langchainrb 0.13.4 → 0.14.0

data/lib/langchain/llm/azure.rb

@@ -42,17 +42,17 @@ module Langchain::LLM
 
     def embed(...)
       @client = @embed_client
-      super(...)
+      super
     end
 
     def complete(...)
       @client = @chat_client
-      super(...)
+      super
     end
 
     def chat(...)
       @client = @chat_client
-      super(...)
+      super
     end
   end
 end

data/lib/langchain/llm/base.rb

@@ -8,6 +8,7 @@ module Langchain::LLM
   # Langchain.rb provides a common interface to interact with all supported LLMs:
   #
   # - {Langchain::LLM::AI21}
+  # - {Langchain::LLM::Anthropic}
   # - {Langchain::LLM::Azure}
   # - {Langchain::LLM::Cohere}
   # - {Langchain::LLM::GooglePalm}

data/lib/langchain/llm/cohere.rb

@@ -74,8 +74,6 @@ module Langchain::LLM
 
       default_params.merge!(params)
 
-      default_params[:max_tokens] = Langchain::Utils::TokenLength::CohereValidator.validate_max_tokens!(prompt, default_params[:model], llm: client)
-
       response = client.generate(**default_params)
       Langchain::LLM::CohereResponse.new response, model: @defaults[:completion_model_name]
     end

data/lib/langchain/llm/google_gemini.rb

@@ -18,7 +18,9 @@ module Langchain::LLM
 
       chat_parameters.update(
         model: {default: @defaults[:chat_completion_model_name]},
-        temperature: {default: @defaults[:temperature]}
+        temperature: {default: @defaults[:temperature]},
+        generation_config: {default: nil},
+        safety_settings: {default: nil}
       )
       chat_parameters.remap(
         messages: :contents,
@@ -42,13 +44,25 @@ module Langchain::LLM
       raise ArgumentError.new("messages argument is required") if Array(params[:messages]).empty?
 
       parameters = chat_parameters.to_params(params)
-      parameters[:generation_config] = {temperature: parameters.delete(:temperature)} if parameters[:temperature]
+      parameters[:generation_config] ||= {}
+      parameters[:generation_config][:temperature] ||= parameters[:temperature] if parameters[:temperature]
+      parameters.delete(:temperature)
+      parameters[:generation_config][:top_p] ||= parameters[:top_p] if parameters[:top_p]
+      parameters.delete(:top_p)
+      parameters[:generation_config][:top_k] ||= parameters[:top_k] if parameters[:top_k]
+      parameters.delete(:top_k)
+      parameters[:generation_config][:max_output_tokens] ||= parameters[:max_tokens] if parameters[:max_tokens]
+      parameters.delete(:max_tokens)
+      parameters[:generation_config][:response_mime_type] ||= parameters[:response_format] if parameters[:response_format]
+      parameters.delete(:response_format)
+      parameters[:generation_config][:stop_sequences] ||= parameters[:stop] if parameters[:stop]
+      parameters.delete(:stop)
 
       uri = URI("https://generativelanguage.googleapis.com/v1beta/models/#{parameters[:model]}:generateContent?key=#{api_key}")
 
       request = Net::HTTP::Post.new(uri)
       request.content_type = "application/json"
-      request.body = parameters.to_json
+      request.body = Langchain::Utils::HashTransformer.deep_transform_keys(parameters) { |key| Langchain::Utils::HashTransformer.camelize_lower(key.to_s).to_sym }.to_json
 
       response = Net::HTTP.start(uri.hostname, uri.port, use_ssl: uri.scheme == "https") do |http|
         http.request(request)

data/lib/langchain/llm/google_palm.rb

@@ -18,7 +18,7 @@ module Langchain::LLM
       chat_completion_model_name: "chat-bison-001",
       embeddings_model_name: "embedding-gecko-001"
     }.freeze
-    LENGTH_VALIDATOR = Langchain::Utils::TokenLength::GooglePalmValidator
+
     ROLE_MAPPING = {
       "assistant" => "ai"
     }
@@ -96,9 +96,6 @@ module Langchain::LLM
         examples: compose_examples(examples)
       }
 
