langchainrb 0.3.14 → 0.4.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 33c9436ac8d6a73dc06d30f63c11e4f246b3705aa8934765a53ee59325c3a9cd
-  data.tar.gz: 9cc85603694f9367dd162e25379029a345aa0b5c88cccf303c2af114d43a4010
+  metadata.gz: ce8728ec2208577809174e154642db161121cb9dd49e0ec5d190d080e68b1d78
+  data.tar.gz: bb0e0ccc4558ca849549f495a4adfacc5f7851c786869974afdaef29f0cde3ca
 SHA512:
-  metadata.gz: ca5e81638625939d11999a64d44c92fc57c762a934aa8fd5b110c3f5aacc9a736ab5f02da4366e7a1b9b9ec0335dd1eb1683f5b9d90bd97c81914ea0a698dc7c
-  data.tar.gz: aff49ef9451bcbc9a97d181757a5b913737cbfbb4fc3ca49d423cbd2e59a4a71091816e98c2996ff7f8292cb6e0c0d69931a4f3e4e36e2a69fdd7f745640e266
+  metadata.gz: b2b4c27e31d730563aeca70a0aa3c4cf129e69773e34f397ba057faa8298a4368c1b9f66f925188f867f1feb47b4e07f77df702fa7c6cb76ad1e1a8464b895f6
+  data.tar.gz: 55dd3fbc21e2cdf9bd84afcd6bb4de0f72c960dec0c6b1d2efff1f9492b3d5c7399f2d14c323597045e64eafb6f2f20992348d640317c64721fb0556f8a64126

data/CHANGELOG.md

@@ -1,5 +1,16 @@
 ## [Unreleased]
 
+## [0.4.0] - 2023-06-01
+- [BREAKING] Everything is namespaced under `Langchain::` now
+- Pgvector similarity search uses the cosine distance by default now
+- OpenAI token length validation using tiktoken_ruby
+
+## [0.3.15] - 2023-05-30
+- Drop Ruby 2.7 support. It had reached EOD.
+- Bump pgvector-ruby to 0.2
+- 🚚 Loaders
+  - Support for options and block to be passed to CSV processor
+
 ## [0.3.14] - 2023-05-28
 - 🔍 Vectorsearch
   - Not relying on Weaviate modules anymore

data/Gemfile.lock

@@ -1,7 +1,8 @@
 PATH
   remote: .
   specs:
-    langchainrb (0.3.14)
+    langchainrb (0.4.0)
+      tiktoken_ruby (~> 0.0.5)
 
 GEM
   remote: https://rubygems.org/
@@ -171,7 +172,7 @@ GEM
       ruby-rc4
       ttfunk
     pg (1.5.3)
-    pgvector (0.1.1)
+    pgvector (0.2.0)
     pinecone (0.1.71)
       dry-struct (~> 1.6.0)
       dry-validation (~> 1.10.0)
@@ -205,6 +206,7 @@ GEM
       zeitwerk (~> 2.5)
     rainbow (3.1.1)
     rake (13.0.6)
+    rb_sys (0.9.78)
     regexp_parser (2.8.0)
     replicate-ruby (0.2.2)
       addressable
@@ -262,6 +264,11 @@ GEM
     standardrb (1.0.1)
       standard
     thor (1.2.1)
+    tiktoken_ruby (0.0.5)
+      rb_sys (~> 0.9.68)
+    tiktoken_ruby (0.0.5-arm64-darwin)
+    tiktoken_ruby (0.0.5-x86_64-darwin)
+    tiktoken_ruby (0.0.5-x86_64-linux)
     treetop (1.6.12)
       polyglot (~> 0.3)
     ttfunk (1.7.0)
@@ -298,7 +305,7 @@ DEPENDENCIES
   nokogiri (~> 1.13)
   pdf-reader (~> 1.4)
   pg (~> 1.5)
-  pgvector (< 0.2)
+  pgvector (~> 0.2)
   pinecone (~> 0.1.6)
   pry-byebug (~> 3.10.0)
   qdrant-ruby (~> 0.9.0)

data/README.md

@@ -44,7 +44,7 @@ Add `gem "weaviate-ruby", "~> 0.8.0"`  to your Gemfile.
 
