boxcars 0.7.4 → 0.7.6

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: b7d7e68a8fe277fba9e416c7f582ee60135ef79f101da5f8da802a900330e807
-  data.tar.gz: 7f7e7f5243dc3a3f646dee5475456b9df640773e6411053b8f8514a5b9ef95be
+  metadata.gz: 998fff351b071b938cb5ba4cb0427b7448ffbb46e71b5272761bac45625dca3a
+  data.tar.gz: c6c945e4cabc226f4eb37e1368b77276a18023a26e1e05a5d60bb02742b3ed05
 SHA512:
-  metadata.gz: 91af273bb15b972e59d0c1c2f142e19fbfc226298ce650da55e7925054e3c9ea2076f2c7ca9e741b2280cea33a913be27bf7f65681d61f6fbfa02d8361726e00
-  data.tar.gz: e4f6d515b34bc9ba1b12b70f7cd0813930b825675b93d263e991858d8c774ffc2030ce27264aac3c458f921fede365651a8344c1a4f279852b0c73c30052e39b
+  metadata.gz: 4e4fad15d9c08ff76efec828697b387241534114f4594605ce5f257ce9de4d354f77aa5481f36961158b97399dba8d7913732b7644984eb864994103bab38839
+  data.tar.gz: 329d630f296a6e193633efe238e04aeb8ad54ebd4f378ee71467caa373786289529d40c9e0d1337dfb7b0b7acc7c94dc9cbe9cd9a8c3b1178bed7680d1d0aeb7

data/.rubocop.yml

@@ -1,7 +1,10 @@
 require:
-  - rubocop-rspec
   - rubocop-rake
 
+plugins:
+  - rubocop-rspec
+
+
 AllCops:
   TargetRubyVersion: 3
   Exclude:

data/Gemfile.lock

@@ -1,7 +1,7 @@
 PATH
   remote: .
   specs:
-    boxcars (0.7.4)
+    boxcars (0.7.6)
       google_search_results (~> 2.2)
       gpt4all (~> 0.0.5)
       hnswlib (~> 0.9)
@@ -58,7 +58,7 @@ GEM
     benchmark (0.4.0)
     bigdecimal (3.1.9)
     concurrent-ruby (1.3.5)
-    connection_pool (2.5.0)
+    connection_pool (2.5.1)
     console (1.30.2)
       fiber-annotation
       fiber-local (~> 1.1)
@@ -91,7 +91,7 @@ GEM
     fiber-annotation (0.2.0)
     fiber-local (1.1.0)
       fiber-storage
-    fiber-storage (1.0.0)
+    fiber-storage (1.0.1)
     github_changelog_generator (1.16.4)
       activesupport
       async (>= 1.25.0)
@@ -131,7 +131,7 @@ GEM
     mime-types (3.6.2)
       logger
       mime-types-data (~> 3.2015)
-    mime-types-data (3.2025.0408)
+    mime-types-data (3.2025.0415)
     minitest (5.25.5)
     multi_json (1.15.0)
     multipart-post (2.4.1)
@@ -139,27 +139,27 @@ GEM
       uri
     netrc (0.11.0)
     nio4r (2.7.4)
-    nokogiri (1.18.7-aarch64-linux-gnu)
+    nokogiri (1.18.8-aarch64-linux-gnu)
       racc (~> 1.4)
-    nokogiri (1.18.7-aarch64-linux-musl)
+    nokogiri (1.18.8-aarch64-linux-musl)
       racc (~> 1.4)
-    nokogiri (1.18.7-arm-linux-gnu)
+    nokogiri (1.18.8-arm-linux-gnu)
       racc (~> 1.4)
-    nokogiri (1.18.7-arm-linux-musl)
+    nokogiri (1.18.8-arm-linux-musl)
       racc (~> 1.4)
-    nokogiri (1.18.7-arm64-darwin)
+    nokogiri (1.18.8-arm64-darwin)
       racc (~> 1.4)
-    nokogiri (1.18.7-x86_64-darwin)
+    nokogiri (1.18.8-x86_64-darwin)
       racc (~> 1.4)
-    nokogiri (1.18.7-x86_64-linux-gnu)
+    nokogiri (1.18.8-x86_64-linux-gnu)
       racc (~> 1.4)
-    nokogiri (1.18.7-x86_64-linux-musl)
+    nokogiri (1.18.8-x86_64-linux-musl)
       racc (~> 1.4)
     octokit (4.25.1)
       faraday (>= 1, < 3)
       sawyer (~> 0.9)
     os (1.1.4)
-    parallel (1.26.3)
+    parallel (1.27.0)
     parser (3.3.8.0)
       ast (~> 2.4.1)
       racc
@@ -223,7 +223,7 @@ GEM
       prism (~> 1.4)
     rubocop-rake (0.6.0)
       rubocop (~> 1.0)
-    rubocop-rspec (3.5.0)
+    rubocop-rspec (3.6.0)
       lint_roller (~> 1.1)
       rubocop (~> 1.72, >= 1.72.1)
     ruby-anthropic (0.4.2)
@@ -247,7 +247,7 @@ GEM
     sqlite3 (2.6.0-x86_64-darwin)
     sqlite3 (2.6.0-x86_64-linux-gnu)
     sqlite3 (2.6.0-x86_64-linux-musl)
-    stringio (3.1.6)
+    stringio (3.1.7)
     strings-ansi (0.2.0)
     timeout (0.4.3)
     timers (4.4.0)