-      # chat-bison-001 is the only model that currently supports countMessageTokens functions
-      LENGTH_VALIDATOR.validate_max_tokens!(default_params[:messages], "chat-bison-001", llm: self)
-
       if options[:stop_sequences]
         default_params[:stop] = options.delete(:stop_sequences)
       end

data/lib/langchain/llm/ollama.rb

@@ -14,7 +14,7 @@ module Langchain::LLM
     attr_reader :url, :defaults
 
     DEFAULTS = {
-      temperature: 0.8,
+      temperature: 0.0,
       completion_model_name: "llama3",
       embeddings_model_name: "llama3",
       chat_completion_model_name: "llama3"

data/lib/langchain/llm/replicate.rb

@@ -64,7 +64,7 @@ module Langchain::LLM
     # Generate a completion for a given prompt
     #
     # @param prompt [String] The prompt to generate a completion for
-    # @return [Langchain::LLM::ReplicateResponse] Reponse object
+    # @return [Langchain::LLM::ReplicateResponse] Response object
     #
     def complete(prompt:, **params)
       response = completion_model.predict(prompt: prompt)

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

@@ -3,7 +3,7 @@
 module Langchain::LLM
   class GoogleGeminiResponse < BaseResponse
     def initialize(raw_response, model: nil)
-      super(raw_response, model: model)
+      super
     end
 
     def chat_completion

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

@@ -36,7 +36,7 @@ module Langchain::LLM
     end
 
     def prompt_tokens
-      raw_response.dig("prompt_eval_count") if done?
+      raw_response.fetch("prompt_eval_count", 0) if done?
     end
 
     def completion_tokens
@@ -47,6 +47,24 @@ module Langchain::LLM
       prompt_tokens + completion_tokens if done?
     end
 
+    def tool_calls
+      if chat_completion && (parsed_tool_calls = JSON.parse(chat_completion))
+        [parsed_tool_calls]
+      elsif completion&.include?("[TOOL_CALLS]") && (
+        parsed_tool_calls = JSON.parse(
+          completion
+            # Slice out the serialize JSON
+            .slice(/\{.*\}/)
+            # Replace hash rocket with colon
+            .gsub("=>", ":")
+        )
+      )
+        [parsed_tool_calls]
+      else
+        []
+      end
+    end
+
     private
 
     def done?

data/lib/langchain/loader.rb

@@ -90,7 +90,9 @@ module Langchain
     private
 
     def load_from_url
-      URI.parse(URI::DEFAULT_PARSER.escape(@path)).open
+      unescaped_url = URI.decode_www_form_component(@path)
+      escaped_url = URI::DEFAULT_PARSER.escape(unescaped_url)
+      URI.parse(escaped_url).open
     end
 
     def load_from_path

data/lib/langchain/utils/hash_transformer.rb

@@ -0,0 +1,25 @@
+module Langchain
+  module Utils
+    class HashTransformer
+      # Converts a string to camelCase
+      def self.camelize_lower(str)
+        str.split("_").inject([]) { |buffer, e| buffer.push(buffer.empty? ? e : e.capitalize) }.join
+      end
+
+      # Recursively transforms the keys of a hash to camel case
+      def self.deep_transform_keys(hash, &block)
+        case hash
+        when Hash
+          hash.each_with_object({}) do |(key, value), result|
+            new_key = block.call(key)
+            result[new_key] = deep_transform_keys(value, &block)
+          end
+        when Array
+          hash.map { |item| deep_transform_keys(item, &block) }
+        else
+          hash
+        end
+      end
+    end
+  end
+end

data/lib/langchain/vectorsearch/chroma.rb

@@ -64,7 +64,9 @@ module Langchain::Vectorsearch
     # @param ids [Array<String>] The list of ids to remove
     # @return [Hash] The response from the server
     def remove_texts(ids:)
-      collection.delete(ids: ids)
+      collection.delete(
+        ids: ids.map(&:to_s)
+      )
     end
 