 Pick the vector search database you'll be using and instantiate the client:
 ```ruby
-client = Vectorsearch::Weaviate.new(
+client = Langchain::Vectorsearch::Weaviate.new(
     url: ENV["WEAVIATE_URL"],
     api_key: ENV["WEAVIATE_API_KEY"],
     llm: :openai, # or :cohere
@@ -52,10 +52,10 @@ client = Vectorsearch::Weaviate.new(
 )
 
 # You can instantiate any other supported vector search database:
-client = Vectorsearch::Milvus.new(...) # `gem "milvus", "~> 0.9.0"`
-client = Vectorsearch::Qdrant.new(...) # `gem"qdrant-ruby", "~> 0.9.0"`
-client = Vectorsearch::Pinecone.new(...) # `gem "pinecone", "~> 0.1.6"`
-client = Vectorsearch::Chroma.new(...) # `gem "chroma-db", "~> 0.3.0"`
+client = Langchain::Vectorsearch::Milvus.new(...) # `gem "milvus", "~> 0.9.0"`
+client = Langchain::Vectorsearch::Qdrant.new(...) # `gem"qdrant-ruby", "~> 0.9.0"`
+client = Langchain::Vectorsearch::Pinecone.new(...) # `gem "pinecone", "~> 0.1.6"`
+client = Langchain::Vectorsearch::Chroma.new(...) # `gem "chroma-db", "~> 0.3.0"`
 ```
 
 ```ruby
@@ -107,7 +107,7 @@ Add `gem "ruby-openai", "~> 4.0.0"` to your Gemfile.
 
 #### OpenAI
 ```ruby
-openai = LLM::OpenAI.new(api_key: ENV["OPENAI_API_KEY"])
+openai = Langchain::LLM::OpenAI.new(api_key: ENV["OPENAI_API_KEY"])
 ```
 ```ruby
 openai.embed(text: "foo bar")
@@ -120,7 +120,7 @@ openai.complete(prompt: "What is the meaning of life?")
 Add `gem "cohere-ruby", "~> 0.9.3"` to your Gemfile.
 
 ```ruby
-cohere = LLM::Cohere.new(api_key: ENV["COHERE_API_KEY"])
+cohere = Langchain::LLM::Cohere.new(api_key: ENV["COHERE_API_KEY"])
 ```
 ```ruby
 cohere.embed(text: "foo bar")
@@ -132,19 +132,19 @@ cohere.complete(prompt: "What is the meaning of life?")
 #### HuggingFace
 Add `gem "hugging-face", "~> 0.3.2"` to your Gemfile.
 ```ruby
-cohere = LLM::HuggingFace.new(api_key: ENV["HUGGING_FACE_API_KEY"])
+cohere = Langchain::LLM::HuggingFace.new(api_key: ENV["HUGGING_FACE_API_KEY"])
 ```
 
 #### Replicate
 Add `gem "replicate-ruby", "~> 0.2.2"` to your Gemfile.
 ```ruby
-cohere = LLM::Replicate.new(api_key: ENV["REPLICATE_API_KEY"])
+cohere = Langchain::LLM::Replicate.new(api_key: ENV["REPLICATE_API_KEY"])
 ```
 
 #### Google PaLM (Pathways Language Model)
 Add `"google_palm_api", "~> 0.1.0"` to your Gemfile.
 ```ruby
-google_palm = LLM::GooglePalm.new(api_key: ENV["GOOGLE_PALM_API_KEY"])
+google_palm = Langchain::LLM::GooglePalm.new(api_key: ENV["GOOGLE_PALM_API_KEY"])
 ```
 
 ### Using Prompts 📋
@@ -154,21 +154,21 @@ google_palm = LLM::GooglePalm.new(api_key: ENV["GOOGLE_PALM_API_KEY"])
 Create a prompt with one input variable:
 
 ```ruby
-prompt = Prompt::PromptTemplate.new(template: "Tell me a {adjective} joke.", input_variables: ["adjective"])
+prompt = Langchain::Prompt::PromptTemplate.new(template: "Tell me a {adjective} joke.", input_variables: ["adjective"])
 prompt.format(adjective: "funny") # "Tell me a funny joke."
 ```
 
 Create a prompt with multiple input variables:
 
 ```ruby
-prompt = Prompt::PromptTemplate.new(template: "Tell me a {adjective} joke about {content}.", input_variables: ["adjective", "content"])
+prompt = Langchain::Prompt::PromptTemplate.new(template: "Tell me a {adjective} joke about {content}.", input_variables: ["adjective", "content"])
 prompt.format(adjective: "funny", content: "chickens") # "Tell me a funny joke about chickens."
 ```
 
 Creating a PromptTemplate using just a prompt and no input_variables:
 
 ```ruby
-prompt = Prompt::PromptTemplate.from_template("Tell me a {adjective} joke about {content}.")
+prompt = Langchain::Prompt::PromptTemplate.from_template("Tell me a {adjective} joke about {content}.")
 prompt.input_variables # ["adjective", "content"]
 prompt.format(adjective: "funny", content: "chickens") # "Tell me a funny joke about chickens."
 ```
@@ -182,7 +182,7 @@ prompt.save(file_path: "spec/fixtures/prompt/prompt_template.json")
 Loading a new prompt template using a JSON file:
 
 ```ruby
-prompt = Prompt.load_from_path(file_path: "spec/fixtures/prompt/prompt_template.json")
+prompt = Langchain::Prompt.load_from_path(file_path: "spec/fixtures/prompt/prompt_template.json")
 prompt.input_variables # ["adjective", "content"]
 ```
 
@@ -191,10 +191,10 @@ prompt.input_variables # ["adjective", "content"]
 Create a prompt with a few shot examples:
 
 ```ruby
-prompt = Prompt::FewShotPromptTemplate.new(
+prompt = Langchain::Prompt::FewShotPromptTemplate.new(
   prefix: "Write antonyms for the following words.",
   suffix: "Input: {adjective}\nOutput:",
-  example_prompt: Prompt::PromptTemplate.new(
+  example_prompt: Langchain::Prompt::PromptTemplate.new(
     input_variables: ["input", "output"],
     template: "Input: {input}\nOutput: {output}"
   ),
@@ -228,7 +228,7 @@ prompt.save(file_path: "spec/fixtures/prompt/few_shot_prompt_template.json")
 Loading a new prompt template using a JSON file:
 
 ```ruby
-prompt = Prompt.load_from_path(file_path: "spec/fixtures/prompt/few_shot_prompt_template.json")
+prompt = Langchain::Prompt.load_from_path(file_path: "spec/fixtures/prompt/few_shot_prompt_template.json")
 prompt.prefix # "Write antonyms for the following words."
 ```
 
@@ -237,10 +237,10 @@ Agents are semi-autonomous bots that can respond to user questions and use avail
 
 #### Chain-of-Thought Agent
 
-Add `gem "openai-ruby"`, `gem "eqn"`, and `gem "google_search_results"` to your Gemfile
+Add `gem "ruby-openai"`, `gem "eqn"`, and `gem "google_search_results"` to your Gemfile
 
 ```ruby
-agent = Agent::ChainOfThoughtAgent.new(llm: :openai, llm_api_key: ENV["OPENAI_API_KEY"], tools: ['search', 'calculator'])
+agent = Langchain::Agent::ChainOfThoughtAgent.new(llm: :openai, llm_api_key: ENV["OPENAI_API_KEY"], tools: ['search', 'calculator'])
 
 agent.tools
 # => ["search", "calculator"]
@@ -273,7 +273,7 @@ Need to read data from various sources? Load it up.
 Just call `Langchan::Loader.load` with the path to the file or a URL you want to load.
 
