langchainrb 0.6.16 → 0.6.18

data/lib/langchain/llm/response/openai_response.rb

@@ -0,0 +1,51 @@
+# frozen_string_literal: true
+
+module Langchain::LLM
+  class OpenAIResponse < BaseResponse
+    def model
+      raw_response["model"]
+    end
+
+    def created_at
+      if raw_response.dig("created")
+        Time.at(raw_response.dig("created"))
+      end
+    end
+
+    def completion
+      completions&.dig(0, "message", "content")
+    end
+
+    def chat_completion
+      completion
+    end
+
+    def embedding
+      embeddings&.first
+    end
+
+    def completions
+      raw_response.dig("choices")
+    end
+
+    def chat_completions
+      raw_response.dig("choices")
+    end
+
+    def embeddings
+      raw_response.dig("data")&.map { |datum| datum.dig("embedding") }
+    end
+
+    def prompt_tokens
+      raw_response.dig("usage", "prompt_tokens")
+    end
+
+    def completion_tokens
+      raw_response.dig("usage", "completion_tokens")
+    end
+
+    def total_tokens
+      raw_response.dig("usage", "total_tokens")
+    end
+  end
+end

data/lib/langchain/llm/response/replicate_response.rb

@@ -0,0 +1,28 @@
+# frozen_string_literal: true
+
+module Langchain::LLM
+  class ReplicateResponse < BaseResponse
+    def completions
+      # Response comes back as an array of strings, e.g.: ["Hi", "how ", "are ", "you?"]
+      # The first array element is missing a space at the end, so we add it manually
+      raw_response.output[0] += " "
+      [raw_response.output.join]
+    end
+
+    def completion
+      completions.first
+    end
+
+    def created_at
+      Time.parse(raw_response.created_at)
+    end
+
+    def embedding
+      embeddings.first
+    end
+
+    def embeddings
+      [raw_response.output]
+    end
+  end
+end

data/lib/langchain/vectorsearch/base.rb

@@ -184,7 +184,7 @@ module Langchain::Vectorsearch
         .flatten
         .map do |path|
           data = Langchain::Loader.new(path)&.load&.chunks
-          data.map { |chunk| chunk[:text] }
+          data.map { |chunk| chunk.text }
         end
 
       texts.flatten!

data/lib/langchain/vectorsearch/chroma.rb

@@ -5,18 +5,17 @@ module Langchain::Vectorsearch
     #
     # Wrapper around Chroma DB
     #
-    # Gem requirements: gem "chroma-db", "~> 0.3.0"
+    # Gem requirements: gem "chroma-db", "~> 0.6.0"
     #
     # Usage:
     # chroma = Langchain::Vectorsearch::Chroma.new(url:, index_name:, llm:, llm_api_key:, api_key: nil)
     #
 
     # Initialize the Chroma client
-    # @param url [String] The URL of the Qdrant server
-    # @param api_key [String] The API key to use
+    # @param url [String] The URL of the Chroma server
     # @param index_name [String] The name of the index to use
     # @param llm [Object] The LLM client to use
-    def initialize(url:, index_name:, llm:, api_key: nil)
+    def initialize(url:, index_name:, llm:)
       depends_on "chroma-db"
 
       ::Chroma.connect_host = url
@@ -29,13 +28,13 @@ module Langchain::Vectorsearch
     end
 
     # Add a list of texts to the index
-    # @param texts [Array] The list of texts to add
+    # @param texts [Array<String>] The list of texts to add
     # @return [Hash] The response from the server
     def add_texts(texts:, ids: [])
       embeddings = Array(texts).map.with_index do |text, i|
         ::Chroma::Resources::Embedding.new(
           id: ids[i] ? ids[i].to_s : SecureRandom.uuid,
-          embedding: llm.embed(text: text),
+          embedding: llm.embed(text: text).embedding,
           # TODO: Add support for passing metadata
           metadata: {}, # metadatas[index],
           document: text # Do we actually need to store the whole original document?
@@ -50,7 +49,7 @@ module Langchain::Vectorsearch
       embeddings = Array(texts).map.with_index do |text, i|
         ::Chroma::Resources::Embedding.new(
           id: ids[i].to_s,
-          embedding: llm.embed(text: text),
+          embedding: llm.embed(text: text).embedding,
           # TODO: Add support for passing metadata
           metadata: [], # metadatas[index],
           document: text # Do we actually need to store the whole original document?
@@ -61,19 +60,19 @@ module Langchain::Vectorsearch
     end
 
