Introduce Langchain::Assistant#parallel_tool_calls options whether …

…to allow the LLM to make multiple parallel tool calls (#827) * Implement assistant.parallel_tool_calls: true/false option * fix linter * Specs * specs * fix linter * CHANGELOG entry
patterns-ai-core · Oct 12, 2024 · 49161bb · 49161bb
1 parent fe33fa2
commit 49161bb
Show file tree

Hide file tree

Showing 25 changed files with 237 additions and 75 deletions.
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -2,6 +2,7 @@
 - [BREAKING] Remove `Langchain::Assistant#clear_thread!` method
 - [BREAKING] `Langchain::Messages::*` namespace had migrated to `Langchain::Assistant::Messages::*`
 - [BREAKING] Modify `Langchain::LLM::AwsBedrock` constructor to pass model options via default_options: {...}
+- Introduce `Langchain::Assistant#parallel_tool_calls` options whether to allow the LLM to make multiple parallel tool calls. Default: true
 - Minor improvements to the Langchain::Assistant class
 - Added support for streaming with Anthropic
 - Bump anthropic gem

diff --git a/README.md b/README.md
@@ -534,6 +534,7 @@ Note that streaming is not currently supported for all LLMs.
 * `tools`: An array of tool instances (optional)
 * `instructions`: System instructions for the assistant (optional)
 * `tool_choice`: Specifies how tools should be selected. Default: "auto". A specific tool function name can be passed. This will force the Assistant to **always** use this function.
+* `parallel_tool_calls`: Whether to make multiple parallel tool calls. Default: true
 * `add_message_callback`: A callback function (proc, lambda) that is called when any message is added to the conversation (optional)
 
 ### Key Methods

diff --git a/lib/langchain/assistant.rb b/lib/langchain/assistant.rb
@@ -12,22 +12,35 @@ module Langchain
   #       tools: [Langchain::Tool::NewsRetriever.new(api_key: ENV["NEWS_API_KEY"])]
   #     )
   class Assistant
-    attr_reader :llm, :instructions, :state, :llm_adapter, :tool_choice
-    attr_reader :total_prompt_tokens, :total_completion_tokens, :total_tokens, :messages
-    attr_accessor :tools, :add_message_callback
+    attr_reader :llm,
+      :instructions,
+      :state,
+      :llm_adapter,
+      :messages,
+      :tool_choice,
+      :total_prompt_tokens,
+      :total_completion_tokens,
+      :total_tokens
+
+    attr_accessor :tools,
+      :add_message_callback,
+      :parallel_tool_calls
 
     # Create a new assistant
     #
     # @param llm [Langchain::LLM::Base] LLM instance that the assistant will use
     # @param tools [Array<Langchain::Tool::Base>] Tools that the assistant has access to
     # @param instructions [String] The system instructions
     # @param tool_choice [String] Specify how tools should be selected. Options: "auto", "any", "none", or <specific function name>
-    # @params add_message_callback [Proc] A callback function (Proc or lambda) that is called when any message is added to the conversation
+    # @param parallel_tool_calls [Boolean] Whether or not to run tools in parallel
+    # @param messages [Array<Langchain::Assistant::Messages::Base>] The messages
+    # @param add_message_callback [Proc] A callback function (Proc or lambda) that is called when any message is added to the conversation
     def initialize(
       llm:,
       tools: [],
       instructions: nil,
       tool_choice: "auto",
+      parallel_tool_calls: true,
       messages: [],
       add_message_callback: nil,
       &block
@@ -47,6 +60,7 @@ def initialize(
 
       self.messages = messages
       @tools = tools
+      @parallel_tool_calls = parallel_tool_calls
       self.tool_choice = tool_choice
       self.instructions = instructions
       @block = block
@@ -326,7 +340,8 @@ def chat_with_llm
         instructions: @instructions,
         messages: array_of_message_hashes,
         tools: @tools,
-        tool_choice: tool_choice
+        tool_choice: tool_choice,
+        parallel_tool_calls: parallel_tool_calls
       )
       @llm.chat(**params, &@block)
     end