 ```ruby
-Langchaing::Loader.load('/path/to/file.pdf')
+Langchain::Loader.load('/path/to/file.pdf')
 ```
 
 or

data/lib/{agent → langchain/agent}/base.rb

@@ -1,6 +1,6 @@
 # frozen_string_literal: true
 
-module Agent
+module Langchain::Agent
   class Base
   end
 end

data/lib/{agent → langchain/agent}/chain_of_thought_agent/chain_of_thought_agent.rb

@@ -1,6 +1,6 @@
 # frozen_string_literal: true
 
-module Agent
+module Langchain::Agent
   class ChainOfThoughtAgent < Base
     attr_reader :llm, :llm_api_key, :llm_client, :tools
 
@@ -11,14 +11,14 @@ module Agent
     # @param tools [Array] The tools to use
     # @return [ChainOfThoughtAgent] The Agent::ChainOfThoughtAgent instance
     def initialize(llm:, llm_api_key:, tools: [])
-      LLM::Base.validate_llm!(llm: llm)
-      Tool::Base.validate_tools!(tools: tools)
+      Langchain::LLM::Base.validate_llm!(llm: llm)
+      Langchain::Tool::Base.validate_tools!(tools: tools)
 
       @llm = llm
       @llm_api_key = llm_api_key
       @tools = tools
 
-      @llm_client = LLM.const_get(LLM::Base::LLMS.fetch(llm)).new(api_key: llm_api_key)
+      @llm_client = Langchain::LLM.const_get(Langchain::LLM::Base::LLMS.fetch(llm)).new(api_key: llm_api_key)
     end
 
     # Validate tools when they're re-assigned
@@ -26,7 +26,7 @@ module Agent
     # @param value [Array] The tools to use
     # @return [Array] The tools that will be used
     def tools=(value)
-      Tool::Base.validate_tools!(tools: value)
+      Langchain::Tool::Base.validate_tools!(tools: value)
       @tools = value
     end
 
@@ -62,8 +62,8 @@ module Agent
           Langchain.logger.info("Agent: Using the \"#{action}\" Tool with \"#{action_input}\"")
 
           # Retrieve the Tool::[ToolName] class and call `execute`` with action_input as the input
-          result = Tool
-            .const_get(Tool::Base::TOOLS[action.strip])
+          result = Langchain::Tool
+            .const_get(Langchain::Tool::Base::TOOLS[action.strip])
             .execute(input: action_input)
 
           # Append the Observation to the prompt
@@ -91,7 +91,7 @@ module Agent
         question: question,
         tool_names: "[#{tools.join(", ")}]",
         tools: tools.map do |tool|
-          "#{tool}: #{Tool.const_get(Tool::Base::TOOLS[tool]).const_get(:DESCRIPTION)}"
+          "#{tool}: #{Langchain::Tool.const_get(Langchain::Tool::Base::TOOLS[tool]).const_get(:DESCRIPTION)}"
         end.join("\n")
       )
     end
@@ -99,8 +99,8 @@ module Agent
     # Load the PromptTemplate from the JSON file
     # @return [PromptTemplate] PromptTemplate instance
     def prompt_template
-      @template ||= Prompt.load_from_path(
-        file_path: Langchain.root.join("agent/chain_of_thought_agent/chain_of_thought_agent_prompt.json")
+      @template ||= Langchain::Prompt.load_from_path(
+        file_path: Langchain.root.join("langchain/agent/chain_of_thought_agent/chain_of_thought_agent_prompt.json")
       )
     end
   end

data/lib/{llm → langchain/llm}/base.rb

@@ -1,6 +1,6 @@
 # frozen_string_literal: true
 
-module LLM
+module Langchain::LLM
   class Base
     attr_reader :client
 
@@ -42,8 +42,8 @@ module LLM
     # @param llm [Symbol] The LLM to use
     def self.validate_llm!(llm:)
       # TODO: Fix so this works when `llm` value is a string instead of a symbol
-      unless LLM::Base::LLMS.key?(llm)
-        raise ArgumentError, "LLM must be one of #{LLM::Base::LLMS.keys}"
+      unless Langchain::LLM::Base::LLMS.key?(llm)
+        raise ArgumentError, "LLM must be one of #{Langchain::LLM::Base::LLMS.keys}"
       end
     end
   end

data/lib/{llm → langchain/llm}/cohere.rb

@@ -1,6 +1,6 @@
 # frozen_string_literal: true
 
-module LLM
+module Langchain::LLM
   class Cohere < Base
     DEFAULTS = {
       temperature: 0.0,

data/lib/{llm → langchain/llm}/google_palm.rb

@@ -1,6 +1,6 @@
 # frozen_string_literal: true
 
-module LLM
+module Langchain::LLM
   class GooglePalm < Base
     # Wrapper around the Google PaLM (Pathways Language Model) APIs.
 
@@ -89,8 +89,8 @@ module LLM
     # @return [String] The summarization
     #
     def summarize(text:)
-      prompt_template = Prompt.load_from_path(
-        file_path: Langchain.root.join("llm/prompts/summarize_template.json")
+      prompt_template = Langchain::Prompt.load_from_path(
+        file_path: Langchain.root.join("langchain/llm/prompts/summarize_template.json")
       )
       prompt = prompt_template.format(text: text)
 

data/lib/{llm → langchain/llm}/hugging_face.rb

@@ -1,6 +1,6 @@
 # frozen_string_literal: true
 
-module LLM
+module Langchain::LLM
   class HuggingFace < Base
     # The gem does not currently accept other models:
     # https://github.com/alchaplinsky/hugging-face/blob/main/lib/hugging_face/inference_api.rb#L32-L34

data/lib/{llm → langchain/llm}/openai.rb

@@ -1,6 +1,6 @@
 # frozen_string_literal: true
 
-module LLM
+module Langchain::LLM
   class OpenAI < Base
     DEFAULTS = {
       temperature: 0.0,
@@ -25,9 +25,13 @@ module LLM
