Ruby版LangChain「LangChain.rb」入門

gem 'langchainrb'

gem install langchainrb

require 'langchain'

gem "ruby-openai", "~> 4.0.0"

export OPENAI_API_KEY="..."

require 'langchain'

openai = Langchain::LLM::OpenAI.new(api_key: ENV['OPENAI_API_KEY'])

openai.complete(prompt: 'What is the meaning of life?')

openai.embed(text: 'foo bar')

# プロンプトの補完
prompt = 'When was Ruby first released?'
openai.chat(prompt: prompt)

# メッセージの補完
messages = [
  {
    role: 'system',
    content: 'You are RubyGPT, a helpful chat bot for helping people learn Ruby'
  },
  {
    role: 'user',
    content: 'When was Ruby first released?'
  }
]
openai.chat(messages: messages)

require 'langchain'

# LLMが使うことができる関数を羅列する
# この例では、引数locationを受け取って天気を返す関数get_current_weatherが事前に定義されている想定
functions = [
  {
    name: 'get_current_weather',
    description: 'Get the current weather in a given location',
    parameters: {
      type: :object,
      properties: {
        location: {
          type: :string,
          description: 'The city and state, e.g. San Francisco, CA'
        },
        unit: {
          type: 'string',
          enum: %w[celsius fahrenheit]
        }
      },
      required: ['location']
    }
  }
]

# OpenAIクライアントの初期化
openai = Langchain::LLM::OpenAI.new(
  api_key: ENV['OPENAI_API_KEY'],
  default_options: {
    chat_completion_model_name: 'gpt-3.5-turbo-16k'
  }
)

chat = Langchain::Conversation.new(llm: openai)

# 会話のコンテキストを設定する。通常、モデルのペルソナを設定するために使用される
chat.set_context('You are the climate bot')
# 関数を設定する
chat.set_functions(functions)

# LLMから回答を得る
user_message = "what's the weather in NYC?"
response = chat.message(user_message)
response_message = response.dig('choices', 0, 'message')

# LLMの回答から関数名と引数を取得する
function_name = response_message.dig('function_call', 'name')
function_args = response_message.dig('function_call', 'arguments')

# 取得した関数名と引数を使って、関数を実行する
...

require 'langchain'

# テンプレートの作成
prompt = Langchain::Prompt::PromptTemplate.new(
  template: 'Tell me a {adjective} joke about {content}.',
  input_variables: %w[adjective content]
)
# テンプレートを使ってプロンプトを作成
prompt_text = prompt.format(adjective: 'funny', content: 'chickens') # "Tell me a funny joke about chickens."

# LLMにプロンプトを送信
openai = Langchain::LLM::OpenAI.new(api_key: ENV['OPENAI_API_KEY'])
openai.complete(prompt: prompt_text)

require 'langchain'

# テンプレートの作成
prompt = Langchain::Prompt::FewShotPromptTemplate.new(
  prefix: 'Write antonyms for the following words.',
  suffix: "Input: {adjective}\nOutput:",
  example_prompt: Langchain::Prompt::PromptTemplate.new(
    input_variables: %w[input output],
    template: "Input: {input}\nOutput: {output}"
  ),
  examples: [
    { "input": 'happy', "output": 'sad' },
    { "input": 'tall', "output": 'short' }
  ],
  input_variables: ['adjective']
)

# テンプレートを使ってプロンプトを作成
prompt_text = prompt.format(adjective: 'good')

# LLMにプロンプトを送信
openai = Langchain::LLM::OpenAI.new(api_key: ENV['OPENAI_API_KEY'])
openai.complete(prompt: prompt_text)

require 'langchain'

prompt = Langchain::Prompt::PromptTemplate.new(
  template: 'Tell me a {adjective} joke about {content}.',
  input_variables: %w[adjective content]
)

# テンプレートの保存
prompt.save(file_path: 'prompt.json')

# テンプレートの読み込み
prompt = Langchain::Prompt.load_from_path(file_path: 'prompt.json')

require 'langchain'

# 以下のjson_schemaに従った回答をLLMに生成させるpromptを作成
json_schema = {
  type: 'object',
  properties: {
    name: {
      type: 'string',
      description: 'Persons name'
    },
    age: {
      type: 'number',
      description: 'Persons age'
    },
    interests: {
      type: 'array',
      items: {
        type: 'object',
        properties: {
          interest: {
            type: 'string',
            description: 'A topic of interest'
          },
          levelOfInterest: {
            type: 'number',
            description: 'A value between 0 and 100 of how interested the person is in this interest'
          }
        },
        required: %w[interest levelOfInterest],
        additionalProperties: false
      },
      minItems: 1,
      maxItems: 3,
      description: "A list of the person's interests"
    }
  },
  required: %w[name age interests],
  additionalProperties: false
}

parser = Langchain::OutputParsers::StructuredOutputParser.from_json_schema(json_schema)
prompt = Langchain::Prompt::PromptTemplate.new(
  template: "Generate details of a fictional character.\n{format_instructions}\nCharacter description: {description}", input_variables: %w[
    description format_instructions
  ]
)

prompt_text = prompt.format(description: 'Korean chemistry student',
                            format_instructions: parser.get_format_instructions)

# LLMにpromptを与え、回答を得る
llm = Langchain::LLM::OpenAI.new(api_key: ENV['OPENAI_API_KEY'])
llm_response = llm.chat(prompt: prompt_text)
# LLMの回答をパースする。パース結果はjson_schemaに従ったHashとなる。
parser.parse(llm_response)

require 'langchain'

# 以下のjson_schemaに従った回答をLLMに生成させるpromptを作成
json_schema = ...

# 省略

parser = Langchain::OutputParsers::StructuredOutputParser.from_json_schema(json_schema)

