langchainrb 0.3.2 → 0.3.4

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: e751a38f6a248db9aabfac3f1b9bc547f61304dd366be999005941045a5adcd6
-  data.tar.gz: 56db69c0a578bdf198bfc12cb53e92f0e1d15654a1dfa29715d5d7833550da33
+  metadata.gz: a5a782dd2282ab5dd4aed3f1d0e421a4f9b227fa4c5450ed27f2f98a86af74f4
+  data.tar.gz: b47bb5d6789d7abb81f56ee1beb0b52323184f578475aec3e92fcc19b4a1314a
 SHA512:
-  metadata.gz: bdd3060863a967b48a6123ea379d9a98632730a726b8da9924bb0061511b8bf88fa05f36c49ac4b8f136472e9aa204c4bd09929d31e0f2c0700ad16361a1a8cf
-  data.tar.gz: 88f04804d10f51d639b8643bd683b8a420920fc4ad2d26ef6c45f4fac034ec3f8b54787de51c4799f4a074029ad5649efca6e950b29fef62b96ce9452e5ec8cc
+  metadata.gz: 4e918b49b2b04a0e7009db732a5b24ab080a0d8d6b4c3be4084aa3a1492ddd6c1627d467b8bff4b34e1e5160b71622a75772ef92658dce2589b9542b8b0a8137
+  data.tar.gz: 3b465f1a05e614d64d416582aeaf7d998bf51422705331b220e08b00774d4beed4dcbc4e2a71b81bad87cddfe8c8f4c8db421d650e7161f16b26a1259922f71c

data/CHANGELOG.md

@@ -1,5 +1,14 @@
 ## [Unreleased]
 
+## [0.3.4] - 2023-05-16
+- LLMs
+  - Introducing support for HuggingFace
+
+## [0.3.3] - 2023-05-16
+- Dependencies are now optionally loaded and required at runtime
+- Start using `standardrb` for linting
+- Use the Ruby logger
+
 ## [0.3.2] - 2023-05-15
 - Agents
   - Fix Chain of Thought prompt loader

data/Gemfile

@@ -8,3 +8,5 @@ gemspec
 gem "rake", "~> 13.0"
 
 gem "rspec", "~> 3.0"
+
+gem "standardrb"

data/Gemfile.lock

@@ -1,16 +1,7 @@
 PATH
   remote: .
   specs:
-    langchainrb (0.3.2)
-      cohere-ruby (~> 0.9.3)
-      eqn (~> 1.6.5)
-      google_search_results (~> 2.0.0)
-      milvus (~> 0.9.0)
-      pinecone (~> 0.1.6)
-      qdrant-ruby (~> 0.9.0)
-      ruby-openai (~> 4.0.0)
-      weaviate-ruby (~> 0.8.0)
-      wikipedia-client (~> 1.17.0)
+    langchainrb (0.3.4)
 