     # @return [Array] The embedding
     #
     def embed(text:)
+      model = DEFAULTS[:embeddings_model_name]
+
+      Langchain::Utils::TokenLengthValidator.validate!(text, model)
+
       response = client.embeddings(
         parameters: {
-          model: DEFAULTS[:embeddings_model_name],
+          model: model,
           input: text
         }
       )
@@ -41,8 +45,12 @@ module LLM
     # @return [String] The completion
     #
     def complete(prompt:, **params)
+      model = DEFAULTS[:completion_model_name]
+
+      Langchain::Utils::TokenLengthValidator.validate!(prompt, model)
+
       default_params = {
-        model: DEFAULTS[:completion_model_name],
+        model: model,
         temperature: DEFAULTS[:temperature],
         prompt: prompt
       }
@@ -64,8 +72,12 @@ module LLM
     # @return [String] The chat completion
     #
     def chat(prompt:, **params)
+      model = DEFAULTS[:chat_completion_model_name]
+
+      Langchain::Utils::TokenLengthValidator.validate!(prompt, model)
+
       default_params = {
-        model: DEFAULTS[:chat_completion_model_name],
+        model: model,
         temperature: DEFAULTS[:temperature],
         # TODO: Figure out how to introduce persisted conversations
         messages: [{role: "user", content: prompt}]
@@ -88,8 +100,8 @@ module LLM
     # @return [String] The summary
     #
     def summarize(text:)
-      prompt_template = Prompt.load_from_path(
-        file_path: Langchain.root.join("llm/prompts/summarize_template.json")
+      prompt_template = Langchain::Prompt.load_from_path(
+        file_path: Langchain.root.join("langchain/llm/prompts/summarize_template.json")
       )
       prompt = prompt_template.format(text: text)
 

data/lib/{llm → langchain/llm}/replicate.rb

@@ -1,6 +1,6 @@
 # frozen_string_literal: true
 
-module LLM
+module Langchain::LLM
   class Replicate < Base
     # Wrapper around Replicate.com LLM provider
     # Use it directly:
@@ -89,8 +89,8 @@ module LLM
     # @return [String] The summary
     #
     def summarize(text:)
-      prompt_template = Prompt.load_from_path(
-        file_path: Langchain.root.join("llm/prompts/summarize_template.json")
+      prompt_template = Langchain::Prompt.load_from_path(
+        file_path: Langchain.root.join("langchain/llm/prompts/summarize_template.json")
       )
       prompt = prompt_template.format(text: text)
 

data/lib/langchain/loader.rb

@@ -14,14 +14,15 @@ module Langchain
     # Equivalent to Langchain::Loader.new(path).load
     # @param path [String | Pathname] path to file or url
     # @return [String] file content
-    def self.load(path)
-      new(path).load
+    def self.load(path, options = {}, &block)
+      new(path, options).load(&block)
     end
 
     # Initialize Langchain::Loader
     # @param path [String | Pathname] path to file or url
     # @return [Langchain::Loader] loader instance
-    def initialize(path)
+    def initialize(path, options = {})
+      @options = options
       @path = path
     end
 
@@ -35,39 +36,38 @@ module Langchain
 
     # Load data from a file or url
     # @return [String] file content
-    def load
-      url? ? from_url(@path) : from_path(@path)
+    def load(&block)
+      @raw_data = url? ? load_from_url : load_from_path
+
+      data = if block
+        yield @raw_data.read, @options
+      else
+        processor_klass.new(@options).parse(@raw_data)
+      end
+
+      Langchain::Data.new(data, source: @path)
     end
 
     private
 
-    def from_url(url)
-      process do
-        data = URI.parse(url).open
-        processor = find_processor(:CONTENT_TYPES, data.content_type)
-        [data, processor]
-      end
+    def load_from_url
+      URI.parse(@path).open
     end
 
-    def from_path(path)
-      raise FileNotFound unless File.exist?(path)
+    def load_from_path
+      raise FileNotFound unless File.exist?(@path)
 
-      process do
-        [File.open(path), find_processor(:EXTENSIONS, File.extname(path))]
-      end
+      File.open(@path)
     end
 
-    def process(&block)
-      raw_data, kind = yield
+    def processor_klass
+      raise UnknownFormatError unless (kind = find_processor)
 
-      raise UnknownFormatError unless kind
-
-      processor = Langchain::Processors.const_get(kind).new
-      Langchain::Data.new(processor.parse(raw_data), source: @path)
+      Langchain::Processors.const_get(kind)
     end
 
-    def find_processor(constant, value)
-      processors.find { |klass| processor_matches? "#{klass}::#{constant}", value }
+    def find_processor
+      processors.find { |klass| processor_matches? "#{klass}::#{lookup_constant}", source_type }
     end
 
     def processor_matches?(constant, value)
@@ -77,5 +77,13 @@ module Langchain
     def processors
       Langchain::Processors.constants
     end
+
+    def source_type
+      url? ? @raw_data.content_type : File.extname(@path)
+    end
+
+    def lookup_constant
+      url? ? :CONTENT_TYPES : :EXTENSIONS
+    end
   end
 end