# 省略

# LLMにpromptを与え、回答を得る
llm = Langchain::LLM::OpenAI.new(api_key: ENV['OPENAI_API_KEY'])
llm_response = llm.chat(prompt: prompt_text)

# OutputFixingParserを使って、LLMのレスポンスのパースが失敗したときに、修正された回答を求める
fix_parser = Langchain::OutputParsers::OutputFixingParser.from_llm(
  llm: llm,
  parser: parser
)
fix_parser.parse(llm_response)

gem 'weaviate-ruby', '~> 0.8.3'

export WEAVIATE_URL="..."
export WEAVIATE_API_KEY="..."

require 'langchain'

# Weaviateクライアントの初期化
# @param url [String] The URL of the Weaviate instance
# @param api_key [String] The API key to use
# @param index_name [String] The capitalized name of the index to use
# @param llm [Object] The LLM client to use
client = Langchain::Vectorsearch::Weaviate.new(
  url: ENV['WEAVIATE_URL'],
  api_key: ENV['WEAVIATE_API_KEY'],
  index_name: 'SampleIndex',
  llm: Langchain::LLM::OpenAI.new(api_key: ENV['OPENAI_API_KEY'])
)

# デフォルトのスキーマの作成
# propertyとして__id（string型）とcontent（text型）を持つindexが作成される
client.create_default_schema

# スキーマの取得
client.get_default_schema

# ベクトル検索データベースにテキストを保存
client.add_texts(
  texts: [
    'Begin by preheating your oven to 375°F (190°C). Prepare four boneless, skinless chicken breasts by cutting a pocket into the side of each breast, being careful not to cut all the way through. Season the chicken with salt and pepper to taste. In a large skillet, melt 2 tablespoons of unsalted butter over medium heat. Add 1 small diced onion and 2 minced garlic cloves, and cook until softened, about 3-4 minutes. Add 8 ounces of fresh spinach and cook until wilted, about 3 minutes. Remove the skillet from heat and let the mixture cool slightly.',
    'In a bowl, combine the spinach mixture with 4 ounces of softened cream cheese, 1/4 cup of grated Parmesan cheese, 1/4 cup of shredded mozzarella cheese, and 1/4 teaspoon of red pepper flakes. Mix until well combined. Stuff each chicken breast pocket with an equal amount of the spinach mixture. Seal the pocket with a toothpick if necessary. In the same skillet, heat 1 tablespoon of olive oil over medium-high heat. Add the stuffed chicken breasts and sear on each side for 3-4 minutes, or until golden brown.'
  ]
)

# クエリー文字列による検索
query = 'oven'
client.similarity_search(query: query, k: 3) # k: 返す結果の数

# ベクトルによる検索
openai = Langchain::LLM::OpenAI.new(api_key: ENV['OPENAI_API_KEY'])
embedding = openai.embed(text: 'oven')
client.similarity_search_by_vector(embedding: embedding, k: 3)

# question answering
question = 'How many chicken breasts do you need?'
client.ask(question: question)

# デフォルトのスキーマの削除
client.destroy_default_schema

% bin/rails g model Recipe \
> title:string \
> description:string

% bin/rails db:create
% bin/rails db:migrate

# app/models/recipe.rb
class Recipe < ActiveRecord::Base
  include Langchain::ActiveRecord::Hooks

  vectorsearch provider: Langchain::Vectorsearch::Weaviate.new(
    api_key: ENV['WEAVIATE_API_KEY'],
    url: ENV['WEAVIATE_URL'],
    index_name: 'Recipes',
    llm: Langchain::LLM::OpenAI.new(api_key: ENV['OPENAI_API_KEY'])
  )

  after_save :upsert_to_vectorsearch

  # Overwriting how the model is serialized before it's indexed
  def as_vector
    [
      "Title: #{title}",
      "Description: #{description}"
    ]
      .compact
      .join("\n")
  end
end

% bin/rails c

irb(main):001:1* Recipe.create(
irb(main):002:1*   title: "curry",
irb(main):003:1*   description: "how to make delicious curry",
irb(main):004:0> )
...
irb(main):005:1* Recipe.create(
irb(main):006:1*   title: "soup",
irb(main):007:1*   description: "how to make wonderfult soup",
irb(main):008:0> )

irb(main):012:0> Recipe.similarity_search("soup", k: 1)
  Recipe Load (0.7ms)  SELECT "recipes".* FROM "recipes" WHERE "recipes"."id" = ?  [["id", 2]]
=>
[#<Recipe:0x00000001144348a0
  id: 2,
  title: "soup",
  description: "how to make wonderfult soup",
  created_at: Tue, 15 Aug 2023 23:52:11.361501000 UTC +00:00,
  updated_at: Tue, 15 Aug 2023 23:52:11.361501000 UTC +00:00>]

gem 'ruby-openai', '~> 4.0.0'
gem 'google_search_results'
gem 'eqn'

export SERPAPI_API_KEY=="..."

require 'langchain'

# ログレベルの設定
# Agentの行動の過程を確認したい場合は下記1行のコメントアウトを解除する
# Langchain.logger.level = :info

# toolの作成
search_tool = Langchain::Tool::GoogleSearch.new(api_key: ENV['SERPAPI_API_KEY'])
calculator = Langchain::Tool::Calculator.new

# agentの作成。作成したtoolを渡す。
openai = Langchain::LLM::OpenAI.new(api_key: ENV['OPENAI_API_KEY'])
agent = Langchain::Agent::ReActAgent.new(llm: openai, tools: [search_tool, calculator])

agent.run(question: 'How many full soccer fields would be needed to cover the distance between NYC and DC in a straight line?')

Langchain.logger.level = :info