langchainrb 0.3.11 → 0.3.12

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: ff7f99d961b09e827df297ddb3144821c9103fd40eabb32688ca92588a73415c
-  data.tar.gz: bb83eaa99055cf45cceaccb18a84e9fd4ee3ea4a93a6a0c66e04ede43e5d4bc0
+  metadata.gz: 974f0a2b8ce3fe42144016bd740ee9d4f7e597834319cc92fbf1d50bd1f4468e
+  data.tar.gz: 3686a42c37eb117e6d7485ef4f7777c0f12968bb9cdcc3a30c7721c86c0a4325
 SHA512:
-  metadata.gz: 40e5362520220d3ffc1b4c29c3e430b051de334c2f281d9cb7d7549a93be40b26b379dbd35d0c91ccb5010c1a495a653e31768f1b7a95bc087059d59339fd1a7
-  data.tar.gz: 04f24944b590ee8b577419a95718ad6796bc4cdc34d52cf05e287806912c05ba8ace22e07181a8537892124881b65a8a221e1c75d8c6245231dd5660c6b4308c
+  metadata.gz: a61f9b36d9d19eb6cf87af18c7fb40f55d39771257d08a6af2ec3384988419dfb158ffa8fc81c3769c0149f1ffa8b03200366bbea55b03b0d1553912af8d9ae6
+  data.tar.gz: 7dc53be923fe5b8587f61617198b24c42e8793fbd8e18c42a17035bf68279c59c37c6c691cabe13c83adc5dc2cff66ea293f198297ab9a9de30aa68ca72bd9c4

data/CHANGELOG.md

@@ -1,5 +1,11 @@
 ## [Unreleased]
 
+## [0.3.12] - 2023-05-25
+- 🔍 Vectorsearch
+  - Introduce namespace support for Pinecone
+- 🚚 Loaders
+  - Loaders overhaul
+
 ## [0.3.11] - 2023-05-23
 - 🗣️ LLMs
   - Introducing support for Google PaLM (Pathways Language Model)

data/Gemfile.lock

@@ -1,7 +1,7 @@
 PATH
   remote: .
   specs:
-    langchainrb (0.3.11)
+    langchainrb (0.3.12)
 
 GEM
   remote: https://rubygems.org/
@@ -271,6 +271,7 @@ GEM
     zeitwerk (2.6.8)
 
 PLATFORMS
+  arm64-darwin-21
   arm64-darwin-22
   x86_64-darwin-19
   x86_64-darwin-22

data/README.md

@@ -268,12 +268,28 @@ agent.run(question: "How many full soccer fields would be needed to cover the di
 
 Need to read data from various sources? Load it up.
 
-| Name | Class         | Gem Requirements             |
-| ---- | ------------- | :--------------------------: |
-| docx | Loaders::Docx | `gem "docx", "~> 0.8.0"`     |
-| html | Loaders::HTML | `gem "nokogiri", "~> 1.13"`  |
-| pdf  | Loaders::PDF  | `gem "pdf-reader", "~> 1.4"` |
-| text | Loaders::Text |                              |
+##### Usage
+
+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')
+```
+
+or
+
+```ruby
+Langchain::Loader.load('https://www.example.com/file.pdf')
+```
+
+##### Supported Formats
+
+| Format | Pocessor         |       Gem Requirements       |
+| ------ | ---------------- | :--------------------------: |
+| docx   | Processors::Docx |   `gem "docx", "~> 0.8.0"`   |
+| html   | Processors::HTML | `gem "nokogiri", "~> 1.13"`  |
+| pdf    | Processors::PDF  | `gem "pdf-reader", "~> 1.4"` |
+| text   | Processors::Text |                              |
 
 ## Examples
 Additional examples available: [/examples](https://github.com/andreibondarev/langchainrb/tree/main/examples)

data/examples/pdf_store_and_query_with_chroma.rb

@@ -26,8 +26,8 @@ docs = [
 ]
 
 # Add data to the index. Weaviate will use OpenAI to generate embeddings behind the scene.
-chroma.add_texts(
-  texts: docs
+chroma.add_data(
+  paths: docs
 )
 