 GEM
   remote: https://rubygems.org/
@@ -35,6 +26,7 @@ GEM
       tzinfo (~> 2.0)
     addressable (2.8.4)
       public_suffix (>= 2.0.2, < 6.0)
+    ast (2.4.2)
     builder (3.2.4)
     byebug (11.1.3)
     coderay (1.1.3)
@@ -125,9 +117,14 @@ GEM
     httparty (0.21.0)
       mini_mime (>= 1.0.0)
       multi_xml (>= 0.5.2)
+    hugging-face (0.3.2)
+      faraday (~> 1.0)
     i18n (1.13.0)
       concurrent-ruby (~> 1.0)
     ice_nine (0.11.2)
+    json (2.6.3)
+    language_server-protocol (3.17.0.3)
+    lint_roller (1.0.0)
     loofah (2.21.1)
       crass (~> 1.0.2)
       nokogiri (>= 1.5.9)
@@ -138,10 +135,15 @@ GEM
     minitest (5.18.0)
     multi_xml (0.6.0)
     multipart-post (2.3.0)
+    nokogiri (1.14.3-arm64-darwin)
+      racc (~> 1.4)
     nokogiri (1.14.3-x86_64-darwin)
       racc (~> 1.4)
     nokogiri (1.14.3-x86_64-linux)
       racc (~> 1.4)
+    parallel (1.23.0)
+    parser (3.2.2.1)
+      ast (~> 2.4.1)
     pinecone (0.1.71)
       dry-struct (~> 1.6.0)
       dry-validation (~> 1.10.0)
@@ -173,7 +175,10 @@ GEM
       rake (>= 12.2)
       thor (~> 1.0)
       zeitwerk (~> 2.5)
+    rainbow (3.1.1)
     rake (13.0.6)
+    regexp_parser (2.8.0)
+    rexml (3.2.5)
     rspec (3.12.0)
       rspec-core (~> 3.12.0)
       rspec-expectations (~> 3.12.0)
@@ -187,15 +192,45 @@ GEM
       diff-lcs (>= 1.2.0, < 2.0)
       rspec-support (~> 3.12.0)
     rspec-support (3.12.0)
+    rubocop (1.50.2)
+      json (~> 2.3)
+      parallel (~> 1.10)
+      parser (>= 3.2.0.0)
+      rainbow (>= 2.2.2, < 4.0)
+      regexp_parser (>= 1.8, < 3.0)
+      rexml (>= 3.2.5, < 4.0)
+      rubocop-ast (>= 1.28.0, < 2.0)
+      ruby-progressbar (~> 1.7)
+      unicode-display_width (>= 2.4.0, < 3.0)
+    rubocop-ast (1.28.1)
+      parser (>= 3.2.1.0)
+    rubocop-performance (1.16.0)
+      rubocop (>= 1.7.0, < 2.0)
+      rubocop-ast (>= 0.4.0)
     ruby-openai (4.0.0)
       faraday (>= 1)
       faraday-multipart (>= 1)
+    ruby-progressbar (1.13.0)
     ruby2_keywords (0.0.5)
+    standard (1.28.2)
+      language_server-protocol (~> 3.17.0.2)
+      lint_roller (~> 1.0)
+      rubocop (~> 1.50.2)
+      standard-custom (~> 1.0.0)
+      standard-performance (~> 1.0.1)
+    standard-custom (1.0.0)
+      lint_roller (~> 1.0)
+    standard-performance (1.0.1)
+      lint_roller (~> 1.0)
+      rubocop-performance (~> 1.16.0)
+    standardrb (1.0.1)
+      standard
     thor (1.2.1)
     treetop (1.6.12)
       polyglot (~> 0.3)
     tzinfo (2.0.6)
       concurrent-ruby (~> 1.0)
+    unicode-display_width (2.4.2)
     weaviate-ruby (0.8.1)
       faraday (~> 1)
       faraday_middleware (~> 1)
@@ -205,15 +240,27 @@ GEM
     zeitwerk (2.6.8)
 
 PLATFORMS
+  arm64-darwin-22
   x86_64-darwin-19
   x86_64-linux
 
 DEPENDENCIES
+  cohere-ruby (~> 0.9.3)
   dotenv-rails (~> 2.7.6)
+  eqn (~> 1.6.5)
+  google_search_results (~> 2.0.0)
+  hugging-face (~> 0.3.2)
   langchainrb!
+  milvus (~> 0.9.0)
+  pinecone (~> 0.1.6)
   pry-byebug (~> 3.10.0)
+  qdrant-ruby (~> 0.9.0)
   rake (~> 13.0)
   rspec (~> 3.0)
+  ruby-openai (~> 4.0.0)
+  standardrb
+  weaviate-ruby (~> 0.8.0)
+  wikipedia-client (~> 1.17.0)
 
 BUNDLED WITH
    2.4.0

data/README.md

@@ -1,5 +1,5 @@
 🦜️🔗 LangChain.rb
---- 
+---
 ⚡ Building applications with LLMs through composability ⚡
 
 👨‍💻👩‍💻 CURRENTLY SEEKING PEOPLE TO FORM THE CORE GROUP OF MAINTAINERS WITH
@@ -39,6 +39,8 @@ require "langchain"
 
 Choose the LLM provider you'll be using (OpenAI or Cohere) and retrieve the API key.
 
+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(
@@ -49,9 +51,9 @@ client = Vectorsearch::Weaviate.new(
 )
 
 # You can instantiate any other supported vector search database:
-client = Vectorsearch::Milvus.new(...)
-client = Vectorsearch::Qdrant.new(...)
-client = Vectorsearch::Pinecone.new(...)
+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"`
 ```
 
 ```ruby
@@ -92,6 +94,8 @@ client.ask(
 
 ### Using Standalone LLMs 🗣️
 
+Add `gem "ruby-openai", "~> 4.0.0"` to your Gemfile.
+
 #### OpenAI
 ```ruby
 openai = LLM::OpenAI.new(api_key: ENV["OPENAI_API_KEY"])
@@ -104,6 +108,8 @@ openai.complete(prompt: "What is the meaning of life?")
 ```
 