     # Create the collection with the default schema

data/lib/langchain/vectorsearch/milvus.rb

@@ -6,7 +6,7 @@ module Langchain::Vectorsearch
     # Wrapper around Milvus REST APIs.
     #
     # Gem requirements:
-    #     gem "milvus", "~> 0.9.2"
+    #     gem "milvus", "~> 0.9.3"
     #
     # Usage:
     # milvus = Langchain::Vectorsearch::Milvus.new(url:, index_name:, llm:, api_key:)
@@ -39,6 +39,21 @@ module Langchain::Vectorsearch
       )
     end
 
+    # Deletes a list of texts in the index
+    #
+    # @param ids [Array<Integer>] The ids of texts to delete
+    # @return [Boolean] The response from the server
+    def remove_texts(ids:)
+      raise ArgumentError, "ids must be an array" unless ids.is_a?(Array)
+      # Convert ids to integers if strings are passed
+      ids = ids.map(&:to_i)
+
+      client.entities.delete(
+        collection_name: index_name,
+        expression: "id in #{ids}"
+      )
+    end
+
     # TODO: Add update_texts method
 
     # Create default schema
@@ -83,7 +98,7 @@ module Langchain::Vectorsearch
     # @return [Boolean] The response from the server
     def create_default_index
       client.indices.create(
-        collection_name: "Documents",
+        collection_name: index_name,
         field_name: "vectors",
         extra_params: [
           {key: "metric_type", value: "L2"},
@@ -125,7 +140,7 @@ module Langchain::Vectorsearch
 
       client.search(
         collection_name: index_name,
-        output_fields: ["id", "content", "vectors"],
+        output_fields: ["id", "content"], # Add "vectors" if need to have full vectors returned.
         top_k: k.to_s,
         vectors: [embedding],
         dsl_type: 1,

data/lib/langchain/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Langchain
-  VERSION = "0.13.4"
+  VERSION = "0.14.0"
 end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: langchainrb
 version: !ruby/object:Gem::Version
-  version: 0.13.4
+  version: 0.14.0
 platform: ruby
 authors:
 - Andrei Bondarev
 autorequire:
 bindir: exe
 cert_chain: []
-date: 2024-06-16 00:00:00.000000000 Z
+date: 2024-07-12 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: baran
@@ -212,14 +212,14 @@ dependencies:
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: '0.2'
+        version: '0.3'
   type: :development
   prerelease: false
   version_requirements: !ruby/object:Gem::Requirement
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: '0.2'
+        version: '0.3'
 - !ruby/object:Gem::Dependency
   name: aws-sdk-bedrockruntime
   requirement: !ruby/object:Gem::Requirement
@@ -408,14 +408,14 @@ dependencies:
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: 0.9.2
+        version: 0.9.3
   type: :development
   prerelease: false
   version_requirements: !ruby/object:Gem::Requirement
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: 0.9.2
+        version: 0.9.3
 - !ruby/object:Gem::Dependency
   name: llama_cpp
   requirement: !ruby/object:Gem::Requirement
@@ -682,20 +682,6 @@ dependencies:
     - - "~>"
       - !ruby/object:Gem::Version
         version: 0.1.0
-- !ruby/object:Gem::Dependency
-  name: tiktoken_ruby
-  requirement: !ruby/object:Gem::Requirement
-    requirements:
-    - - "~>"
-      - !ruby/object:Gem::Version
-        version: 0.0.9
-  type: :development
-  prerelease: false
-  version_requirements: !ruby/object:Gem::Requirement
-    requirements:
-    - - "~>"
-      - !ruby/object:Gem::Version
-        version: 0.0.9
 description: Build LLM-backed Ruby applications with Ruby's Langchain.rb
 email:
 - andrei.bondarev13@gmail.com
@@ -711,6 +697,7 @@ files:
 - lib/langchain/assistants/messages/anthropic_message.rb
 - lib/langchain/assistants/messages/base.rb
 - lib/langchain/assistants/messages/google_gemini_message.rb
+- lib/langchain/assistants/messages/ollama_message.rb
 - lib/langchain/assistants/messages/openai_message.rb
 - lib/langchain/assistants/thread.rb
 - lib/langchain/chunk.rb
@@ -809,12 +796,7 @@ files:
 - lib/langchain/tool/wikipedia/wikipedia.json
 - lib/langchain/tool/wikipedia/wikipedia.rb
 - lib/langchain/utils/cosine_similarity.rb
-- lib/langchain/utils/token_length/ai21_validator.rb
-- lib/langchain/utils/token_length/base_validator.rb
-- lib/langchain/utils/token_length/cohere_validator.rb
-- lib/langchain/utils/token_length/google_palm_validator.rb
-- lib/langchain/utils/token_length/openai_validator.rb
-- lib/langchain/utils/token_length/token_limit_exceeded.rb
+- lib/langchain/utils/hash_transformer.rb
 - lib/langchain/vectorsearch/base.rb
 - lib/langchain/vectorsearch/chroma.rb
 - lib/langchain/vectorsearch/elasticsearch.rb
@@ -852,7 +834,7 @@ required_rubygems_version: !ruby/object:Gem::Requirement
     - !ruby/object:Gem::Version
       version: '0'
 requirements: []
-rubygems_version: 3.5.11
+rubygems_version: 3.5.14
 signing_key:
 specification_version: 4
 summary: Build LLM-backed Ruby applications with Ruby's Langchain.rb