 # Query your data

data/lib/langchain/loader.rb

@@ -0,0 +1,80 @@
+# frozen_string_literal: true
+
+require "open-uri"
+
+module Langchain
+  class Loader
+    class FileNotFound < StandardError; end
+
+    class UnknownFormatError < StandardError; end
+
+    URI_REGEX = %r{\A[A-Za-z][A-Za-z0-9+\-.]*://}
+
+    # Load data from a file or url
+    # 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
+    end
+
+    # Initialize Langchain::Loader
+    # @param path [String | Pathname] path to file or url
+    # @return [Langchain::Loader] loader instance
+    def initialize(path)
+      @path = path
+    end
+
+    # Check if path is url
+    # @return [Boolean] true if path is url
+    def url?
+      return false if @path.is_a?(Pathname)
+
+      !!(@path =~ URI_REGEX)
+    end
+
+    # Load data from a file or url
+    # @return [String] file content
+    def load
+      url? ? from_url(@path) : from_path(@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
+    end
+
+    def from_path(path)
+      raise FileNotFound unless File.exist?(path)
+
+      process do
+        [File.open(path), find_processor(:EXTENSIONS, File.extname(path))]
+      end
+    end
+
+    def process(&block)
+      data, processor = yield
+
+      raise UnknownFormatError unless processor
+
+      Langchain::Processors.const_get(processor).new.parse(data)
+    end
+
+    def find_processor(constant, value)
+      processors.find { |klass| processor_matches? "#{klass}::#{constant}", value }
+    end
+
+    def processor_matches?(constant, value)
+      Langchain::Processors.const_get(constant).include?(value)
+    end
+
+    def processors
+      Langchain::Processors.constants
+    end
+  end
+end

data/lib/langchain/processors/base.rb

@@ -0,0 +1,14 @@
+# frozen_string_literal: true
+
+module Langchain
+  module Processors
+    class Base
+      EXTENSIONS = []
+      CONTENT_TYPES = []
+
+      def parse(data)
+        raise NotImplementedError
+      end
+    end
+  end
+end

data/lib/langchain/processors/docx.rb

@@ -0,0 +1,24 @@
+# frozen_string_literal: true
+
+module Langchain
+  module Processors
+    class Docx < Base
+      EXTENSIONS = [".docx"]
+      CONTENT_TYPES = ["application/vnd.openxmlformats-officedocument.wordprocessingml.document"]
+
+      def initialize
+        depends_on "docx"
+        require "docx"
+      end
+
+      # Parse the document and return the text
+      # @param [File] data
+      # @return [String]
+      def parse(data)
+        ::Docx::Document
+          .open(StringIO.new(data.read))
+          .text
+      end
+    end
+  end
+end

data/lib/langchain/processors/html.rb

@@ -0,0 +1,28 @@
+# frozen_string_literal: true
+
+module Langchain
+  module Processors
+    class HTML < Base
+      EXTENSIONS = [".html", ".htm"]
+      CONTENT_TYPES = ["text/html"]
+
+      # We only look for headings and paragraphs
+      TEXT_CONTENT_TAGS = %w[h1 h2 h3 h4 h5 h6 p]
+
+      def initialize
+        depends_on "nokogiri"
+        require "nokogiri"
+      end
+
+      # Parse the document and return the text
+      # @param [File] data
+      # @return [String]
+      def parse(data)
+        Nokogiri::HTML(data.read)
+          .css(TEXT_CONTENT_TAGS.join(","))
+          .map(&:inner_text)
+          .join("\n\n")
+      end
+    end
+  end
+end

data/lib/langchain/processors/pdf.rb

@@ -0,0 +1,26 @@
+# frozen_string_literal: true
+
+module Langchain
+  module Processors
+    class PDF < Base
+      EXTENSIONS = [".pdf"]
+      CONTENT_TYPES = ["application/pdf"]
+
+      def initialize
+        depends_on "pdf-reader"
+        require "pdf-reader"
+      end
+
+      # Parse the document and return the text
+      # @param [File] data
+      # @return [String]
+      def parse(data)
+        ::PDF::Reader
+          .new(StringIO.new(data.read))
+          .pages
+          .map(&:text)
+          .join("\n\n")
+      end
+    end
+  end
+end