data/lib/langchain/processors/base.rb

@@ -6,6 +6,10 @@ module Langchain
       EXTENSIONS = []
       CONTENT_TYPES = []
 
+      def initialize(options = {})
+        @options = options
+      end
+
       def parse(data)
         raise NotImplementedError
       end

data/lib/langchain/processors/csv.rb

@@ -12,10 +12,16 @@ module Langchain
       # @param [File] data
       # @return [Array of Hash]
       def parse(data)
-        ::CSV.new(data.read).map do |row|
+        ::CSV.new(data.read, col_sep: separator).map do |row|
           row.map(&:strip)
         end
       end
+
+      private
+
+      def separator
+        @options[:col_sep] || ","
+      end
     end
   end
 end

data/lib/langchain/processors/docx.rb

@@ -6,7 +6,7 @@ module Langchain
       EXTENSIONS = [".docx"]
       CONTENT_TYPES = ["application/vnd.openxmlformats-officedocument.wordprocessingml.document"]
 
-      def initialize
+      def initialize(*)
         depends_on "docx"
         require "docx"
       end

data/lib/langchain/processors/html.rb

@@ -9,7 +9,7 @@ module Langchain
       # We only look for headings and paragraphs
       TEXT_CONTENT_TAGS = %w[h1 h2 h3 h4 h5 h6 p]
 
-      def initialize
+      def initialize(*)
         depends_on "nokogiri"
         require "nokogiri"
       end

data/lib/langchain/processors/pdf.rb

@@ -6,7 +6,7 @@ module Langchain
       EXTENSIONS = [".pdf"]
       CONTENT_TYPES = ["application/pdf"]
 
-      def initialize
+      def initialize(*)
         depends_on "pdf-reader"
         require "pdf-reader"
       end

data/lib/{prompt → langchain/prompt}/base.rb

@@ -3,7 +3,7 @@
 require "strscan"
 require "json"
 
-module Prompt
+module Langchain::Prompt
   class Base
     def format(**kwargs)
       raise NotImplementedError
@@ -29,7 +29,7 @@ module Prompt
     #
     def validate(template:, input_variables:)
       input_variables_set = @input_variables.uniq
-      variables_from_template = Prompt::Base.extract_variables_from_template(template)
+      variables_from_template = Langchain::Prompt::Base.extract_variables_from_template(template)
 
       missing_variables = variables_from_template - input_variables_set
       extra_variables = input_variables_set - variables_from_template

data/lib/{prompt → langchain/prompt}/few_shot_prompt_template.rb

@@ -1,6 +1,6 @@
 # frozen_string_literal: true
 
-module Prompt
+module Langchain::Prompt
   class FewShotPromptTemplate < Base
     attr_reader :examples, :example_prompt, :input_variables, :prefix, :suffix, :example_separator
 

data/lib/{prompt → langchain/prompt}/loading.rb

@@ -3,10 +3,10 @@
 require "strscan"
 require "pathname"
 
-module Prompt
+module Langchain::Prompt
   TYPE_TO_LOADER = {
-    "prompt" => ->(config) { Prompt.load_prompt(config) },
-    "few_shot" => ->(config) { Prompt.load_few_shot_prompt(config) }
+    "prompt" => ->(config) { load_prompt(config) },
+    "few_shot" => ->(config) { load_few_shot_prompt(config) }
   }
 
   class << self

data/lib/{prompt → langchain/prompt}/prompt_template.rb

@@ -1,6 +1,6 @@
 # frozen_string_literal: true
 
-module Prompt
+module Langchain::Prompt
   class PromptTemplate < Base
     attr_reader :template, :input_variables, :validate_template
 

data/lib/{tool → langchain/tool}/base.rb

@@ -1,6 +1,6 @@
 # frozen_string_literal: true
 
-module Tool
+module Langchain::Tool
   class Base
     # How to add additional Tools?
     # 1. Create a new file in lib/tool/your_tool_name.rb
@@ -10,9 +10,9 @@ module Tool
     # 4. Add your tool to the README.md
 
     TOOLS = {
-      "calculator" => "Tool::Calculator",
-      "search" => "Tool::SerpApi",
-      "wikipedia" => "Tool::Wikipedia"
+      "calculator" => "Langchain::Tool::Calculator",
+      "search" => "Langchain::Tool::SerpApi",
+      "wikipedia" => "Langchain::Tool::Wikipedia"
     }
 
     def self.description(value)
@@ -40,7 +40,7 @@ module Tool
     # @raise [ArgumentError] If any of the tools are not supported
     #
     def self.validate_tools!(tools:)
-      unrecognized_tools = tools - Tool::Base::TOOLS.keys
+      unrecognized_tools = tools - Langchain::Tool::Base::TOOLS.keys
 
       if unrecognized_tools.any?
         raise ArgumentError, "Unrecognized Tools: #{unrecognized_tools}"

data/lib/{tool → langchain/tool}/calculator.rb

@@ -1,6 +1,6 @@
 # frozen_string_literal: true
 
-module Tool
+module Langchain::Tool
   class Calculator < Base
     description <<~DESC
       Useful for getting the result of a math expression.
@@ -22,7 +22,7 @@ module Tool
     rescue Eqn::ParseError, Eqn::NoVariableValueError
       # Sometimes the input is not a pure math expression, e.g: "12F in Celsius"
       # We can use the google answer box to evaluate this expression
-      hash_results = Tool::SerpApi.execute_search(input: input)
+      hash_results = Langchain::Tool::SerpApi.execute_search(input: input)
       hash_results.dig(:answer_box, :to)
     end
   end

data/lib/{tool → langchain/tool}/serp_api.rb

@@ -1,6 +1,6 @@
 # frozen_string_literal: true
 
-module Tool