diff --git a/lib/langchain/assistant/llm/adapters/anthropic.rb b/lib/langchain/assistant/llm/adapters/anthropic.rb
@@ -7,16 +7,23 @@ module Adapters
         class Anthropic < Base
           # Build the chat parameters for the Anthropic API
           #
-          # @param tools [Array<Hash>] The tools to use
-          # @param instructions [String] The system instructions
           # @param messages [Array<Hash>] The messages
+          # @param instructions [String] The system instructions
+          # @param tools [Array<Hash>] The tools to use
           # @param tool_choice [String] The tool choice
+          # @param parallel_tool_calls [Boolean] Whether to make parallel tool calls
           # @return [Hash] The chat parameters
-          def build_chat_params(tools:, instructions:, messages:, tool_choice:)
+          def build_chat_params(
+            messages:,
+            instructions:,
+            tools:,
+            tool_choice:,
+            parallel_tool_calls:
+          )
             params = {messages: messages}
             if tools.any?
               params[:tools] = build_tools(tools)
-              params[:tool_choice] = build_tool_choice(tool_choice)
+              params[:tool_choice] = build_tool_choice(tool_choice, parallel_tool_calls)
             end
             params[:system] = instructions if instructions
             params
@@ -31,7 +38,7 @@ def build_chat_params(tools:, instructions:, messages:, tool_choice:)
           # @param tool_call_id [String] The tool call ID
           # @return [Messages::AnthropicMessage] The Anthropic message
           def build_message(role:, content: nil, image_url: nil, tool_calls: [], tool_call_id: nil)
-            warn "Image URL is not supported by Anthropic currently" if image_url
+            Langchain.logger.warn "WARNING: Image URL is not supported by Anthropic currently" if image_url
 
             Messages::AnthropicMessage.new(role: role, content: content, tool_calls: tool_calls, tool_call_id: tool_call_id)
           end
@@ -76,15 +83,20 @@ def support_system_message?
 
           private
 
-          def build_tool_choice(choice)
+          def build_tool_choice(choice, parallel_tool_calls)
+            tool_choice_object = {disable_parallel_tool_use: !parallel_tool_calls}
+
             case choice
             when "auto"
-              {type: "auto"}
+              tool_choice_object[:type] = "auto"
             when "any"
-              {type: "any"}
+              tool_choice_object[:type] = "any"
             else
-              {type: "tool", name: choice}
+              tool_choice_object[:type] = "tool"
+              tool_choice_object[:name] = choice
             end
+
+            tool_choice_object
           end
         end
       end

diff --git a/lib/langchain/assistant/llm/adapters/base.rb b/lib/langchain/assistant/llm/adapters/base.rb
@@ -7,12 +7,19 @@ module Adapters
         class Base
           # Build the chat parameters for the LLM
           #
-          # @param tools [Array] The tools to use
-          # @param instructions [String] The system instructions
           # @param messages [Array] The messages
+          # @param instructions [String] The system instructions
+          # @param tools [Array] The tools to use
           # @param tool_choice [String] The tool choice
+          # @param parallel_tool_calls [Boolean] Whether to make parallel tool calls
           # @return [Hash] The chat parameters
-          def build_chat_params(tools:, instructions:, messages:, tool_choice:)
+          def build_chat_params(
+            messages:,
+            instructions:,
+            tools:,
+            tool_choice:,
+            parallel_tool_calls:
+          )
             raise NotImplementedError, "Subclasses must implement build_chat_params"
           end
 

diff --git a/lib/langchain/assistant/llm/adapters/google_gemini.rb b/lib/langchain/assistant/llm/adapters/google_gemini.rb
@@ -7,12 +7,21 @@ module Adapters
         class GoogleGemini < Base
           # Build the chat parameters for the Google Gemini LLM
           #
-          # @param tools [Array] The tools to use
-          # @param instructions [String] The system instructions
           # @param messages [Array] The messages
+          # @param instructions [String] The system instructions
+          # @param tools [Array] The tools to use
           # @param tool_choice [String] The tool choice
+          # @param parallel_tool_calls [Boolean] Whether to make parallel tool calls
           # @return [Hash] The chat parameters
-          def build_chat_params(tools:, instructions:, messages:, tool_choice:)
+          def build_chat_params(
+            messages:,
+            instructions:,
+            tools:,
+            tool_choice:,
+            parallel_tool_calls:
+          )
+            Langchain.logger.warn "WARNING: `parallel_tool_calls:` is not supported by Google Gemini currently"
+
             params = {messages: messages}
             if tools.any?
               params[:tools] = build_tools(tools)
@@ -31,7 +40,7 @@ def build_chat_params(tools:, instructions:, messages:, tool_choice:)
           # @param tool_call_id [String] The tool call ID
           # @return [Messages::GoogleGeminiMessage] The Google Gemini message
           def build_message(role:, content: nil, image_url: nil, tool_calls: [], tool_call_id: nil)
-            warn "Image URL is not supported by Google Gemini" if image_url
+            Langchain.logger.warn "Image URL is not supported by Google Gemini" if image_url
 