data/lib/langchain/utils/token_length/ai21_validator.rb

@@ -1,41 +0,0 @@
-# frozen_string_literal: true
-
-module Langchain
-  module Utils
-    module TokenLength
-      #
-      # This class is meant to validate the length of the text passed in to AI21's API.
-      # It is used to validate the token length before the API call is made
-      #
-
-      class AI21Validator < BaseValidator
-        TOKEN_LIMITS = {
-          "j2-ultra" => 8192,
-          "j2-mid" => 8192,
-          "j2-light" => 8192
-        }.freeze
-
-        #
-        # Calculate token length for a given text and model name
-        #
-        # @param text [String] The text to calculate the token length for
-        # @param model_name [String] The model name to validate against
-        # @return [Integer] The token length of the text
-        #
-        def self.token_length(text, model_name, options = {})
-          res = options[:llm].tokenize(text)
-          res.dig(:tokens).length
-        end
-
-        def self.token_limit(model_name)
-          TOKEN_LIMITS[model_name]
-        end
-        singleton_class.alias_method :completion_token_limit, :token_limit
-
-        def self.token_length_from_messages(messages, model_name, options)
-          messages.sum { |message| token_length(message.to_json, model_name, options) }
-        end
-      end
-    end
-  end
-end

data/lib/langchain/utils/token_length/base_validator.rb

@@ -1,42 +0,0 @@
-# frozen_string_literal: true
-
-module Langchain
-  module Utils
-    module TokenLength
-      #
-      # Calculate the `max_tokens:` parameter to be set by calculating the context length of the text minus the prompt length
-      #
-      # @param content [String | Array<String>] The text or array of texts to validate
-      # @param model_name [String] The model name to validate against
-      # @return [Integer] Whether the text is valid or not
-      # @raise [TokenLimitExceeded] If the text is too long
-      #
-      class BaseValidator
-        def self.validate_max_tokens!(content, model_name, options = {})
-          text_token_length = if content.is_a?(Array)
-            token_length_from_messages(content, model_name, options)
-          else
-            token_length(content, model_name, options)
-          end
-
-          leftover_tokens = token_limit(model_name) - text_token_length
-
-          # Some models have a separate token limit for completions (e.g. GPT-4 Turbo)
-          # We want the lower of the two limits
-          max_tokens = [leftover_tokens, completion_token_limit(model_name)].min
-
-          # Raise an error even if whole prompt is equal to the model's token limit (leftover_tokens == 0)
-          if max_tokens < 0
-            raise limit_exceeded_exception(token_limit(model_name), text_token_length)
-          end
-
-          max_tokens
-        end
-
-        def self.limit_exceeded_exception(limit, length)
-          TokenLimitExceeded.new("This model's maximum context length is #{limit} tokens, but the given text is #{length} tokens long.", length - limit)
-        end
-      end
-    end
-  end
-end

