langchainrb 0.5.1 → 0.5.2

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 2673339c5bbe874a8bdf1722a2556f26d9fe13394875af914b5203632714f2f0
-  data.tar.gz: 216ab880c2c6094b267cbf3efcaf19ce74bea7cc665442fbf2b23108a9cb087b
+  metadata.gz: d36de4206b792714ba9b6773c03272e9638b14caf7140e0bc00c3e767aa5fdef
+  data.tar.gz: 819fab9de55a34e4e6dc865febc19bb9979df55fa8fc6a753774cf1961c40103
 SHA512:
-  metadata.gz: 408cf6194d85a4af076adbfd8be4a360d094200d127672f218d8914fbcd67d1a8a803645219532f66d4e79214571d61b629570df071de871b013e2d9d6c0d3a5
-  data.tar.gz: 123016bd42d1d2539c13f7d68074ddc19dc8a5880ae0b02b103e20bf7f058adfe2659beb90263a852bbf38b4b169622a1b7ac8a245c3791ab9b9ae8f8fc4e3cb
+  metadata.gz: 6e180b41bbca96bd5523c276923f223bbebe470314086c6a909df440890793bcc70dbd66ecf59bf5d0fd52426650cc5d2684c56cc8fc643209cc1679527cbef4
+  data.tar.gz: af5db76c2b22b5c7bdc1170de437921e8464a16566f46a5cad465d69e6da47c97a82f7331a5ea5747840e58acc71463aa8456b03e9bc8851efda7b734e5d23cc

data/CHANGELOG.md

@@ -1,5 +1,9 @@
 ## [Unreleased]
 
+## [0.5.2] - 2023-06-07
+- 🗣️ LLMs
+  - Auto-calculate the max_tokens: setting to be passed on to OpenAI
+
 ## [0.5.1] - 2023-06-06
 - 🛠️ Tools
   - Modified Tool usage. Agents now accept Tools instances instead of Tool strings.

data/Gemfile.lock

@@ -1,7 +1,7 @@
 PATH
   remote: .
   specs:
-    langchainrb (0.5.1)
+    langchainrb (0.5.2)
       colorize (~> 0.8.1)
       tiktoken_ruby (~> 0.0.5)
 

data/README.md

@@ -281,11 +281,10 @@ Add `gem "sequel"` to your Gemfile
 ```ruby
 database = Langchain::Tool::Database.new(connection_string: "postgres://user:password@localhost:5432/db_name")
 
-agent = Langchain::Agent::SQLQueryAgent.new(llm: Langchain::LLM::OpenAI.new(api_key: ENV["OPENAI_API_KEY"]), tools: [database])
-
+agent = Langchain::Agent::SQLQueryAgent.new(llm: Langchain::LLM::OpenAI.new(api_key: ENV["OPENAI_API_KEY"]), db: database)
 ```
 ```ruby
-agent.ask(question: "How many users have a name with length greater than 5 in the users table?")
+agent.run(question: "How many users have a name with length greater than 5 in the users table?")
 #=> "14 users have a name with length greater than 5 in the users table."
 ```
 

data/examples/create_and_manage_few_shot_prompt_templates.rb

@@ -1,10 +1,10 @@
 require "langchain"
 
 # Create a prompt with a few shot examples
-prompt = Prompt::FewShotPromptTemplate.new(
+prompt = Langchain::Prompt::FewShotPromptTemplate.new(
   prefix: "Write antonyms for the following words.",
   suffix: "Input: {adjective}\nOutput:",
-  example_prompt: Prompt::PromptTemplate.new(
+  example_prompt: Langchain::Prompt::PromptTemplate.new(
     input_variables: ["input", "output"],
     template: "Input: {input}\nOutput: {output}"
   ),
@@ -32,5 +32,5 @@ prompt.format(adjective: "good")
 prompt.save(file_path: "spec/fixtures/prompt/few_shot_prompt_template.json")
 