 #### Cohere
+Add `gem "cohere-ruby", "~> 0.9.3"` to your Gemfile.
+
 ```ruby
 cohere = LLM::Cohere.new(api_key: ENV["COHERE_API_KEY"])
 ```
@@ -114,6 +120,9 @@ cohere.embed(text: "foo bar")
 cohere.complete(prompt: "What is the meaning of life?")
 ```
 
+#### HuggingFace
+Add `gem "hugging-face", "~> 0.3.2"` to your Gemfile.
+
 ### Using Prompts 📋
 
 #### Prompt Templates
@@ -204,6 +213,8 @@ 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
+
 ```ruby
 agent = Agent::ChainOfThoughtAgent.new(llm: :openai, llm_api_key: ENV["OPENAI_API_KEY"], tools: ['search', 'calculator'])
 
@@ -211,7 +222,7 @@ agent.tools
 # => ["search", "calculator"]
 ```
 ```ruby
-agent.run(question: "How many full soccer fields would be needed to cover the distance between NYC and DC in a straight line?", logging: true)
+agent.run(question: "How many full soccer fields would be needed to cover the distance between NYC and DC in a straight line?")
 #=> "Approximately 2,945 soccer fields would be needed to cover the distance between NYC and DC in a straight line."
 ```
 
@@ -228,6 +239,16 @@ agent.run(question: "How many full soccer fields would be needed to cover the di
 | "search" | A wrapper around Google Search | `ENV["SERPAPI_API_KEY"]` (https://serpapi.com/manage-api-key)
 | "wikipedia" | Calls Wikipedia API to retrieve the summary | |
 
+
+## Logging
+
+LangChain.rb uses standard logging mechanisms and defaults to `:debug` level. Most messages are at info level, but we will add debug or warn statements as needed.
+To show all log messages:
+
+```ruby
+Langchain.logger.level = :info
+```
+
 ## Development
 
 After checking out the repo, run `bin/setup` to install dependencies. Then, run `rake spec` to run the tests. You can also run `bin/console` for an interactive prompt that will allow you to experiment.

data/Rakefile

@@ -2,7 +2,12 @@
 
 require "bundler/gem_tasks"
 require "rspec/core/rake_task"
+require "standard/rake"
 
 RSpec::Core::RakeTask.new(:spec)
 
 task default: :spec
+
+Rake::Task["spec"].enhance do
+  Rake::Task["standard:fix"].invoke
+end

data/examples/create_and_manage_few_shot_prompt_templates.rb

@@ -9,10 +9,10 @@ prompt = Prompt::FewShotPromptTemplate.new(
     template: "Input: {input}\nOutput: {output}"
   ),
   examples: [
-    { "input": "happy", "output": "sad" },
-    { "input": "tall", "output": "short" }
+    {input: "happy", output: "sad"},
+    {input: "tall", output: "short"}
   ],
-   input_variables: ["adjective"]
+  input_variables: ["adjective"]
 )
 
 prompt.format(adjective: "good")

data/examples/store_and_query_with_pinecone.rb

@@ -1,5 +1,8 @@
 require "langchain"
 
+# gem install pinecone
+# or add `gem "pinecone"` to your Gemfile
+
 # Instantiate the Qdrant client
 pinecone = Vectorsearch::Pinecone.new(
   environment: ENV["PINECONE_ENVIRONMENT"],
@@ -35,9 +38,9 @@ pinecone.ask(
 )
 
 # Generate your an embedding and search by it
-openai = LLM::OpenAI.new(api_key: ENV['OPENAI_API_KEY'])
+openai = LLM::OpenAI.new(api_key: ENV["OPENAI_API_KEY"])
 embedding = openai.embed(text: "veggie")
 
 pinecone.similarity_search_by_vector(
   embedding: embedding
-)
+)

data/examples/store_and_query_with_qdrant.rb

@@ -1,5 +1,8 @@
 require "langchain"
 
+# gem install qdrant-ruby
+# or add `gem "qdrant-ruby"` to your Gemfile
+
 # Instantiate the Qdrant client
 qdrant = Vectorsearch::Qdrant.new(
   url: ENV["QDRANT_URL"],
@@ -9,7 +12,6 @@ qdrant = Vectorsearch::Qdrant.new(
   llm_api_key: ENV["COHERE_API_KEY"]
 )
 
-
 # Create the default schema.
 qdrant.create_default_schema
 
@@ -33,4 +35,4 @@ qdrant.similarity_search(
 # Interact with your index through Q&A
 qdrant.ask(
   question: "What is the best recipe for chicken?"
-)
+)

data/examples/store_and_query_with_weaviate.rb

@@ -1,5 +1,8 @@
 require "langchain"
 
+# gem install weaviate-ruby
+# or add `gem "weaviate-ruby"` to your Gemfile
+
 # Instantiate the Weaviate client
 weaviate = Vectorsearch::Weaviate.new(
   url: ENV["WEAVIATE_URL"],
@@ -27,4 +30,4 @@ weaviate.add_texts(
 weaviate.similarity_search(
   query: "chicken",