data/lib/boxcars/boxcar/active_record.rb

@@ -33,6 +33,30 @@ module Boxcars
       { model_info: model_info }.merge super
     end
 
+    CTEMPLATE = [
+      syst("You are a Ruby on Rails Active Record code generator"),
+      syst("Given an input question, first create a syntactically correct Rails Active Record code to run, ",
+           "then look at the results of the code and return the answer. Unless the user specifies ",
+           "in her question a specific number of examples she wishes to obtain, limit your code ",
+           "to at most %<top_k>s results.\n",
+           "Never query for all the columns from a specific model, ",
+           "only ask for the relevant attributes given the question.\n",
+           "Also, pay attention to which attribute is in which model.\n\n",
+           "Use the following format:\n",
+           "Question: ${{Question here}}\n",
+           "ARChanges: ${{Active Record code to compute the number of records going to change}} - ",
+           "Only add this line if the ARCode on the next line will make data changes.\n",
+           "ARCode: ${{Active Record code to run}} - make sure you use valid code\n",
+           "Answer: ${{Final answer here}}\n\n",
+           "Only use the following Active Record models: %<model_info>s\n",
+           "Pay attention to use only the attribute names that you can see in the model description.\n",
+           "Do not make up variable or attribute names, and do not share variables between the code in ARChanges and ARCode\n",
+           "Be careful to not query for attributes that do not exist, and to use the format specified above.\n",
+           "Finally, try not to use print or puts in your code"
+          ),
+      user("Question: %<question>s")
+    ].freeze
+
     private
 
     def get_name
@@ -240,30 +264,6 @@ module Boxcars
       end
     end
 
-    CTEMPLATE = [
-      syst("You are a Ruby on Rails Active Record code generator"),
-      syst("Given an input question, first create a syntactically correct Rails Active Record code to run, ",
-           "then look at the results of the code and return the answer. Unless the user specifies ",
-           "in her question a specific number of examples she wishes to obtain, limit your code ",
-           "to at most %<top_k>s results.\n",
-           "Never query for all the columns from a specific model, ",
-           "only ask for the relevant attributes given the question.\n",
-           "Also, pay attention to which attribute is in which model.\n\n",
-           "Use the following format:\n",
-           "Question: ${{Question here}}\n",
-           "ARChanges: ${{Active Record code to compute the number of records going to change}} - ",
-           "Only add this line if the ARCode on the next line will make data changes.\n",
-           "ARCode: ${{Active Record code to run}} - make sure you use valid code\n",
-           "Answer: ${{Final answer here}}\n\n",
-           "Only use the following Active Record models: %<model_info>s\n",
-           "Pay attention to use only the attribute names that you can see in the model description.\n",
-           "Do not make up variable or attribute names, and do not share variables between the code in ARChanges and ARCode\n",
-           "Be careful to not query for attributes that do not exist, and to use the format specified above.\n",
-           "Finally, try not to use print or puts in your code"
-          ),
-      user("Question: %<question>s")
-    ].freeze
-
     # The prompt to use for the engine.
     def my_prompt
       @conversation ||= Conversation.new(lines: CTEMPLATE)

data/lib/boxcars/boxcar/calculator.rb

@@ -18,27 +18,6 @@ module Boxcars
       super(engine: engine, prompt: the_prompt, **kwargs)
     end
 