data/lib/langchain/processors/text.rb

@@ -0,0 +1,17 @@
+# frozen_string_literal: true
+
+module Langchain
+  module Processors
+    class Text < Base
+      EXTENSIONS = [".txt"]
+      CONTENT_TYPES = ["text/plain"]
+
+      # Parse the document and return the text
+      # @param [File] data
+      # @return [String]
+      def parse(data)
+        data.read
+      end
+    end
+  end
+end

data/lib/langchain.rb

@@ -7,7 +7,6 @@ require_relative "./version"
 require_relative "./dependency_helper"
 module Langchain
   class << self
-    attr_accessor :default_loaders
     attr_accessor :logger
 
     attr_reader :root
@@ -16,6 +15,16 @@ module Langchain
   @logger ||= ::Logger.new($stdout, level: :warn, formatter: ->(severity, datetime, progname, msg) { "[LangChain.rb] #{msg}\n" })
 
   @root = Pathname.new(__dir__)
+
+  autoload :Loader, "langchain/loader"
+
+  module Processors
+    autoload :Base, "langchain/processors/base"
+    autoload :PDF, "langchain/processors/pdf"
+    autoload :HTML, "langchain/processors/html"
+    autoload :Text, "langchain/processors/text"
+    autoload :Docx, "langchain/processors/docx"
+  end
 end
 
 module Agent
@@ -55,16 +64,3 @@ module Tool
   autoload :SerpApi, "tool/serp_api"
   autoload :Wikipedia, "tool/wikipedia"
 end
-
-module Loaders
-  autoload :Base, "loaders/base"
-  autoload :Docx, "loaders/docx"
-  autoload :PDF, "loaders/pdf"
-  autoload :Text, "loaders/text"
-  autoload :HTML, "loaders/html"
-end
-
-autoload :Loader, "loader"
-
-# Load the default Loaders
-Langchain.default_loaders ||= [::Loaders::Text, ::Loaders::PDF, ::Loaders::Docx]

data/lib/langchainrb.rb

@@ -0,0 +1 @@
+require "langchain"

data/lib/vectorsearch/base.rb

@@ -19,8 +19,6 @@ module Vectorsearch
       @llm_api_key = llm_api_key
 
       @llm_client = LLM.const_get(LLM::Base::LLMS.fetch(llm)).new(api_key: llm_api_key)
-
-      @loaders = Langchain.default_loaders
     end
 
     # Method supported by Vectorsearch DB to create a default schema
@@ -74,18 +72,12 @@ module Vectorsearch
       raise ArgumentError, "Either path or paths must be provided" if path.nil? && paths.nil?
       raise ArgumentError, "Either path or paths must be provided, not both" if !path.nil? && !paths.nil?
 
-      texts =
-        Loader
-          .with(*loaders)
-          .load(path || paths)
+      texts = Array(path || paths)
+        .flatten
+        .map { |path| Langchain::Loader.new(path)&.load }
+        .compact
 
       add_texts(texts: texts)
     end
-
-    attr_reader :loaders
-
-    def add_loader(*loaders)
-      loaders.each { |loader| @loaders << loader }
-    end
   end
 end

data/lib/vectorsearch/pinecone.rb

@@ -25,20 +25,22 @@ module Vectorsearch
 
     # Add a list of texts to the index
     # @param texts [Array] The list of texts to add
+    # @param namespace [String] The namespace to add the texts to
+    # @param metadata [Hash] The metadata to use for the texts
     # @return [Hash] The response from the server
-    def add_texts(texts:)
+    def add_texts(texts:, namespace: "", metadata: nil)
       vectors = texts.map do |text|
         {
           # TODO: Allows passing in your own IDs
           id: SecureRandom.uuid,
-          metadata: {content: text},
+          metadata: metadata || {content: text},
           values: llm_client.embed(text: text)
         }
       end
 
       index = client.index(index_name)
 