+module Langchain::Tool
   class SerpApi < Base
     # Wrapper around SerpAPI
     # Set ENV["SERPAPI_API_KEY"] to use it

data/lib/{tool → langchain/tool}/wikipedia.rb

@@ -1,6 +1,6 @@
 # frozen_string_literal: true
 
-module Tool
+module Langchain::Tool
   class Wikipedia < Base
     # Tool that adds the capability to search using the Wikipedia API
 

data/lib/langchain/utils/token_length_validator.rb

@@ -0,0 +1,57 @@
+# frozen_string_literal: true
+
+require "tiktoken_ruby"
+
+module Langchain
+  module Utils
+    class TokenLimitExceeded < StandardError; end
+
+    class TokenLengthValidator
+      #
+      # This class is meant to validate the length of the text passed in to OpenAI's API.
+      # It is used to validate the token length before the API call is made
+      #
+      TOKEN_LIMITS = {
+        # Source:
+        # https://platform.openai.com/docs/api-reference/embeddings
+        # https://platform.openai.com/docs/models/gpt-4
+        "text-embedding-ada-002" => 8191,
+        "gpt-3.5-turbo" => 4096,
+        "gpt-3.5-turbo-0301" => 4096,
+        "text-davinci-003" => 4097,
+        "text-davinci-002" => 4097,
+        "code-davinci-002" => 8001,
+        "gpt-4" => 8192,
+        "gpt-4-0314" => 8192,
+        "gpt-4-32k" => 32768,
+        "gpt-4-32k-0314" => 32768,
+        "text-curie-001" => 2049,
+        "text-babbage-001" => 2049,
+        "text-ada-001" => 2049,
+        "davinci" => 2049,
+        "curie" => 2049,
+        "babbage" => 2049,
+        "ada" => 2049
+      }.freeze
+
+      #
+      # Validate the length of the text passed in to OpenAI's API
+      #
+      # @param text [String] The text to validate
+      # @param model_name [String] The model name to validate against
+      # @return [Boolean] Whether the text is valid or not
+      # @raise [TokenLimitExceeded] If the text is too long
+      #
+      def self.validate!(text, model_name)
+        encoder = Tiktoken.encoding_for_model(model_name)
+        token_length = encoder.encode(text).length
+
+        if token_length > TOKEN_LIMITS[model_name]
+          raise TokenLimitExceeded, "This model's maximum context length is #{TOKEN_LIMITS[model_name]} tokens, but the given text is #{token_length} tokens long."
+        end
+
+        true
+      end
+    end
+  end
+end

data/lib/{vectorsearch → langchain/vectorsearch}/base.rb

@@ -2,7 +2,7 @@
 
 require "forwardable"
 
-module Vectorsearch
+module Langchain::Vectorsearch
   class Base
     extend Forwardable
 
@@ -13,12 +13,12 @@ module Vectorsearch
     # @param llm [Symbol] The LLM to use
     # @param llm_api_key [String] The API key for the LLM
     def initialize(llm:, llm_api_key:)
-      LLM::Base.validate_llm!(llm: llm)
+      Langchain::LLM::Base.validate_llm!(llm: llm)
 
       @llm = llm
       @llm_api_key = llm_api_key
 
-      @llm_client = LLM.const_get(LLM::Base::LLMS.fetch(llm)).new(api_key: llm_api_key)
+      @llm_client = Langchain::LLM.const_get(Langchain::LLM::Base::LLMS.fetch(llm)).new(api_key: llm_api_key)
     end
 
     # Method supported by Vectorsearch DB to create a default schema
@@ -51,10 +51,10 @@ module Vectorsearch
       :default_dimension
 
     def generate_prompt(question:, context:)
-      prompt_template = Prompt::FewShotPromptTemplate.new(
+      prompt_template = Langchain::Prompt::FewShotPromptTemplate.new(
         prefix: "Context:",
         suffix: "---\nQuestion: {question}\n---\nAnswer:",
-        example_prompt: Prompt::PromptTemplate.new(
+        example_prompt: Langchain::Prompt::PromptTemplate.new(
           template: "{context}",
           input_variables: ["context"]
         ),

data/lib/{vectorsearch → langchain/vectorsearch}/chroma.rb

@@ -1,6 +1,6 @@
 # frozen_string_literal: true
 
-module Vectorsearch
+module Langchain::Vectorsearch
   class Chroma < Base
     # Initialize the Chroma client
     # @param url [String] The URL of the Qdrant server

data/lib/{vectorsearch → langchain/vectorsearch}/milvus.rb

@@ -1,6 +1,6 @@
 # frozen_string_literal: true
 
-module Vectorsearch
+module Langchain::Vectorsearch
   class Milvus < Base
     def initialize(url:, index_name:, llm:, llm_api_key:, api_key: nil)
       depends_on "milvus"

data/lib/{vectorsearch → langchain/vectorsearch}/pgvector.rb

@@ -1,8 +1,17 @@
 # frozen_string_literal: true
 
-module Vectorsearch
+module Langchain::Vectorsearch
   # The PostgreSQL vector search adapter
   class Pgvector < Base
+    # The operators supported by the PostgreSQL vector search adapter
+    OPERATORS = {
+      "cosine_distance" => "<=>",
+      "euclidean_distance" => "<->"
+    }
+    DEFAULT_OPERATOR = "cosine_distance"
+
+    attr_reader :operator, :quoted_table_name
+
     # @param url [String] The URL of the PostgreSQL database
     # @param index_name [String] The name of the table to use for the index
     # @param llm [String] The URL of the Language Layer API
@@ -18,6 +27,8 @@ module Vectorsearch
       @client.type_map_for_results = PG::BasicTypeMapForResults.new(@client, registry: registry)
 