     # Create the collection with the default schema
-    # @return [Hash] The response from the server
+    # @return [::Chroma::Resources::Collection] Created collection
     def create_default_schema
       ::Chroma::Resources::Collection.create(index_name)
     end
 
     # Get the default schema
-    # @return [Hash] The response from the server
+    # @return [::Chroma::Resources::Collection] Default schema
     def get_default_schema
       ::Chroma::Resources::Collection.get(index_name)
     end
 
     # Delete the default schema
-    # @return [Hash] The response from the server
+    # @return [bool] Success or failure
     def destroy_default_schema
       ::Chroma::Resources::Collection.delete(index_name)
     end
@@ -86,7 +85,7 @@ module Langchain::Vectorsearch
       query:,
       k: 4
     )
-      embedding = llm.embed(text: query)
+      embedding = llm.embed(text: query).embedding
 
       similarity_search_by_vector(
         embedding: embedding,
@@ -95,7 +94,7 @@ module Langchain::Vectorsearch
     end
 
     # Search for similar texts by embedding
-    # @param embedding [Array] The embedding to search for
+    # @param embedding [Array<Float>] The embedding to search for
     # @param k [Integer] The number of results to return
     # @return [Chroma::Resources::Embedding] The response from the server
     def similarity_search_by_vector(

data/lib/langchain/vectorsearch/hnswlib.rb

@@ -35,15 +35,15 @@ module Langchain::Vectorsearch
     #
     # Add a list of texts and corresponding IDs to the index
     #
-    # @param texts [Array] The list of texts to add
-    # @param ids [Array] The list of corresponding IDs (integers) to the texts
+    # @param texts [Array<String>] The list of texts to add
+    # @param ids [Array<Integer>] The list of corresponding IDs (integers) to the texts
     # @return [Boolean] The response from the HNSW library
     #
     def add_texts(texts:, ids:)
       resize_index(texts.size)
 
       Array(texts).each_with_index do |text, i|
-        embedding = llm.embed(text: text)
+        embedding = llm.embed(text: text).embedding
 
         client.add_point(embedding, ids[i])
       end
@@ -64,7 +64,7 @@ module Langchain::Vectorsearch
       query:,
       k: 4
     )
-      embedding = llm.embed(text: query)
+      embedding = llm.embed(text: query).embedding
 
       similarity_search_by_vector(
         embedding: embedding,
@@ -75,7 +75,7 @@ module Langchain::Vectorsearch
     #
     # Search for the K nearest neighbors of a given vector
     #
-    # @param embedding [Array] The embedding to search for
+    # @param embedding [Array<Float>] The embedding to search for
     # @param k [Integer] The number of results to return
     # @return [Array] Results in the format `[[id1, distance3], [id2, distance2]]`
     #

data/lib/langchain/vectorsearch/milvus.rb

@@ -32,7 +32,7 @@ module Langchain::Vectorsearch
           }, {
             field_name: "vectors",
             type: ::Milvus::DATA_TYPES["float_vector"],
-            field: Array(texts).map { |text| llm.embed(text: text) }
+            field: Array(texts).map { |text| llm.embed(text: text).embedding }
           }
         ]
       )
@@ -111,7 +111,7 @@ module Langchain::Vectorsearch
     end
 
     def similarity_search(query:, k: 4)
-      embedding = llm.embed(text: query)
+      embedding = llm.embed(text: query).embedding
 