 # Loading a new prompt template using a JSON file
-prompt = Prompt.load_from_path(file_path: "spec/fixtures/prompt/few_shot_prompt_template.json")
+prompt = Langchain::Prompt.load_from_path(file_path: "spec/fixtures/prompt/few_shot_prompt_template.json")
 prompt.prefix # "Write antonyms for the following words."

data/examples/create_and_manage_prompt_templates.rb

@@ -1,15 +1,15 @@
 require "langchain"
 
 # Create a prompt with one input variable
-prompt = Prompt::PromptTemplate.new(template: "Tell me a {adjective} joke.", input_variables: ["adjective"])
+prompt = Langchain::Prompt::PromptTemplate.new(template: "Tell me a {adjective} joke.", input_variables: ["adjective"])
 prompt.format(adjective: "funny") # "Tell me a funny joke."
 
 # Create a prompt with multiple input variables
-prompt = Prompt::PromptTemplate.new(template: "Tell me a {adjective} joke about {content}.", input_variables: ["adjective", "content"])
+prompt = Langchain::Prompt::PromptTemplate.new(template: "Tell me a {adjective} joke about {content}.", input_variables: ["adjective", "content"])
 prompt.format(adjective: "funny", content: "chickens") # "Tell me a funny joke about chickens."
 
 # Creating a PromptTemplate using just a prompt and no input_variables
-prompt = Prompt::PromptTemplate.from_template("Tell me a {adjective} joke about {content}.")
+prompt = Langchain::Prompt::PromptTemplate.from_template("Tell me a {adjective} joke about {content}.")
 prompt.input_variables # ["adjective", "content"]
 prompt.format(adjective: "funny", content: "chickens") # "Tell me a funny joke about chickens."
 
@@ -17,5 +17,9 @@ prompt.format(adjective: "funny", content: "chickens") # "Tell me a funny joke a
 prompt.save(file_path: "spec/fixtures/prompt/prompt_template.json")
 
 # Loading a new prompt template using a JSON file
-prompt = Prompt.load_from_path(file_path: "spec/fixtures/prompt/prompt_template.json")
+prompt = Langchain::Prompt.load_from_path(file_path: "spec/fixtures/prompt/prompt_template.json")
+prompt.input_variables # ["adjective", "content"]
+
+# Loading a new prompt template using a YAML file
+prompt = Langchain::Prompt.load_from_path(file_path: "spec/fixtures/prompt/prompt_template.yaml")
 prompt.input_variables # ["adjective", "content"]

data/examples/pdf_store_and_query_with_chroma.rb

@@ -4,7 +4,7 @@ require "langchain"
 # or add `gem "chroma-db", "~> 0.3.0"` to your Gemfile
 
 # Instantiate the Chroma client
-chroma = Vectorsearch::Chroma.new(
+chroma = Langchain::Vectorsearch::Chroma.new(
   url: ENV["CHROMA_URL"],
   index_name: "documents",
   llm: Langchain::LLM::OpenAI.new(api_key: ENV["OPENAI_API_KEY"])

data/examples/store_and_query_with_pinecone.rb

@@ -4,7 +4,7 @@ require "langchain"
 # or add `gem "pinecone"` to your Gemfile
 
 # Instantiate the Qdrant client
-pinecone = Vectorsearch::Pinecone.new(
+pinecone = Langchain::Vectorsearch::Pinecone.new(
   environment: ENV["PINECONE_ENVIRONMENT"],
   api_key: ENV["PINECONE_API_KEY"],
   index_name: "recipes",
@@ -37,7 +37,7 @@ pinecone.ask(
 )
 
 # Generate your an embedding and search by it
-openai = LLM::OpenAI.new(api_key: ENV["OPENAI_API_KEY"])
+openai = Langchain::LLM::OpenAI.new(api_key: ENV["OPENAI_API_KEY"])
 embedding = openai.embed(text: "veggie")
 
 pinecone.similarity_search_by_vector(

data/examples/store_and_query_with_qdrant.rb

@@ -4,7 +4,7 @@ require "langchain"
 # or add `gem "qdrant-ruby"` to your Gemfile
 