       @index_name = index_name
+      @quoted_table_name = @client.quote_ident(index_name)
+      @operator = OPERATORS[DEFAULT_OPERATOR]
 
       super(llm: llm, llm_api_key: llm_api_key)
     end
@@ -31,7 +42,7 @@ module Vectorsearch
       end
       values = texts.length.times.map { |i| "($#{2 * i + 1}, $#{2 * i + 2})" }.join(",")
       client.exec_params(
-        "INSERT INTO #{@index_name} (content, vectors) VALUES #{values};",
+        "INSERT INTO #{quoted_table_name} (content, vectors) VALUES #{values};",
         data
       )
     end
@@ -42,7 +53,7 @@ module Vectorsearch
       client.exec("CREATE EXTENSION IF NOT EXISTS vector;")
       client.exec(
         <<~SQL
-          CREATE TABLE IF NOT EXISTS #{@index_name} (
+          CREATE TABLE IF NOT EXISTS #{quoted_table_name} (
             id serial PRIMARY KEY,
             content TEXT,
             vectors VECTOR(#{default_dimension})
@@ -73,7 +84,7 @@ module Vectorsearch
       result = client.transaction do |conn|
         conn.exec("SET LOCAL ivfflat.probes = 10;")
         query = <<~SQL
-          SELECT id, content FROM #{@index_name} ORDER BY vectors <-> $1 ASC LIMIT $2;
+          SELECT id, content FROM #{quoted_table_name} ORDER BY vectors #{operator} $1 ASC LIMIT $2;
         SQL
         conn.exec_params(query, [embedding, k])
       end

data/lib/{vectorsearch → langchain/vectorsearch}/pinecone.rb

@@ -1,6 +1,6 @@
 # frozen_string_literal: true
 
-module Vectorsearch
+module Langchain::Vectorsearch
   class Pinecone < Base
     # Initialize the Pinecone client
     # @param environment [String] The environment to use

data/lib/{vectorsearch → langchain/vectorsearch}/qdrant.rb

@@ -1,6 +1,6 @@
 # frozen_string_literal: true
 
-module Vectorsearch
+module Langchain::Vectorsearch
   class Qdrant < Base
     # Initialize the Qdrant client
     # @param url [String] The URL of the Qdrant server

data/lib/{vectorsearch → langchain/vectorsearch}/weaviate.rb

@@ -1,6 +1,6 @@
 # frozen_string_literal: true
 
-module Vectorsearch
+module Langchain::Vectorsearch
   class Weaviate < Base
     # Initialize the Weaviate adapter
     # @param url [String] The URL of the Weaviate instance

data/lib/langchain.rb

@@ -5,6 +5,7 @@ require "pathname"
 
 require_relative "./version"
 require_relative "./dependency_helper"
+
 module Langchain
   class << self
     attr_accessor :logger
@@ -19,6 +20,18 @@ module Langchain
   autoload :Loader, "langchain/loader"
   autoload :Data, "langchain/data"
 
+  module Agent
+    autoload :Base, "langchain/agent/base"
+    autoload :ChainOfThoughtAgent, "langchain/agent/chain_of_thought_agent/chain_of_thought_agent.rb"
+  end
+
+  module Tool
+    autoload :Base, "langchain/tool/base"
+    autoload :Calculator, "langchain/tool/calculator"
+    autoload :SerpApi, "langchain/tool/serp_api"
+    autoload :Wikipedia, "langchain/tool/wikipedia"
+  end
+
   module Processors
     autoload :Base, "langchain/processors/base"
     autoload :CSV, "langchain/processors/csv"
@@ -29,43 +42,35 @@ module Langchain
     autoload :PDF, "langchain/processors/pdf"
     autoload :Text, "langchain/processors/text"
   end
-end
-
-module Agent
-  autoload :Base, "agent/base"
-  autoload :ChainOfThoughtAgent, "agent/chain_of_thought_agent/chain_of_thought_agent.rb"
-end
 
-module Vectorsearch
-  autoload :Base, "vectorsearch/base"
-  autoload :Chroma, "vectorsearch/chroma"
-  autoload :Milvus, "vectorsearch/milvus"
-  autoload :Pinecone, "vectorsearch/pinecone"
-  autoload :Pgvector, "vectorsearch/pgvector"
-  autoload :Qdrant, "vectorsearch/qdrant"
-  autoload :Weaviate, "vectorsearch/weaviate"
-end
+  module Utils
+    autoload :TokenLengthValidator, "langchain/utils/token_length_validator"
+  end
 
-module LLM
-  autoload :Base, "llm/base"
-  autoload :Cohere, "llm/cohere"
-  autoload :GooglePalm, "llm/google_palm"
-  autoload :HuggingFace, "llm/hugging_face"
-  autoload :OpenAI, "llm/openai"
-  autoload :Replicate, "llm/replicate"
-end
+  module Vectorsearch
+    autoload :Base, "langchain/vectorsearch/base"
+    autoload :Chroma, "langchain/vectorsearch/chroma"
+    autoload :Milvus, "langchain/vectorsearch/milvus"
+    autoload :Pinecone, "langchain/vectorsearch/pinecone"
+    autoload :Pgvector, "langchain/vectorsearch/pgvector"
+    autoload :Qdrant, "langchain/vectorsearch/qdrant"
+    autoload :Weaviate, "langchain/vectorsearch/weaviate"
+  end
 
-module Prompt
-  require_relative "prompt/loading"
+  module LLM
+    autoload :Base, "langchain/llm/base"
+    autoload :Cohere, "langchain/llm/cohere"
+    autoload :GooglePalm, "langchain/llm/google_palm"
+    autoload :HuggingFace, "langchain/llm/hugging_face"
+    autoload :OpenAI, "langchain/llm/openai"
+    autoload :Replicate, "langchain/llm/replicate"
+  end
 
-  autoload :Base, "prompt/base"
-  autoload :PromptTemplate, "prompt/prompt_template"
-  autoload :FewShotPromptTemplate, "prompt/few_shot_prompt_template"
-end
+  module Prompt