   k: 1
-)
+)

data/lib/agent/chain_of_thought_agent/chain_of_thought_agent.rb

@@ -5,7 +5,7 @@ module Agent
     attr_reader :llm, :llm_api_key, :llm_client, :tools
 
     # Initializes the Agent
-    # 
+    #
     # @param llm [Symbol] The LLM to use
     # @param llm_api_key [String] The API key for the LLM
     # @param tools [Array] The tools to use
@@ -22,7 +22,7 @@ module Agent
     end
 
     # Validate tools when they're re-assigned
-    # 
+    #
     # @param value [Array] The tools to use
     # @return [Array] The tools that will be used
     def tools=(value)
@@ -31,11 +31,10 @@ module Agent
     end
 
     # Run the Agent!
-    # 
+    #
     # @param question [String] The question to ask
-    # @param logging [Boolean] Whether or not to log the Agent's actions
     # @return [String] The answer to the question
-    def run(question:, logging: false)
+    def run(question:)
       question = question.strip
       prompt = create_prompt(
         question: question,
@@ -43,7 +42,7 @@ module Agent
       )
 
       loop do
-        puts("Agent: Passing the prompt to the #{llm} LLM") if logging
+        Langchain.logger.info("Agent: Passing the prompt to the #{llm} LLM")
         response = llm_client.generate_completion(
           prompt: prompt,
           stop_sequences: ["Observation:"],
@@ -51,16 +50,16 @@ module Agent
         )
 
         # Append the response to the prompt
-        prompt += response;
-    
+        prompt += response
+
         # Find the requested action in the "Action: search" format
         action = response.match(/Action: (.*)/)&.send(:[], -1)
-        
+
         if action
           # Find the input to the action in the "Action Input: [action_input]" format
           action_input = response.match(/Action Input: "?(.*)"?/)&.send(:[], -1)
 
-          puts("Agent: Using the \"#{action}\" Tool with \"#{action_input}\"") if logging
+          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
@@ -68,10 +67,10 @@ module Agent
             .execute(input: action_input)
 
           # Append the Observation to the prompt
-          if prompt.end_with?("Observation:")
-            prompt += " #{result}\nThought:"
+          prompt += if prompt.end_with?("Observation:")
+            " #{result}\nThought:"
           else
-            prompt += "\nObservation: #{result}\nThought:"
+            "\nObservation: #{result}\nThought:"
           end
         else
           # Return the final answer
@@ -92,7 +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}: #{Tool.const_get(Tool::Base::TOOLS[tool]).const_get(:DESCRIPTION)}"
         end.join("\n")
       )
     end

data/lib/dependency_helper.rb

@@ -0,0 +1,18 @@
+# frozen_string_literal: true
+
+def depends_on(gem_name)
+  gem(gem_name) # require the gem
+
+  return(true) unless defined?(Bundler) # If we're in a non-bundler environment, we're no longer able to determine if we'll meet requirements
+
+  gem_version = Gem.loaded_specs[gem_name].version
+  gem_requirement = Bundler.load.dependencies.find { |g| g.name == gem_name }.requirement
+
+  if !gem_requirement.satisfied_by?(gem_version)
+    raise "The #{gem_name} gem is installed, but version #{gem_requirement} is required. You have #{gem_version}."
+  end
+
+  true
+rescue LoadError
+  raise LoadError, "Could not load #{gem_name}. Please ensure that the #{gem_name} gem is installed."
+end

data/lib/langchain.rb

@@ -1,6 +1,8 @@
 # frozen_string_literal: true
 
 require_relative "./version"
+require_relative "./dependency_helper"
+require_relative "./logging"
 
 module Agent
   autoload :Base, "agent/base"
@@ -18,6 +20,7 @@ end
 module LLM
   autoload :Base, "llm/base"
   autoload :Cohere, "llm/cohere"
+  autoload :HuggingFace, "llm/hugging_face"
   autoload :OpenAI, "llm/openai"
 end
 

data/lib/llm/base.rb

@@ -12,16 +12,16 @@ module LLM
     }.freeze
 
     def default_dimension
-      self.class.const_get("DEFAULTS").dig(:dimension)
+      self.class.const_get(:DEFAULTS).dig(:dimension)
     end
 