-      index.upsert(vectors: vectors)
+      index.upsert(vectors: vectors, namespace: namespace)
     end
 
     # Create the index with the default schema
@@ -54,40 +56,54 @@ module Vectorsearch
     # Search for similar texts
     # @param query [String] The text to search for
     # @param k [Integer] The number of results to return
+    # @param namespace [String] The namespace to search in
+    # @param filter [String] The filter to use
     # @return [Array] The list of results
     def similarity_search(
       query:,
-      k: 4
+      k: 4,
+      namespace: "",
+      filter: nil
     )
       embedding = llm_client.embed(text: query)
 
       similarity_search_by_vector(
         embedding: embedding,
-        k: k
+        k: k,
+        namespace: namespace,
+        filter: filter
       )
     end
 
     # Search for similar texts by embedding
     # @param embedding [Array] The embedding to search for
     # @param k [Integer] The number of results to return
+    # @param namespace [String] The namespace to search in
+    # @param filter [String] The filter to use
     # @return [Array] The list of results
-    def similarity_search_by_vector(embedding:, k: 4)
+    def similarity_search_by_vector(embedding:, k: 4, namespace: "", filter: nil)
       index = client.index(index_name)
 
-      response = index.query(
+      query_params = {
         vector: embedding,
+        namespace: namespace,
+        filter: filter,
         top_k: k,
         include_values: true,
         include_metadata: true
-      )
+      }.compact
+
+      response = index.query(query_params)
       response.dig("matches")
     end
 
     # Ask a question and return the answer
     # @param question [String] The question to ask
+    # @param namespace [String] The namespace to search in
+    # @param filter [String] The filter to use
     # @return [String] The answer to the question
-    def ask(question:)
-      search_results = similarity_search(query: question)
+    def ask(question:, namespace: "", filter: nil)
+      search_results = similarity_search(query: question, namespace: namespace, filter: filter)
 
       context = search_results.map do |result|
         result.dig("metadata").to_s

data/lib/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Langchain
-  VERSION = "0.3.11"
+  VERSION = "0.3.12"
 end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: langchainrb
 version: !ruby/object:Gem::Version
-  version: 0.3.11
+  version: 0.3.12
 platform: ruby
 authors:
 - Andrei Bondarev
 autorequire:
 bindir: exe
 cert_chain: []
-date: 2023-05-24 00:00:00.000000000 Z
+date: 2023-05-25 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: dotenv-rails
@@ -288,18 +288,19 @@ files:
 - lib/agent/chain_of_thought_agent/chain_of_thought_agent_prompt.json
 - lib/dependency_helper.rb
 - lib/langchain.rb
+- lib/langchain/loader.rb
+- lib/langchain/processors/base.rb
+- lib/langchain/processors/docx.rb
+- lib/langchain/processors/html.rb
+- lib/langchain/processors/pdf.rb
+- lib/langchain/processors/text.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/replicate.rb
-- lib/loader.rb
-- lib/loaders/base.rb
-- lib/loaders/docx.rb
-- lib/loaders/html.rb
-- lib/loaders/pdf.rb
-- lib/loaders/text.rb
 - lib/prompt/base.rb
 - lib/prompt/few_shot_prompt_template.rb
 - lib/prompt/loading.rb

data/lib/loader.rb

@@ -1,26 +0,0 @@
-module Loader
-  def self.with(*loaders)
-    LoaderSet.new(loaders)
-  end
-
-  class LoaderSet
-    def initialize(loaders)
-      @loaders = Array(loaders)
-    end
-
-    def load(*paths)
-      Array(paths)
-        .flatten
-        .map { |path| first_loadable_loader(path)&.load }
-        .compact
-    end
-
-    def first_loadable_loader(path)
-      @loaders
-        .each do |loader_klass|
-          loader_instance = loader_klass.new(path)
-          return(loader_instance) if loader_instance.loadable?
-        end
-    end
-  end
-end