 # Instantiate the Qdrant client
-qdrant = Vectorsearch::Qdrant.new(
+qdrant = Langchain::Vectorsearch::Qdrant.new(
   url: ENV["QDRANT_URL"],
   api_key: ENV["QDRANT_API_KEY"],
   index_name: "recipes",

data/examples/store_and_query_with_weaviate.rb

@@ -4,7 +4,7 @@ require "langchain"
 # or add `gem "weaviate-ruby"` to your Gemfile
 
 # Instantiate the Weaviate client
-weaviate = Vectorsearch::Weaviate.new(
+weaviate = Langchain::Vectorsearch::Weaviate.new(
   url: ENV["WEAVIATE_URL"],
   api_key: ENV["WEAVIATE_API_KEY"],
   index_name: "Recipes",

data/lib/langchain/agent/chain_of_thought_agent/chain_of_thought_agent.rb

@@ -39,11 +39,8 @@ module Langchain::Agent
 
       loop do
         Langchain.logger.info("[#{self.class.name}]".red + ": Sending the prompt to the #{llm.class} LLM")
-        response = llm.complete(
-          prompt: prompt,
-          stop_sequences: ["Observation:"],
-          max_tokens: 500
-        )
+
+        response = llm.complete(prompt: prompt, stop_sequences: ["Observation:"])
 
         # Append the response to the prompt
         prompt += response

data/lib/langchain/agent/sql_query_agent/sql_query_agent.rb

@@ -22,12 +22,12 @@ module Langchain::Agent
     # @param question [String] Question to ask the LLM/Database
     # @return [String] Answer to the question
     #
-    def ask(question:)
+    def run(question:)
       prompt = create_prompt_for_sql(question: question)
 
       # Get the SQL string to execute
       Langchain.logger.info("[#{self.class.name}]".red + ":  Passing the inital prompt to the #{llm.class} LLM")
-      sql_string = llm.complete(prompt: prompt, max_tokens: 500)
+      sql_string = llm.complete(prompt: prompt)
 
       # Execute the SQL string and collect the results
       Langchain.logger.info("[#{self.class.name}]".red + ":  Passing the SQL to the Database: #{sql_string}")
@@ -36,7 +36,7 @@ module Langchain::Agent
       # Pass the results and get the LLM to synthesize the answer to the question
       Langchain.logger.info("[#{self.class.name}]".red + ":  Passing the synthesize prompt to the #{llm.class} LLM with results: #{results}")
       prompt2 = create_prompt_for_answer(question: question, sql_query: sql_string, results: results)
-      llm.complete(prompt: prompt2, max_tokens: 500)
+      llm.complete(prompt: prompt2)
     end
 
     private

data/lib/langchain/llm/openai.rb

@@ -35,7 +35,7 @@ module Langchain::LLM
     def embed(text:, **params)
       parameters = {model: DEFAULTS[:embeddings_model_name], input: text}
 
-      Langchain::Utils::TokenLengthValidator.validate!(text, parameters[:model])
+      Langchain::Utils::TokenLengthValidator.validate_max_tokens!(text, parameters[:model])
 
       response = client.embeddings(parameters: parameters.merge(params))
       response.dig("data").first.dig("embedding")
@@ -50,9 +50,8 @@ module Langchain::LLM
     def complete(prompt:, **params)
       parameters = compose_parameters DEFAULTS[:completion_model_name], params
 
-      Langchain::Utils::TokenLengthValidator.validate!(prompt, parameters[:model])
-
       parameters[:prompt] = prompt
+      parameters[:max_tokens] = Langchain::Utils::TokenLengthValidator.validate_max_tokens!(prompt, parameters[:model])
 
       response = client.completions(parameters: parameters)
       response.dig("choices", 0, "text")
@@ -67,9 +66,8 @@ module Langchain::LLM
     def chat(prompt:, **params)
       parameters = compose_parameters DEFAULTS[:chat_completion_model_name], params
 