+    require_relative "langchain/prompt/loading"
 
-module Tool
-  autoload :Base, "tool/base"
-  autoload :Calculator, "tool/calculator"
-  autoload :SerpApi, "tool/serp_api"
-  autoload :Wikipedia, "tool/wikipedia"
+    autoload :Base, "langchain/prompt/base"
+    autoload :PromptTemplate, "langchain/prompt/prompt_template"
+    autoload :FewShotPromptTemplate, "langchain/prompt/few_shot_prompt_template"
+  end
 end

data/lib/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Langchain
-  VERSION = "0.3.14"
+  VERSION = "0.4.0"
 end

metadata

@@ -1,15 +1,29 @@
 --- !ruby/object:Gem::Specification
 name: langchainrb
 version: !ruby/object:Gem::Version
-  version: 0.3.14
+  version: 0.4.0
 platform: ruby
 authors:
 - Andrei Bondarev
 autorequire:
 bindir: exe
 cert_chain: []
-date: 2023-05-28 00:00:00.000000000 Z
+date: 2023-06-01 00:00:00.000000000 Z
 dependencies:
+- !ruby/object:Gem::Dependency
+  name: tiktoken_ruby
+  requirement: !ruby/object:Gem::Requirement
+    requirements:
+    - - "~>"
+      - !ruby/object:Gem::Version
+        version: 0.0.5
+  type: :runtime
+  prerelease: false
+  version_requirements: !ruby/object:Gem::Requirement
+    requirements:
+    - - "~>"
+      - !ruby/object:Gem::Version
+        version: 0.0.5
 - !ruby/object:Gem::Dependency
   name: dotenv-rails
   requirement: !ruby/object:Gem::Requirement
@@ -182,14 +196,14 @@ dependencies:
   name: pgvector
   requirement: !ruby/object:Gem::Requirement
     requirements:
-    - - "<"
+    - - "~>"
       - !ruby/object:Gem::Version
         version: '0.2'
   type: :development
   prerelease: false
   version_requirements: !ruby/object:Gem::Requirement
     requirements:
-    - - "<"
+    - - "~>"
       - !ruby/object:Gem::Version
         version: '0.2'
 - !ruby/object:Gem::Dependency
@@ -290,7 +304,7 @@ dependencies:
     - - "~>"
       - !ruby/object:Gem::Version
         version: 1.17.0
-description: Build ML/AI-powered applications with Ruby's LangChain
+description: Build LLM-backed Ruby applications with Ruby's LangChain
 email:
 - andrei.bondarev13@gmail.com
 executables: []
@@ -311,12 +325,19 @@ files:
 - examples/store_and_query_with_pinecone.rb
 - examples/store_and_query_with_qdrant.rb
 - examples/store_and_query_with_weaviate.rb
-- lib/agent/base.rb
-- lib/agent/chain_of_thought_agent/chain_of_thought_agent.rb
-- lib/agent/chain_of_thought_agent/chain_of_thought_agent_prompt.json
 - lib/dependency_helper.rb
 - lib/langchain.rb
+- lib/langchain/agent/base.rb
+- lib/langchain/agent/chain_of_thought_agent/chain_of_thought_agent.rb
+- lib/langchain/agent/chain_of_thought_agent/chain_of_thought_agent_prompt.json
 - lib/langchain/data.rb
+- lib/langchain/llm/base.rb
+- lib/langchain/llm/cohere.rb
+- lib/langchain/llm/google_palm.rb
+- lib/langchain/llm/hugging_face.rb
+- lib/langchain/llm/openai.rb
+- lib/langchain/llm/prompts/summarize_template.json
+- lib/langchain/llm/replicate.rb
 - lib/langchain/loader.rb
 - lib/langchain/processors/base.rb
 - lib/langchain/processors/csv.rb
@@ -326,29 +347,23 @@ files:
 - lib/langchain/processors/jsonl.rb
 - lib/langchain/processors/pdf.rb
 - lib/langchain/processors/text.rb
+- lib/langchain/prompt/base.rb
+- lib/langchain/prompt/few_shot_prompt_template.rb
+- lib/langchain/prompt/loading.rb
+- lib/langchain/prompt/prompt_template.rb
+- lib/langchain/tool/base.rb
+- lib/langchain/tool/calculator.rb
+- lib/langchain/tool/serp_api.rb
+- lib/langchain/tool/wikipedia.rb
+- lib/langchain/utils/token_length_validator.rb
+- lib/langchain/vectorsearch/base.rb
+- lib/langchain/vectorsearch/chroma.rb
+- lib/langchain/vectorsearch/milvus.rb
+- lib/langchain/vectorsearch/pgvector.rb
+- lib/langchain/vectorsearch/pinecone.rb
+- lib/langchain/vectorsearch/qdrant.rb
+- lib/langchain/vectorsearch/weaviate.rb
 - lib/langchainrb.rb
-- lib/llm/base.rb
-- lib/llm/cohere.rb
-- lib/llm/google_palm.rb
-- lib/llm/hugging_face.rb
-- lib/llm/openai.rb
-- lib/llm/prompts/summarize_template.json
-- lib/llm/replicate.rb
-- lib/prompt/base.rb
-- lib/prompt/few_shot_prompt_template.rb
-- lib/prompt/loading.rb
-- lib/prompt/prompt_template.rb
-- lib/tool/base.rb
-- lib/tool/calculator.rb
-- lib/tool/serp_api.rb
-- lib/tool/wikipedia.rb
-- lib/vectorsearch/base.rb
-- lib/vectorsearch/chroma.rb
-- lib/vectorsearch/milvus.rb
-- lib/vectorsearch/pgvector.rb
-- lib/vectorsearch/pinecone.rb
-- lib/vectorsearch/qdrant.rb
-- lib/vectorsearch/weaviate.rb
 - lib/version.rb
 - sig/langchain.rbs
 homepage: https://rubygems.org/gems/langchainrb
@@ -366,7 +381,7 @@ required_ruby_version: !ruby/object:Gem::Requirement
   requirements:
   - - ">="
     - !ruby/object:Gem::Version
-      version: 2.6.0
+      version: 3.0.0
 required_rubygems_version: !ruby/object:Gem::Requirement
   requirements:
   - - ">="
@@ -376,5 +391,5 @@ requirements: []
 rubygems_version: 3.2.3
 signing_key:
 specification_version: 4
-summary: Build ML/AI-powered applications with Ruby's LangChain
+summary: Build LLM-backed Ruby applications with Ruby's LangChain
 test_files: []