-    private
-
-    def get_embedded_ruby_answer(text)
-      code = text.split("```ruby\n").last.split("```").first.strip
-      # code = text[8..-4].split("```").first.strip
-      ruby_executor = Boxcars::RubyREPL.new
-      ruby_executor.call(code: code)
-    end
-
-    def get_answer(text)
-      case text
-      when /^```ruby/
-        get_embedded_ruby_answer(text)
-      when /^Answer:/
-        Result.from_text(text)
-      else
-        Result.new(status: :error,
-                   explanation: "Error: expecting your response to begin with '```ruby'. Try answering the question again.")
-      end
-    end
-
     # our template
     CTEMPLATE = [
       syst("You can do basic math, but for any hard calculations that a human could not do ",
@@ -62,6 +41,27 @@ module Boxcars
       user("%<question>s")
     ].freeze
 
+    private
+
+    def get_embedded_ruby_answer(text)
+      code = text.split("```ruby\n").last.split("```").first.strip
+      # code = text[8..-4].split("```").first.strip
+      ruby_executor = Boxcars::RubyREPL.new
+      ruby_executor.call(code: code)
+    end
+
+    def get_answer(text)
+      case text
+      when /^```ruby/
+        get_embedded_ruby_answer(text)
+      when /^Answer:/
+        Result.from_text(text)
+      else
+        Result.new(status: :error,
+                   explanation: "Error: expecting your response to begin with '```ruby'. Try answering the question again.")
+      end
+    end
+
     # The prompt to use for the engine.
     def my_prompt
       @conversation ||= Conversation.new(lines: CTEMPLATE)

data/lib/boxcars/boxcar/google_search.rb

@@ -41,8 +41,6 @@ module Boxcars
       answer
     end
 
-    private
-
     ANSWER_LOCATIONS = [
       %i[answer_box answer],
       %i[answer_box snippet],
@@ -54,6 +52,8 @@ module Boxcars
       [:organic_results, 0, :title]
     ].freeze
 
+    private
+
     def find_answer(res)
       raise Error, "Got error from SerpAPI: #{res[:error]}" if res[:error]
 

data/lib/boxcars/boxcar/sql_base.rb

@@ -32,6 +32,24 @@ module Boxcars
       { schema: schema, dialect: dialect }.merge super
     end
 
+    CTEMPLATE = [
+      syst("Given an input question, first create a syntactically correct %<dialect>s SQL query to run, ",
+           "then look at the results of the query and return the answer. Unless the user specifies ",
+           "in her question a specific number of examples he wishes to obtain, always limit your query ",
+           "to at most %<top_k>s results using a LIMIT clause. You can order the results by a relevant column ",
+           "to return the most interesting examples in the database.\n",
+           "Never query for all the columns from a specific table, only ask for the elevant columns given the question.\n",
+           "Pay attention to use only the column names that you can see in the schema description. Be careful to ",
+           "not query for columns that do not exist. Also, pay attention to which column is in which table.\n",
+           "Use the following format:\n",
+           "Question: 'Question here'\n",
+           "SQLQuery: 'SQL Query to run'\n",
+           "SQLResult: 'Result of the SQLQuery'\n",
+           "Answer: 'Final answer here'"),
+      syst("Only use the following tables:\n%<schema>s"),
+      user("Question: %<question>s")
+    ].freeze
+
     private
 
     def check_tables(rtables, exceptions)
@@ -106,24 +124,6 @@ module Boxcars
       end
     end
 