-      Langchain::Utils::TokenLengthValidator.validate!(prompt, parameters[:model])
-
       parameters[:messages] = [{role: "user", content: prompt}]
+      parameters[:max_tokens] = Langchain::Utils::TokenLengthValidator.validate_max_tokens!(prompt, parameters[:model])
 
       response = client.chat(parameters: parameters)
       response.dig("choices", 0, "message", "content")
@@ -87,12 +85,7 @@ module Langchain::LLM
       )
       prompt = prompt_template.format(text: text)
 
-      complete(
-        prompt: prompt,
-        temperature: DEFAULTS[:temperature],
-        # Most models have a context length of 2048 tokens (except for the newest models, which support 4096).
-        max_tokens: 2048
-      )
+      complete(prompt: prompt, temperature: DEFAULTS[:temperature])
     end
 
     private

data/lib/langchain/tool/calculator.rb

@@ -38,7 +38,8 @@ module Langchain::Tool
       hash_results = Langchain::Tool::SerpApi
         .new(api_key: ENV["SERPAPI_API_KEY"])
         .execute_search(input: input)
-      hash_results.dig(:answer_box, :to)
+      hash_results.dig(:answer_box, :to) ||
+        hash_results.dig(:answer_box, :result)
     end
   end
 end

data/lib/langchain/utils/token_length_validator.rb

@@ -34,23 +34,50 @@ module Langchain
         "ada" => 2049
       }.freeze
 
+      # GOOGLE_PALM_TOKEN_LIMITS = {
+      #   "chat-bison-001" => {
+      #     "inputTokenLimit"=>4096,
+      #     "outputTokenLimit"=>1024
+      #   },
+      #   "text-bison-001" => {
+      #     "inputTokenLimit"=>8196,
+      #     "outputTokenLimit"=>1024
+      #   },
+      #   "embedding-gecko-001" => {
+      #     "inputTokenLimit"=>1024
+      #   }
+      # }.freeze
+
       #
-      # Validate the length of the text passed in to OpenAI's API
+      # Calculate the `max_tokens:` parameter to be set by calculating the context length of the text minus the prompt length
       #
       # @param text [String] The text to validate
       # @param model_name [String] The model name to validate against
-      # @return [Boolean] Whether the text is valid or not
+      # @return [Integer] Whether the text is valid or not
       # @raise [TokenLimitExceeded] If the text is too long
       #
-      def self.validate!(text, model_name)
-        encoder = Tiktoken.encoding_for_model(model_name)
-        token_length = encoder.encode(text).length
+      def self.validate_max_tokens!(text, model_name)
+        text_token_length = token_length(text, model_name)
+        max_tokens = TOKEN_LIMITS[model_name] - text_token_length
 
-        if token_length > TOKEN_LIMITS[model_name]
-          raise TokenLimitExceeded, "This model's maximum context length is #{TOKEN_LIMITS[model_name]} tokens, but the given text is #{token_length} tokens long."
+        # Raise an error even if whole prompt is equal to the model's token limit (max_tokens == 0) since not response will be returned
+        if max_tokens <= 0
+          raise TokenLimitExceeded, "This model's maximum context length is #{TOKEN_LIMITS[model_name]} tokens, but the given text is #{text_token_length} tokens long."
         end
 
-        true
+        max_tokens
+      end
+
+      #
+      # Calculate token length for a given text and model name
+      #
+      # @param text [String] The text to validate
+      # @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)
+        encoder = Tiktoken.encoding_for_model(model_name)
+        encoder.encode(text).length
       end
     end
   end

data/lib/langchain/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Langchain
-  VERSION = "0.5.1"
+  VERSION = "0.5.2"
 end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: langchainrb
 version: !ruby/object:Gem::Version
-  version: 0.5.1
+  version: 0.5.2
 platform: ruby
 authors:
 - Andrei Bondarev
 autorequire:
 bindir: exe
 cert_chain: []
-date: 2023-06-06 00:00:00.000000000 Z
+date: 2023-06-08 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: tiktoken_ruby