data/lib/langchain/utils/token_length/cohere_validator.rb

@@ -1,49 +0,0 @@
-# frozen_string_literal: true
-
-module Langchain
-  module Utils
-    module TokenLength
-      #
-      # This class is meant to validate the length of the text passed in to Cohere's API.
-      # It is used to validate the token length before the API call is made
-      #
-
-      class CohereValidator < BaseValidator
-        TOKEN_LIMITS = {
-          # Source:
-          # https://docs.cohere.com/docs/models
-          "command-light" => 4096,
-          "command" => 4096,
-          "base-light" => 2048,
-          "base" => 2048,
-          "embed-english-light-v2.0" => 512,
-          "embed-english-v2.0" => 512,
-          "embed-multilingual-v2.0" => 256,
-          "summarize-medium" => 2048,
-          "summarize-xlarge" => 2048
-        }.freeze
-
-        #
-        # Calculate token length for a given text and model name
-        #
-        # @param text [String] The text to calculate the token length for
-        # @param model_name [String] The model name to validate against
-        # @return [Integer] The token length of the text
-        #
-        def self.token_length(text, model_name, options = {})
-          res = options[:llm].tokenize(text: text)
-          res["tokens"].length
-        end
-
-        def self.token_limit(model_name)
-          TOKEN_LIMITS[model_name]
-        end
-        singleton_class.alias_method :completion_token_limit, :token_limit
-
-        def self.token_length_from_messages(messages, model_name, options)
-          messages.sum { |message| token_length(message.to_json, model_name, options) }
-        end
-      end
-    end
-  end
-end

data/lib/langchain/utils/token_length/google_palm_validator.rb

@@ -1,57 +0,0 @@
-# frozen_string_literal: true
-
-module Langchain
-  module Utils
-    module TokenLength
-      #
-      # This class is meant to validate the length of the text passed in to Google Palm's API.
-      # It is used to validate the token length before the API call is made
-      #
-      class GooglePalmValidator < BaseValidator
-        TOKEN_LIMITS = {
-          # Source:
-          # This data can be pulled when `list_models()` method is called: https://github.com/andreibondarev/google_palm_api#usage
-
-          # chat-bison-001 is the only model that currently supports countMessageTokens functions
-          "chat-bison-001" => {
-            "input_token_limit" => 4000, # 4096 is the limit but the countMessageTokens does not return anything higher than 4000
-            "output_token_limit" => 1024
-          }
-          # "text-bison-001" => {
-          #   "input_token_limit" => 8196,
-          #   "output_token_limit" => 1024
-          # },
-          # "embedding-gecko-001" => {
-          #   "input_token_limit" => 1024
-          # }
-        }.freeze
-
-        #
-        # Calculate token length for a given text and model name
-        #
-        # @param text [String] The text to calculate the token length for
-        # @param model_name [String] The model name to validate against
-        # @param options [Hash] the options to create a message with
-        # @option options [Langchain::LLM:GooglePalm] :llm The Langchain::LLM:GooglePalm instance
-        # @return [Integer] The token length of the text
-        #
-        def self.token_length(text, model_name = "chat-bison-001", options = {})
-          response = options[:llm].client.count_message_tokens(model: model_name, prompt: text)
-
-          raise Langchain::LLM::ApiError.new(response["error"]["message"]) unless response["error"].nil?
-
-          response.dig("tokenCount")
-        end
-
-        def self.token_length_from_messages(messages, model_name, options = {})
-          messages.sum { |message| token_length(message.to_json, model_name, options) }
-        end
-
-        def self.token_limit(model_name)
-          TOKEN_LIMITS.dig(model_name, "input_token_limit")
-        end
-        singleton_class.alias_method :completion_token_limit, :token_limit
-      end
-    end
-  end
-end