-    CTEMPLATE = [
-      syst("Given an input question, first create a syntactically correct %<dialect>s SQL query to run, ",
-           "then look at the results of the query and return the answer. Unless the user specifies ",
-           "in her question a specific number of examples he wishes to obtain, always limit your query ",
-           "to at most %<top_k>s results using a LIMIT clause. You can order the results by a relevant column ",
-           "to return the most interesting examples in the database.\n",
-           "Never query for all the columns from a specific table, only ask for the elevant columns given the question.\n",
-           "Pay attention to use only the column names that you can see in the schema description. Be careful to ",
-           "not query for columns that do not exist. Also, pay attention to which column is in which table.\n",
-           "Use the following format:\n",
-           "Question: 'Question here'\n",
-           "SQLQuery: 'SQL Query to run'\n",
-           "SQLResult: 'Result of the SQLQuery'\n",
-           "Answer: 'Final answer here'"),
-      syst("Only use the following tables:\n%<schema>s"),
-      user("Question: %<question>s")
-    ].freeze
-
     # The prompt to use for the engine.
     def my_prompt
       @conversation ||= Conversation.new(lines: CTEMPLATE)

data/lib/boxcars/boxcar/swagger.rb

@@ -29,6 +29,24 @@ module Boxcars
       { swagger_url: swagger_url, context: context }.merge super
     end
 
+    # our template
+    CTEMPLATE = [
+      syst("Study this Open API Swagger file %<swagger_url>s\n",
+           "and write a Ruby Program that prints the answer to the following questions using the appropriate API calls:\n",
+           "Additional context that you might need in the Ruby program: (%<context>s)\n",
+           "Use the following format:\n",
+           "${{Question needing API calls and code}}\n",
+           "reply only with the following format:\n",
+           "```ruby\n${{Ruby code with API calls and code that prints the answer}}\n```\n",
+           "```output\n${{Output of your code}}\n```\n\n",
+           "Otherwise, if you know the answer and do not need any API calls, you should use this simpler format:\n",
+           "${{Question not needing API calls}}\n",
+           "Answer: ${{Answer}}\n\n",
+           "Do not give an explanation of the answer and make sure your answer starts with either 'Answer:' or '```ruby'. ",
+           "Make use of the rest-client gem to make your requests to the API. Just print the answer."),
+      user("%<question>s")
+    ].freeze
+
     private
 
     def get_embedded_ruby_answer(text)
@@ -49,24 +67,6 @@ module Boxcars
       end
     end
 
-    # our template
-    CTEMPLATE = [
-      syst("Study this Open API Swagger file %<swagger_url>s\n",
-           "and write a Ruby Program that prints the answer to the following questions using the appropriate API calls:\n",
-           "Additional context that you might need in the Ruby program: (%<context>s)\n",
-           "Use the following format:\n",
-           "${{Question needing API calls and code}}\n",
-           "reply only with the following format:\n",
-           "```ruby\n${{Ruby code with API calls and code that prints the answer}}\n```\n",
-           "```output\n${{Output of your code}}\n```\n\n",
-           "Otherwise, if you know the answer and do not need any API calls, you should use this simpler format:\n",
-           "${{Question not needing API calls}}\n",
-           "Answer: ${{Answer}}\n\n",
-           "Do not give an explanation of the answer and make sure your answer starts with either 'Answer:' or '```ruby'. ",
-           "Make use of the rest-client gem to make your requests to the API. Just print the answer."),
-      user("%<question>s")
-    ].freeze
-
     # The prompt to use for the engine.
     def my_prompt
       @conversation ||= Conversation.new(lines: CTEMPLATE)

data/lib/boxcars/boxcar/vector_answer.rb

@@ -30,6 +30,14 @@ module Boxcars
       { search_content: get_search_content(inputs[:question]) }.merge super
     end
 
+    # our template
+    CTEMPLATE = [
+      syst("You are tasked with answering a question using these possibly relevant excerpts from a large volume of text:\n" \
+           "```text\n%<search_content>s\n```\n\n",
+           "Using the above, just answer the question as if you were answering directly."),
+      user("%<question>s")
+    ].freeze
+
     private
 
     # @param results [Array] The results from the vector search.
@@ -50,14 +58,6 @@ module Boxcars
       @search_content = get_results_content(results)
     end
 