       similarity_search_by_vector(
         embedding: embedding,

data/lib/langchain/vectorsearch/pgvector.rb

@@ -52,7 +52,7 @@ module Langchain::Vectorsearch
     # the added or updated texts.
     def upsert_texts(texts:, ids:)
       data = texts.zip(ids).flat_map do |(text, id)|
-        {id: id, content: text, vectors: llm.embed(text: text).to_s, namespace: namespace}
+        {id: id, content: text, vectors: llm.embed(text: text).embedding.to_s, namespace: namespace}
       end
       # @db[table_name.to_sym].multi_insert(data, return: :primary_key)
       @db[table_name.to_sym]
@@ -70,7 +70,7 @@ module Langchain::Vectorsearch
     def add_texts(texts:, ids: nil)
       if ids.nil? || ids.empty?
         data = texts.map do |text|
-          {content: text, vectors: llm.embed(text: text).to_s, namespace: namespace}
+          {content: text, vectors: llm.embed(text: text).embedding.to_s, namespace: namespace}
         end
 
         @db[table_name.to_sym].multi_insert(data, return: :primary_key)
@@ -110,7 +110,7 @@ module Langchain::Vectorsearch
     # @param k [Integer] The number of top results to return
     # @return [Array<Hash>] The results of the search
     def similarity_search(query:, k: 4)
-      embedding = llm.embed(text: query)
+      embedding = llm.embed(text: query).embedding
 
       similarity_search_by_vector(
         embedding: embedding,

data/lib/langchain/vectorsearch/pinecone.rb

@@ -31,7 +31,7 @@ module Langchain::Vectorsearch
     end
 
     # Find records by ids
-    # @param ids [Array] The ids to find
+    # @param ids [Array<Integer>] The ids to find
     # @param namespace String The namespace to search through
     # @return [Hash] The response from the server
     def find(ids: [], namespace: "")
@@ -44,8 +44,8 @@ module Langchain::Vectorsearch
     end
 
     # Add a list of texts to the index
-    # @param texts [Array] The list of texts to add
-    # @param ids [Array] The list of IDs to add
+    # @param texts [Array<String>] The list of texts to add
+    # @param ids [Array<Integer>] The list of IDs 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
@@ -54,7 +54,7 @@ module Langchain::Vectorsearch
         {
           id: ids[i] ? ids[i].to_s : SecureRandom.uuid,
           metadata: metadata || {content: text},
-          values: llm.embed(text: text)
+          values: llm.embed(text: text).embedding
         }
       end
 
@@ -70,7 +70,7 @@ module Langchain::Vectorsearch
         .flatten
         .map do |path|
           data = Langchain::Loader.new(path)&.load&.chunks
-          data.map { |chunk| chunk[:text] }
+          data.map { |chunk| chunk.text }
         end
 
       texts.flatten!
@@ -79,8 +79,8 @@ module Langchain::Vectorsearch
     end
 
     # Update a list of texts in the index
-    # @param texts [Array] The list of texts to update
-    # @param ids [Array] The list of IDs to update
+    # @param texts [Array<String>] The list of texts to update
+    # @param ids [Array<Integer>] The list of IDs to update
     # @param namespace [String] The namespace to update the texts in
     # @param metadata [Hash] The metadata to use for the texts
     # @return [Array] The response from the server
@@ -90,7 +90,7 @@ module Langchain::Vectorsearch
         index.update(
           namespace: namespace,
           id: ids[i].to_s,
-          values: llm.embed(text: text),
+          values: llm.embed(text: text).embedding,
           set_metadata: metadata
         )
       end
@@ -130,7 +130,7 @@ module Langchain::Vectorsearch
       namespace: "",
       filter: nil
     )
-      embedding = llm.embed(text: query)
+      embedding = llm.embed(text: query).embedding
 
       similarity_search_by_vector(
         embedding: embedding,
@@ -141,7 +141,7 @@ module Langchain::Vectorsearch
     end
 