             Messages::GoogleGeminiMessage.new(role: role, content: content, tool_calls: tool_calls, tool_call_id: tool_call_id)
           end

diff --git a/lib/langchain/assistant/llm/adapters/mistral_ai.rb b/lib/langchain/assistant/llm/adapters/mistral_ai.rb
@@ -7,12 +7,21 @@ module Adapters
         class MistralAI < Base
           # Build the chat parameters for the Mistral AI LLM
           #
-          # @param tools [Array] The tools to use
-          # @param instructions [String] The system instructions
           # @param messages [Array] The messages
+          # @param instructions [String] The system instructions
+          # @param tools [Array] The tools to use
           # @param tool_choice [String] The tool choice
+          # @param parallel_tool_calls [Boolean] Whether to make parallel tool calls
           # @return [Hash] The chat parameters
-          def build_chat_params(tools:, instructions:, messages:, tool_choice:)
+          def build_chat_params(
+            messages:,
+            instructions:,
+            tools:,
+            tool_choice:,
+            parallel_tool_calls:
+          )
+            Langchain.logger.warn "WARNING: `parallel_tool_calls:` is not supported by Mistral AI currently"
+
             params = {messages: messages}
             if tools.any?
               params[:tools] = build_tools(tools)

diff --git a/lib/langchain/assistant/llm/adapters/ollama.rb b/lib/langchain/assistant/llm/adapters/ollama.rb
@@ -7,12 +7,22 @@ module Adapters
         class Ollama < Base
           # Build the chat parameters for the Ollama LLM
           #
-          # @param tools [Array] The tools to use
-          # @param instructions [String] The system instructions
           # @param messages [Array] The messages
+          # @param instructions [String] The system instructions
+          # @param tools [Array] The tools to use
           # @param tool_choice [String] The tool choice
+          # @param parallel_tool_calls [Boolean] Whether to make parallel tool calls
           # @return [Hash] The chat parameters
-          def build_chat_params(tools:, instructions:, messages:, tool_choice:)
+          def build_chat_params(
+            messages:,
+            instructions:,
+            tools:,
+            tool_choice:,
+            parallel_tool_calls:
+          )
+            Langchain.logger.warn "WARNING: `parallel_tool_calls:` is not supported by Ollama currently"
+            Langchain.logger.warn "WARNING: `tool_choice:` is not supported by Ollama currently"
+
             params = {messages: messages}
             if tools.any?
               params[:tools] = build_tools(tools)
@@ -29,7 +39,7 @@ def build_chat_params(tools:, instructions:, messages:, tool_choice:)
           # @param tool_call_id [String] The tool call ID
           # @return [Messages::OllamaMessage] The Ollama message
           def build_message(role:, content: nil, image_url: nil, tool_calls: [], tool_call_id: nil)
-            warn "Image URL is not supported by Ollama currently" if image_url
+            Langchain.logger.warn "WARNING: Image URL is not supported by Ollama currently" if image_url
 
             Messages::OllamaMessage.new(role: role, content: content, tool_calls: tool_calls, tool_call_id: tool_call_id)
           end

diff --git a/lib/langchain/assistant/llm/adapters/openai.rb b/lib/langchain/assistant/llm/adapters/openai.rb
@@ -7,16 +7,24 @@ module Adapters
         class OpenAI < Base
           # Build the chat parameters for the OpenAI LLM
           #
-          # @param tools [Array] The tools to use
-          # @param instructions [String] The system instructions
           # @param messages [Array] The messages
+          # @param instructions [String] The system instructions
+          # @param tools [Array] The tools to use
           # @param tool_choice [String] The tool choice
+          # @param parallel_tool_calls [Boolean] Whether to make parallel tool calls
           # @return [Hash] The chat parameters
-          def build_chat_params(tools:, instructions:, messages:, tool_choice:)
+          def build_chat_params(
+            messages:,
+            instructions:,
+            tools:,
+            tool_choice:,
+            parallel_tool_calls:
+          )
             params = {messages: messages}
             if tools.any?
               params[:tools] = build_tools(tools)
               params[:tool_choice] = build_tool_choice(tool_choice)
+              params[:parallel_tool_calls] = parallel_tool_calls
             end
             params
           end

diff --git a/lib/langchain/llm/ai21.rb b/lib/langchain/llm/ai21.rb
@@ -8,7 +8,7 @@ module Langchain::LLM
   #   gem "ai21", "~> 0.2.1"
   #
   # Usage:
-  #     ai21 = Langchain::LLM::AI21.new(api_key: ENV["AI21_API_KEY"])
+  #     llm = Langchain::LLM::AI21.new(api_key: ENV["AI21_API_KEY"])
   #
   class AI21 < Base
     DEFAULTS = {

diff --git a/lib/langchain/llm/anthropic.rb b/lib/langchain/llm/anthropic.rb
@@ -8,7 +8,7 @@ module Langchain::LLM
   #   gem "anthropic", "~> 0.3.2"
   #
   # Usage:
-  #     anthropic = Langchain::LLM::Anthropic.new(api_key: ENV["ANTHROPIC_API_KEY"])
+  #     llm = Langchain::LLM::Anthropic.new(api_key: ENV["ANTHROPIC_API_KEY"])
   #
   class Anthropic < Base
     DEFAULTS = {