-    # our template
-    CTEMPLATE = [
-      syst("You are tasked with answering a question using these possibly relevant excerpts from a large volume of text:\n" \
-           "```text\n%<search_content>s\n```\n\n",
-           "Using the above, just answer the question as if you were answering directly."),
-      user("%<question>s")
-    ].freeze
-
     # The prompt to use for the engine.
     def my_prompt
       @conversation ||= Conversation.new(lines: CTEMPLATE)

data/lib/boxcars/engine/openai.rb

@@ -66,10 +66,7 @@ module Boxcars
           params.delete(:response_format)
           params.delete(:stop)
         end
-        params = prompt.as_messages(inputs).merge(params)
-        if Boxcars.configuration.log_prompts
-          Boxcars.debug(params[:messages].last(2).map { |p| ">>>>>> Role: #{p[:role]} <<<<<<\n#{p[:content]}" }.join("\n"), :cyan)
-        end
+        params = get_params(prompt, inputs, params)
         clnt.chat(parameters: params)
       else
         params = prompt.as_prompt(inputs: inputs).merge(params)
@@ -151,4 +148,14 @@ module Boxcars
     max_size = modelname_to_contextsize(model_name)
     max_size - num_tokens
   end
+
+  def get_params(prompt, inputs, params)
+    params = prompt.as_messages(inputs).merge(params)
+    # Handle models like o1-mini that don't support the system role
+    params[:messages].first[:role] = :user if params[:model] =~ /^o/ && params[:messages].first&.fetch(:role) == :system
+    if Boxcars.configuration.log_prompts
+      Boxcars.debug(params[:messages].last(2).map { |p| ">>>>>> Role: #{p[:role]} <<<<<<\n#{p[:content]}" }.join("\n"), :cyan)
+    end
+    params
+  end
 end

data/lib/boxcars/engine/together.rb

@@ -0,0 +1,36 @@
+# frozen_string_literal: true
+
+module Boxcars
+  # A engine that uses Google's API
+  class Together < IntelligenceBase
+    # The default parameters to use when asking the engine
+    DEFAULT_PARAMS = {
+      model: "deepseek-ai/DeepSeek-R1-Distill-Llama-70B",
+      temperature: 0.1
+    }.freeze
+
+    # the default name of the engine
+    DEFAULT_NAME = "DeepSeek R1 Distill Llama 70B AI engine"
+    # the default description of the engine
+    DEFAULT_DESCRIPTION = "useful for when you need to use DeepSeek AI to process complex content. " \
+                          "Supports text, images, and other content types"
+
+    # A DeepSeek Engine is used by Boxcars to generate output from prompts
+    # @param name [String] The name of the Engine. Defaults to classname.
+    # @param description [String] A description of the Engine.
+    # @param prompts [Array<Prompt>] The prompts to use for the Engine.
+    # @param batch_size [Integer] The number of prompts to send to the Engine at a time.
+    # @param kwargs [Hash] Additional parameters to pass to the Engine.
+    def initialize(name: DEFAULT_NAME, description: DEFAULT_DESCRIPTION, prompts: [], batch_size: 20, **kwargs)
+      super(provider: :together_ai, description: description, name: name, prompts: prompts, batch_size: batch_size, **kwargs)
+    end
+
+    def default_model_params
+      DEFAULT_PARAMS
+    end
+
+    def lookup_provider_api_key(params:)
+      Boxcars.configuration.together_api_key(**params)
+    end
+  end
+end

data/lib/boxcars/engine.rb

@@ -89,3 +89,4 @@ require "boxcars/engine/gemini_ai"
 require "boxcars/engine/intelligence_base"
 require "boxcars/engine/cerebras"
 require "boxcars/engine/google"
+require "boxcars/engine/together"

data/lib/boxcars/train/xml_zero_shot.rb

@@ -22,8 +22,6 @@ module Boxcars
       super(engine: engine, boxcars: boxcars, prompt: prompt, name: name, description: description, **kwargs)
     end
 
-    private
-
     CTEMPLATE = [
       syst("<training>Answer the following questions as best you can. You have access to the following tools for actions:\n",
            "%<boxcars_xml>s",
@@ -47,6 +45,8 @@ module Boxcars
       assi("%<agent_scratchpad>s")
     ].freeze
 