     # Search for similar texts by embedding
-    # @param embedding [Array] The embedding to search for
+    # @param embedding [Array<Float>] 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

data/lib/langchain/vectorsearch/qdrant.rb

@@ -29,7 +29,7 @@ module Langchain::Vectorsearch
     end
 
     # Find records by ids
-    # @param ids [Array] The ids to find
+    # @param ids [Array<Integer>] The ids to find
     # @return [Hash] The response from the server
     def find(ids: [])
       client.points.get_all(
@@ -41,7 +41,7 @@ module Langchain::Vectorsearch
     end
 
     # Add a list of texts to the index
-    # @param texts [Array] The list of texts to add
+    # @param texts [Array<String>] The list of texts to add
     # @return [Hash] The response from the server
     def add_texts(texts:, ids: [])
       batch = {ids: [], vectors: [], payloads: []}
@@ -49,7 +49,7 @@ module Langchain::Vectorsearch
       Array(texts).each_with_index do |text, i|
         id = ids[i] || SecureRandom.uuid
         batch[:ids].push(id)
-        batch[:vectors].push(llm.embed(text: text))
+        batch[:vectors].push(llm.embed(text: text).embedding)
         batch[:payloads].push({content: text})
       end
 
@@ -95,7 +95,7 @@ module Langchain::Vectorsearch
       query:,
       k: 4
     )
-      embedding = llm.embed(text: query)
+      embedding = llm.embed(text: query).embedding
 
       similarity_search_by_vector(
         embedding: embedding,
@@ -104,7 +104,7 @@ module Langchain::Vectorsearch
     end
 
     # Search for similar texts by embedding
-    # @param embedding [Array] The embedding to search for
+    # @param embedding [Array<Float>] The embedding to search for
     # @param k [Integer] The number of results to return
     # @return [Hash] The response from the server
     def similarity_search_by_vector(

data/lib/langchain/vectorsearch/weaviate.rb

@@ -32,7 +32,7 @@ module Langchain::Vectorsearch
     end
 
     # Add a list of texts to the index
-    # @param texts [Array] The list of texts to add
+    # @param texts [Array<String>] The list of texts to add
     # @return [Hash] The response from the server
     def add_texts(texts:, ids: [])
       client.objects.batch_create(
@@ -41,7 +41,7 @@ module Langchain::Vectorsearch
     end
 
     # Update a list of texts in the index
-    # @param texts [Array] The list of texts to update
+    # @param texts [Array<String>] The list of texts to update
     # @return [Hash] The response from the server
     def update_texts(texts:, ids:)
       uuids = []
@@ -65,7 +65,7 @@ module Langchain::Vectorsearch
             __id: ids[i].to_s,
             content: text
           },
-          vector: llm.embed(text: text)
+          vector: llm.embed(text: text).embedding
         )
       end
     end
@@ -101,13 +101,13 @@ module Langchain::Vectorsearch
     # @param k [Integer|String] The number of results to return
     # @return [Hash] The search results
     def similarity_search(query:, k: 4)
-      embedding = llm.embed(text: query)
+      embedding = llm.embed(text: query).embedding
 
       similarity_search_by_vector(embedding: embedding, k: k)
     end
 
     # Return documents similar to the vector
-    # @param embedding [Array] The vector to search for
+    # @param embedding [Array<Float>] The vector to search for
     # @param k [Integer|String] The number of results to return
     # @return [Hash] The search results
     def similarity_search_by_vector(embedding:, k: 4)
@@ -154,7 +154,7 @@ module Langchain::Vectorsearch
           __id: id.to_s,
           content: text
         },
-        vector: llm.embed(text: text)
+        vector: llm.embed(text: text).embedding
       }
     end
   end

data/lib/langchain/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Langchain
-  VERSION = "0.6.16"
+  VERSION = "0.6.18"
 end