data/lib/loaders/base.rb

@@ -1,19 +0,0 @@
-# frozen_string_literal: true
-
-# TODO: Add chunking options to the loaders
-
-module Loaders
-  class Base
-    def self.load(path)
-      new.load(path)
-    end
-
-    def initialize(path)
-      @path = path
-    end
-
-    def loadable?
-      raise NotImplementedError
-    end
-  end
-end

data/lib/loaders/docx.rb

@@ -1,34 +0,0 @@
-# frozen_string_literal: true
-
-module Loaders
-  class Docx < Base
-    #
-    # This Loader parses Docx files into text.
-    # If you'd like to use it directly you can do so like this:
-    # Loaders::Docx.new("path/to/my.docx").load
-    #
-    # This parser is also invoked when you're adding data to a Vectorsearch DB:
-    # qdrant = Vectorsearch::Qdrant.new(...)
-    # path = Langchain.root.join("path/to/my.docx")
-    # qdrant.add_data(path: path)
-    #
-
-    def initialize(path)
-      depends_on "docx"
-      require "docx"
-
-      @path = path
-    end
-
-    # Check that the file is a `.docx` file
-    def loadable?
-      @path.to_s.end_with?(".docx")
-    end
-
-    def load
-      ::Docx::Document
-        .open(@path.to_s)
-        .text
-    end
-  end
-end

data/lib/loaders/html.rb

@@ -1,38 +0,0 @@
-# frozen_string_literal: true
-
-require "open-uri"
-
-module Loaders
-  class HTML < Base
-    # We only look for headings and paragraphs
-    TEXT_CONTENT_TAGS = %w[h1 h2 h3 h4 h5 h6 p]
-
-    #
-    # This Loader parses URL into a text.
-    # If you'd like to use it directly you can do so like this:
-    # Loaders::URL.new("https://nokogiri.org/").load
-    #
-    def initialize(url)
-      depends_on "nokogiri"
-      require "nokogiri"
-
-      @url = url
-    end
-
-    # Check that url is a valid URL
-    def loadable?
-      !!(@url =~ URI::DEFAULT_PARSER.make_regexp)
-    end
-
-    def load
-      return unless response.status.first == "200"
-
-      doc = Nokogiri::HTML(response.read)
-      doc.css(TEXT_CONTENT_TAGS.join(",")).map(&:inner_text).join("\n\n")
-    end
-
-    def response
-      @response ||= URI.parse(@url).open
-    end
-  end
-end

data/lib/loaders/pdf.rb

@@ -1,36 +0,0 @@
-# frozen_string_literal: true
-
-module Loaders
-  class PDF < Base
-    #
-    # This Loader parses PDF files into text.
-    # If you'd like to use it directly you can do so like this:
-    # Loaders::PDF.new("path/to/my.pdf").load
-    #
-    # This parser is also invoked when you're adding data to a Vectorsearch DB:
-    # qdrant = Vectorsearch::Qdrant.new(...)
-    # path = Langchain.root.join("path/to/my.pdf")
-    # qdrant.add_data(path: path)
-    #
-
-    def initialize(path)
-      depends_on "pdf-reader"
-      require "pdf-reader"
-
-      @path = path
-    end
-
-    # Check that the file is a PDF file
-    def loadable?
-      @path.to_s.end_with?(".pdf")
-    end
-
-    def load
-      ::PDF::Reader
-        .new(@path)
-        .pages
-        .map(&:text)
-        .join("\n\n")
-    end
-  end
-end

data/lib/loaders/text.rb

@@ -1,24 +0,0 @@
-# frozen_string_literal: true
-
-module Loaders
-  class Text < Base
-    #
-    # This Loader parses .txt files.
-    # If you'd like to use it directly you can do so like this:
-    # Loaders::Text.new("path/to/my.txt").load
-    #
-    # This parser is also invoked when you're adding data to a Vectorsearch DB:
-    # qdrant = Vectorsearch::Qdrant.new(...)
-    # path = Langchain.root.join("path/to/my.txt")
-    # qdrant.add_data(path: path)
-    #
-
-    def loadable?
-      @path.to_s.end_with?(".txt")
-    end
-
-    def load
-      @path.read
-    end
-  end
-end