     # Ensure that the LLM value passed in is supported
     # @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.keys.include?(llm)
+      unless LLM::Base::LLMS.key?(llm)
         raise ArgumentError, "LLM must be one of #{LLM::Base::LLMS.keys}"
       end
     end
   end
-end
+end

data/lib/llm/cohere.rb

@@ -1,10 +1,7 @@
 # frozen_string_literal: true
 
-require "cohere"
-
 module LLM
   class Cohere < Base
-
     DEFAULTS = {
       temperature: 0.0,
       completion_model_name: "base",
@@ -13,6 +10,9 @@ module LLM
     }.freeze
 
     def initialize(api_key:)
+      depends_on "cohere-ruby"
+      require "cohere"
+
       @client = ::Cohere::Client.new(api_key: api_key)
     end
 
@@ -22,7 +22,7 @@ module LLM
     def embed(text:)
       response = client.embed(
         texts: [text],
-        model: DEFAULTS[:embeddings_model_name],
+        model: DEFAULTS[:embeddings_model_name]
       )
       response.dig("embeddings").first
     end
@@ -50,4 +50,4 @@ module LLM
     alias_method :generate_completion, :complete
     alias_method :generate_embedding, :embed
   end
-end
+end

data/lib/llm/hugging_face.rb

@@ -0,0 +1,32 @@
+# frozen_string_literal: true
+
+module 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
+    DEFAULTS = {
+      embeddings_model_name: "sentence-transformers/all-MiniLM-L6-v2"      
+    }.freeze
+    
+    # 
+    # Intialize the HuggingFace LLM
+    # @param api_key [String] The API key to use
+    # 
+    def initialize(api_key:)
+      depends_on "hugging-face"
+      require "hugging_face"
+
+      @client = ::HuggingFace::InferenceApi.new(api_token: api_key)
+    end
+
+    # Generate an embedding for a given text
+    # @param text [String] The text to embed
+    # @return [Array] The embedding
+    def embed(text:)
+      response = client.embedding(
+        input: text,
+        model: DEFAULTS[:embeddings_model_name]
+      )
+    end
+  end
+end

data/lib/llm/openai.rb

@@ -1,10 +1,7 @@
 # frozen_string_literal: true
 
-require "openai"
-
 module LLM
   class OpenAI < Base
-
     DEFAULTS = {
       temperature: 0.0,
       completion_model_name: "text-davinci-003",
@@ -13,6 +10,9 @@ module LLM
     }.freeze
 
     def initialize(api_key:)
+      depends_on "ruby-openai"
+      require "openai"
+
       # TODO: Add support to pass `organization_id:`
       @client = ::OpenAI::Client.new(access_token: api_key)
     end
@@ -53,4 +53,4 @@ module LLM
     alias_method :generate_completion, :complete
     alias_method :generate_embedding, :embed
   end
-end
+end

data/lib/logging.rb

@@ -0,0 +1,13 @@
+# frozen_string_literal: true
+
+require "logger"
+
+module Langchain
+  def self.logger
+    @@logger ||= Logger.new($stdout, level: :warn, formatter: ->(severity, datetime, progname, msg) { "[LangChain.rb] #{msg}\n" })
+  end
+
+  def self.logger=(instance)
+    @@logger = instance
+  end
+end

data/lib/prompt/base.rb

@@ -1,6 +1,7 @@
 # frozen_string_literal: true
 
-require 'strscan'
+require "strscan"
+require "json"
 
 module Prompt
   class Base
@@ -52,14 +53,12 @@ module Prompt
       FileUtils.mkdir_p(directory_path) unless directory_path.directory?
 
       if save_path.extname == ".json"
-        File.open(file_path, "w") { |f| f.write(to_h.to_json) }
+        File.write(file_path, to_h.to_json)
       else
         raise ArgumentError, "#{file_path} must be json"
       end
     end
 
-    private
-
     #
     # Extracts variables from a template string.
     #

data/lib/prompt/loading.rb

@@ -1,7 +1,7 @@
 # frozen_string_literal: true
 
-require 'strscan'
-require 'pathname'
+require "strscan"
+require "pathname"
 
 module Prompt
   TYPE_TO_LOADER = {
@@ -70,7 +70,7 @@ module Prompt
     def load_from_config(config)
       # If `_type` key is not present in the configuration hash, add it with a default value of `prompt`
       unless config.key?("_type")
-        puts "[WARN] No `_type` key found, defaulting to `prompt`"
+        Langchain.logger.warn "No `_type` key found, defaulting to `prompt`"
         config["_type"] = "prompt"
       end