data/lib/langchain.rb

@@ -7,8 +7,8 @@ require "zeitwerk"
 loader = Zeitwerk::Loader.for_gem
 loader.ignore("#{__dir__}/langchainrb.rb")
 loader.inflector.inflect(
-  "ai_message" => "AIMessage",
   "ai21" => "AI21",
+  "ai21_response" => "AI21Response",
   "ai21_validator" => "AI21Validator",
   "csv" => "CSV",
   "html" => "HTML",
@@ -17,10 +17,12 @@ loader.inflector.inflect(
   "llm" => "LLM",
   "openai" => "OpenAI",
   "openai_validator" => "OpenAIValidator",
+  "openai_response" => "OpenAIResponse",
   "pdf" => "PDF",
   "react_agent" => "ReActAgent",
   "sql_query_agent" => "SQLQueryAgent"
 )
+loader.collapse("#{__dir__}/langchain/llm/response")
 loader.setup
 
 # Langchain.rb a is library for building LLM-backed Ruby applications. It is an abstraction layer that sits on top of the emerging AI-related tools that makes it easy for developers to consume and string those services together.

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: langchainrb
 version: !ruby/object:Gem::Version
-  version: 0.6.16
+  version: 0.6.18
 platform: ruby
 authors:
 - Andrei Bondarev
 autorequire:
 bindir: exe
 cert_chain: []
-date: 2023-10-03 00:00:00.000000000 Z
+date: 2023-10-17 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: baran
@@ -184,14 +184,14 @@ dependencies:
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: 0.3.0
+        version: 0.6.0
   type: :development
   prerelease: false
   version_requirements: !ruby/object:Gem::Requirement
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: 0.3.0
+        version: 0.6.0
 - !ruby/object:Gem::Dependency
   name: cohere-ruby
   requirement: !ruby/object:Gem::Requirement
@@ -492,14 +492,14 @@ dependencies:
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: 0.8.7
+        version: 0.8.9
   type: :development
   prerelease: false
   version_requirements: !ruby/object:Gem::Requirement
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: 0.8.7
+        version: 0.8.9
 - !ruby/object:Gem::Dependency
   name: wikipedia-client
   requirement: !ruby/object:Gem::Requirement
@@ -532,17 +532,22 @@ files:
 - lib/langchain/agent/sql_query_agent.rb
 - lib/langchain/agent/sql_query_agent/sql_query_agent_answer_prompt.yaml
 - lib/langchain/agent/sql_query_agent/sql_query_agent_sql_prompt.yaml
-- lib/langchain/ai_message.rb
+- lib/langchain/chunk.rb
 - lib/langchain/chunker/base.rb
+- lib/langchain/chunker/prompts/semantic_prompt_template.yml
 - lib/langchain/chunker/recursive_text.rb
+- lib/langchain/chunker/semantic.rb
 - lib/langchain/chunker/sentence.rb
 - lib/langchain/chunker/text.rb
 - lib/langchain/contextual_logger.rb
 - lib/langchain/conversation.rb
-- lib/langchain/conversation_memory.rb
+- lib/langchain/conversation/context.rb
+- lib/langchain/conversation/memory.rb
+- lib/langchain/conversation/message.rb
+- lib/langchain/conversation/prompt.rb
+- lib/langchain/conversation/response.rb
 - lib/langchain/data.rb
 - lib/langchain/dependency_helper.rb
-- lib/langchain/human_message.rb
 - lib/langchain/llm/ai21.rb
 - lib/langchain/llm/anthropic.rb
 - lib/langchain/llm/base.rb
@@ -554,8 +559,16 @@ files:
 - lib/langchain/llm/openai.rb
 - lib/langchain/llm/prompts/summarize_template.yaml
 - lib/langchain/llm/replicate.rb
+- lib/langchain/llm/response/ai21_response.rb
+- lib/langchain/llm/response/anthropic_response.rb
+- lib/langchain/llm/response/base_response.rb
+- lib/langchain/llm/response/cohere_response.rb
+- lib/langchain/llm/response/google_palm_response.rb
+- lib/langchain/llm/response/hugging_face_response.rb
+- lib/langchain/llm/response/ollama_response.rb
+- lib/langchain/llm/response/openai_response.rb
+- lib/langchain/llm/response/replicate_response.rb
 - lib/langchain/loader.rb
-- lib/langchain/message.rb
 - lib/langchain/output_parsers/base.rb
 - lib/langchain/output_parsers/output_fixing_parser.rb
 - lib/langchain/output_parsers/prompts/naive_fix_prompt.yaml
@@ -575,7 +588,6 @@ files:
 - lib/langchain/prompt/loading.rb
 - lib/langchain/prompt/prompt_template.rb
 - lib/langchain/railtie.rb
-- lib/langchain/system_message.rb
 - lib/langchain/tool/base.rb
 - lib/langchain/tool/calculator.rb
 - lib/langchain/tool/database.rb