diff --git a/lib/langchain/llm/aws_bedrock.rb b/lib/langchain/llm/aws_bedrock.rb
@@ -7,7 +7,7 @@ module Langchain::LLM
   #    gem 'aws-sdk-bedrockruntime', '~> 1.1'
   #
   # Usage:
-  #    bedrock = Langchain::LLM::AwsBedrock.new(llm_options: {})
+  #    llm = Langchain::LLM::AwsBedrock.new(llm_options: {})
   #
   class AwsBedrock < Base
     DEFAULTS = {

diff --git a/lib/langchain/llm/azure.rb b/lib/langchain/llm/azure.rb
@@ -7,7 +7,7 @@ module Langchain::LLM
   #    gem "ruby-openai", "~> 6.3.0"
   #
   # Usage:
-  #    openai = Langchain::LLM::Azure.new(api_key:, llm_options: {}, embedding_deployment_url: chat_deployment_url:)
+  #    llm = Langchain::LLM::Azure.new(api_key:, llm_options: {}, embedding_deployment_url: chat_deployment_url:)
   #
   class Azure < OpenAI
     attr_reader :embed_client

diff --git a/lib/langchain/llm/hugging_face.rb b/lib/langchain/llm/hugging_face.rb
@@ -8,7 +8,7 @@ module Langchain::LLM
   #     gem "hugging-face", "~> 0.3.4"
   #
   # Usage:
-  #     hf = Langchain::LLM::HuggingFace.new(api_key: ENV["HUGGING_FACE_API_KEY"])
+  #     llm = Langchain::LLM::HuggingFace.new(api_key: ENV["HUGGING_FACE_API_KEY"])
   #
   class HuggingFace < Base
     DEFAULTS = {

diff --git a/lib/langchain/llm/ollama.rb b/lib/langchain/llm/ollama.rb
@@ -5,7 +5,6 @@ module Langchain::LLM
   # Available models: https://ollama.ai/library
   #
   # Usage:
-  #    llm = Langchain::LLM::Ollama.new
   #    llm = Langchain::LLM::Ollama.new(url: ENV["OLLAMA_URL"], default_options: {})
   #
   class Ollama < Base

diff --git a/lib/langchain/llm/openai.rb b/lib/langchain/llm/openai.rb
@@ -7,7 +7,7 @@ module Langchain::LLM
   #    gem "ruby-openai", "~> 6.3.0"
   #
   # Usage:
-  #    openai = Langchain::LLM::OpenAI.new(
+  #    llm = Langchain::LLM::OpenAI.new(
   #      api_key: ENV["OPENAI_API_KEY"],
   #      llm_options: {}, # Available options: https://github.com/alexrudall/ruby-openai/blob/main/lib/openai/client.rb#L5-L13
   #      default_options: {}
@@ -100,7 +100,7 @@ def embed(
     # @param params [Hash] The parameters to pass to the `chat()` method
     # @return [Langchain::LLM::OpenAIResponse] Response object
     def complete(prompt:, **params)
-      warn "DEPRECATED: `Langchain::LLM::OpenAI#complete` is deprecated, and will be removed in the next major version. Use `Langchain::LLM::OpenAI#chat` instead."
+      Langchain.logger.warn "DEPRECATED: `Langchain::LLM::OpenAI#complete` is deprecated, and will be removed in the next major version. Use `Langchain::LLM::OpenAI#chat` instead."
 
       if params[:stop_sequences]
         params[:stop] = params.delete(:stop_sequences)

diff --git a/lib/langchain/llm/parameters/chat.rb b/lib/langchain/llm/parameters/chat.rb
@@ -34,6 +34,7 @@ class Chat < SimpleDelegator
       # Function-calling
       tools: {default: []},
       tool_choice: {},
+      parallel_tool_calls: {},
 
       # Additional optional parameters
       logit_bias: {}

diff --git a/lib/langchain/llm/replicate.rb b/lib/langchain/llm/replicate.rb
@@ -7,16 +7,8 @@ module Langchain::LLM
   # Gem requirements:
   #     gem "replicate-ruby", "~> 0.2.2"
   #
-  # Use it directly:
-  #     replicate = Langchain::LLM::Replicate.new(api_key: ENV["REPLICATE_API_KEY"])
-  #
-  # Or pass it to be used by a vector search DB:
-  #     chroma = Langchain::Vectorsearch::Chroma.new(
-  #       url: ENV["CHROMA_URL"],
-  #       index_name: "...",
-  #       llm: replicate
-  #     )
-  #
+  # Usage:
+  #     llm = Langchain::LLM::Replicate.new(api_key: ENV["REPLICATE_API_KEY"])
   class Replicate < Base
     DEFAULTS = {
       # TODO: Figure out how to send the temperature to the API