b6714a527cf83ac3d5b271d50ecd004064ae7aec5b5671e5dca2c83dba3e458794f0a87c0c12f19c27357a89b9ad61

data/README.md

@@ -76,6 +76,33 @@ last_message = chat.messages.last
 puts "Conversation used #{last_message.input_tokens} input tokens and #{last_message.output_tokens} output tokens"
 ```
 
+## Text Embeddings
+
+Need vector embeddings for your text? RubyLLM makes it simple:
+
+```ruby
+# Get embeddings with the default model
+vector = RubyLLM.embed(text: "Hello, world!")
+
+# Use a specific model
+vector = RubyLLM.embed(
+  "Ruby is awesome!",
+  model: "text-embedding-3-large"
+)
+
+# Process multiple texts at once
+vectors = RubyLLM.embed([
+  "First document",
+  "Second document",
+  "Third document"
+])
+
+# Configure the default model
+RubyLLM.configure do |config|
+  config.default_embedding_model = 'text-embedding-3-large'
+end
+```
+
 ## Using Tools
 
 Give your AI assistants access to your Ruby code by creating tool classes that do one thing well:

data/lib/ruby_llm/active_record/acts_as.rb

@@ -9,7 +9,7 @@ module RubyLLM
       extend ActiveSupport::Concern
 
       class_methods do
-        def acts_as_chat(message_class: 'Message')
+        def acts_as_chat(message_class: 'Message') # rubocop:disable Metrics/MethodLength
           include ChatMethods
 
           has_many :messages,
@@ -17,13 +17,23 @@ module RubyLLM
                    class_name: message_class.to_s,
                    dependent: :destroy
 
-          delegate :complete, to: :chat
+          delegate :complete,
+                   :with_tool,
+                   :with_tools,
+                   :with_model,
+                   :with_temperature,
+                   :on_new_message,
+                   :on_end_message,
+                   :add_message,
+                   to: :to_llm
         end
 
         def acts_as_message(chat_class: 'Chat')
           include MessageMethods
 
           belongs_to :chat, class_name: chat_class.to_s
+
+          delegate :tool_call?, :tool_result?, :tool_results, to: :to_llm
         end
       end
     end
@@ -33,7 +43,7 @@ module RubyLLM
     module ChatMethods
       extend ActiveSupport::Concern
 
-      def chat
+      def to_llm
         chat = RubyLLM.chat(model: model_id)
 
         # Load existing messages into chat
@@ -44,8 +54,6 @@ module RubyLLM
         # Set up message persistence
         chat.on_new_message { persist_new_message }
             .on_end_message { |msg| persist_message_completion(msg) }
-
-        chat
       end
 
       def ask(message, &block)
@@ -59,7 +67,10 @@ module RubyLLM
       private
 
       def persist_new_message
-        messages.create!
+        messages.create!(
+          role: :assistant,
+          content: String.new
+        )
       end
 
       def persist_message_completion(message)

data/lib/ruby_llm/configuration.rb

@@ -10,11 +10,16 @@ module RubyLLM
   #     config.anthropic_api_key = ENV['ANTHROPIC_API_KEY']
   #   end
   class Configuration
-    attr_accessor :openai_api_key, :anthropic_api_key, :default_model, :request_timeout
+    attr_accessor :openai_api_key,
+                  :anthropic_api_key,
+                  :default_model,
+                  :default_embedding_model,
+                  :request_timeout
 
     def initialize
       @request_timeout = 30
       @default_model = 'gpt-4o-mini'
+      @default_embedding_model = 'text-embedding-3-small'
     end
   end
 end

data/lib/ruby_llm/embedding.rb

@@ -0,0 +1,17 @@
+# frozen_string_literal: true
+
+module RubyLLM
+  # Core embedding interface. Provides a clean way to generate embeddings
+  # from text using various provider models.
+  module Embedding
+    module_function
+
+    def embed(text, model: nil)
+      model_id = model || RubyLLM.config.default_embedding_model
+      Models.find(model_id)
+
+      provider = Provider.for(model_id)
+      provider.embed(text, model: model_id)
+    end
+  end
+end

data/lib/ruby_llm/provider.rb

@@ -30,17 +30,23 @@ module RubyLLM
         parse_list_models_response response
       end
 
+      def embed(text, model:)
+        payload = build_embedding_payload text, model: model
+        response = post embedding_url, payload
+        parse_embedding_response response
+      end
+
       private
 
       def sync_response(payload)
-        response = post payload
+        response = post completion_url, payload
         parse_completion_response response
       end
 
       def stream_response(payload, &block)
         accumulator = StreamAccumulator.new
 
-        post payload do |req|
+        post completion_url, payload do |req|
           req.options.on_data = handle_stream do |chunk|
             accumulator.add chunk
             block.call chunk
@@ -50,8 +56,8 @@ module RubyLLM
         accumulator.to_message
       end
 
-      def post(payload)
-        connection.post completion_url, payload do |req|
+      def post(url, payload)
+        connection.post url, payload do |req|
           req.headers.merge! headers
           yield req if block_given?
         end

data/lib/ruby_llm/providers/openai.rb

@@ -28,6 +28,10 @@ module RubyLLM
         '/v1/models'
       end
 
+      def embedding_url
+        '/v1/embeddings'
+      end
+
       def build_payload(messages, tools:, temperature:, model:, stream: false) # rubocop:disable Metrics/MethodLength
         {
           model: model,
@@ -53,6 +57,18 @@ module RubyLLM
         end
       end
 
+      def build_embedding_payload(text, model:)
+        {
+          model: model,
+          input: text
+        }
+      end
+
+      def parse_embedding_response(response)
+        embeddings = response.body['data'].map { |d| d['embedding'] }
+        embeddings.size == 1 ? embeddings.first : embeddings
+      end
+
       def format_tool_calls(tool_calls) # rubocop:disable Metrics/MethodLength
         return nil unless tool_calls&.any?
 

data/lib/ruby_llm/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module RubyLLM
-  VERSION = '0.1.0.pre15'
+  VERSION = '0.1.0.pre17'
 end

data/lib/ruby_llm.rb

@@ -27,6 +27,10 @@ module RubyLLM
       Chat.new(model: model)
     end
 
+    def embed(...)
+      Embedding.embed(...)
+    end
+
     def models
       Models
     end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: ruby_llm
 version: !ruby/object:Gem::Version
-  version: 0.1.0.pre15
+  version: 0.1.0.pre17
 platform: ruby
 authors:
 - Carmine Paolino
 autorequire:
 bindir: exe
 cert_chain: []
-date: 2025-02-04 00:00:00.000000000 Z
+date: 2025-02-06 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: event_stream_parser
@@ -348,6 +348,7 @@ files:
 - lib/ruby_llm/chat.rb
 - lib/ruby_llm/chunk.rb
 - lib/ruby_llm/configuration.rb
+- lib/ruby_llm/embedding.rb
 - lib/ruby_llm/message.rb
 - lib/ruby_llm/model_capabilities/anthropic.rb
 - lib/ruby_llm/model_capabilities/openai.rb