data/lib/langchain/conversation_memory.rb

@@ -1,84 +0,0 @@
-# frozen_string_literal: true
-
-module Langchain
-  class ConversationMemory
-    attr_reader :examples, :messages
-
-    # The least number of tokens we want to be under the limit by
-    TOKEN_LEEWAY = 20
-
-    def initialize(llm:, messages: [], **options)
-      @llm = llm
-      @context = nil
-      @summary = nil
-      @examples = []
-      @messages = messages
-      @strategy = options.delete(:strategy) || :truncate
-      @options = options
-    end
-
-    def set_context(message)
-      @context = message
-    end
-
-    def add_examples(examples)
-      @examples.concat examples
-    end
-
-    def append_message(message)
-      @messages.append(message)
-    end
-
-    def reduce_messages(exception)
-      case @strategy
-      when :truncate
-        truncate_messages(exception)
-      when :summarize
-        summarize_messages
-      else
-        raise "Unknown strategy: #{@options[:strategy]}"
-      end
-    end
-
-    def context
-      return if @context.nil? && @summary.nil?
-
-      SystemMessage.new([@context, @summary].compact.join("\n"))
-    end
-
-    private
-
-    def truncate_messages(exception)
-      raise exception if @messages.size == 1
-
-      token_overflow = exception.token_overflow
-
-      @messages = @messages.drop_while do |message|
-        proceed = token_overflow > -TOKEN_LEEWAY
-        token_overflow -= token_length(message.to_json, model_name, llm: @llm)
-
-        proceed
-      end
-    end
-
-    def summarize_messages
-      history = [@summary, @messages.to_json].compact.join("\n")
-      partitions = [history[0, history.size / 2], history[history.size / 2, history.size]]
-
-      @summary = partitions.map { |messages| @llm.summarize(text: messages.to_json) }.join("\n")
-
-      @messages = [@messages.last]
-    end
-
-    def partition_messages
-    end
-
-    def model_name
-      @llm.class::DEFAULTS[:chat_completion_model_name]
-    end
-
-    def token_length(content, model_name, options)
-      @llm.class::LENGTH_VALIDATOR.token_length(content, model_name, options)
-    end
-  end
-end

data/lib/langchain/message.rb

@@ -1,35 +0,0 @@
-# frozen_string_literal: true
-
-module Langchain
-  class Message
-    attr_reader :content, :additional_kwargs
-
-    def initialize(content, additional_kwargs = nil)
-      @content = content
-      @additional_kwargs = additional_kwargs
-    end
-
-    def type
-      raise NotImplementedError
-    end
-
-    def to_s
-      content
-    end
-
-    def ==(other)
-      to_json == other.to_json
-    end
-
-    def to_json(options = {})
-      hash = {
-        type: type,
-        content: content
-      }
-
-      hash[:additional_kwargs] = additional_kwargs unless additional_kwargs.nil? || additional_kwargs.empty?
-
-      hash.to_json
-    end
-  end
-end