+    private
+
     # The prompt to use for the train.
     def my_prompt
       @conversation ||= Conversation.new(lines: CTEMPLATE)

data/lib/boxcars/train/zero_shot.rb

@@ -33,11 +33,35 @@ module Boxcars
       [:error, e.message]
     end
 
-    private
-
     # the final answer action string
     FINAL_ANSWER_ACTION = "Final Answer:"
 
+    CTEMPLATE = [
+      syst("Answer the following questions as best you can. You have access to the following actions:\n",
+           "%<boxcar_descriptions>s\n",
+           "Use the following format:\n",
+           "Question: the input question you must answer\n",
+           "Thought: you should always think about what to do\n",
+           "Action: the action to take, should be one from this list: %<boxcar_names>s\n",
+           "Action Input: an input question to the action\n",
+           "Observation: the result of the action\n",
+           "... (this Thought/Action/Action Input/Observation sequence can repeat N times)\n",
+           "Thought: I know the final answer\n",
+           "Final Answer: the final answer to the original input question\n",
+           "%<next_actions>s\n",
+           "Remember to start a line with \"Final Answer:\" to give me the final answer.\n",
+           "Also make sure to specify a question for the Action Input.\n",
+           "Finally, if you can deduct the answer from the question or observation, you can ",
+           "start with \"Final Answer:\" and give me the answer.\n",
+           "Begin!"),
+      # insert thoughts here from previous runs
+      hist,
+      user("Question: %<input>s"),
+      assi("Thought: %<agent_scratchpad>s")
+    ].freeze
+
+    private
+
     # Parse out the action and input from the engine output.
     # @param engine_output [String] The output from the engine.
     # @return [Array<String>] The action and input.
@@ -70,30 +94,6 @@ module Boxcars
       end
     end
 
-    CTEMPLATE = [
-      syst("Answer the following questions as best you can. You have access to the following actions:\n",
-           "%<boxcar_descriptions>s\n",
-           "Use the following format:\n",
-           "Question: the input question you must answer\n",
-           "Thought: you should always think about what to do\n",
-           "Action: the action to take, should be one from this list: %<boxcar_names>s\n",
-           "Action Input: an input question to the action\n",
-           "Observation: the result of the action\n",
-           "... (this Thought/Action/Action Input/Observation sequence can repeat N times)\n",
-           "Thought: I know the final answer\n",
-           "Final Answer: the final answer to the original input question\n",
-           "%<next_actions>s\n",
-           "Remember to start a line with \"Final Answer:\" to give me the final answer.\n",
-           "Also make sure to specify a question for the Action Input.\n",
-           "Finally, if you can deduct the answer from the question or observation, you can ",
-           "start with \"Final Answer:\" and give me the answer.\n",
-           "Begin!"),
-      # insert thoughts here from previous runs
-      hist,
-      user("Question: %<input>s"),
-      assi("Thought: %<agent_scratchpad>s")
-    ].freeze
-
     # The prompt to use for the train.
     def my_prompt
       @conversation ||= Conversation.new(lines: CTEMPLATE)

data/lib/boxcars/version.rb

@@ -2,5 +2,5 @@
 
 module Boxcars
   # The current version of the gem.
-  VERSION = "0.7.4"
+  VERSION = "0.7.6"
 end

data/lib/boxcars.rb

@@ -72,6 +72,11 @@ module Boxcars
       key_lookup(:gemini_api_key, kwargs)
     end
 
+    # @return [String] The Together AI API key either from arg or env.
+    def together_api_key(**kwargs)
+      key_lookup(:together_api_key, kwargs)
+    end
+
     private
 
     def check_key(key, val)

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: boxcars
 version: !ruby/object:Gem::Version
-  version: 0.7.4
+  version: 0.7.6
 platform: ruby
 authors:
 - Francis Sullivan
@@ -9,7 +9,7 @@ authors:
 autorequire:
 bindir: exe
 cert_chain: []
-date: 2025-04-16 00:00:00.000000000 Z
+date: 2025-04-22 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: google_search_results
@@ -176,6 +176,7 @@ files:
 - lib/boxcars/engine/ollama.rb
 - lib/boxcars/engine/openai.rb
 - lib/boxcars/engine/perplexityai.rb
+- lib/boxcars/engine/together.rb
 - lib/boxcars/generation.rb
 - lib/boxcars/observation.rb
 - lib/boxcars/prompt.rb