llm.rb 0.14.2 → 0.15.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 41efcc6a142e5863b3b1d2ab1b1c8a452b5c8e8098bf48918c3bd5b16abfcdd2
-  data.tar.gz: d2bfdfd690070d6dbca8411874d2970e470b37ed67139976026120b2806e3421
+  metadata.gz: 68bbdddd157d6df71729378a46b4759140d7fe13c44a1e14940a0a3a367d277c
+  data.tar.gz: f24aa6b6042b58857ca419c0039422f678e92355f5fdfd0064d2931108338866
 SHA512:
-  metadata.gz: 7a10131e21e424e4620b155ef4e8fd23617aecc5ee9141e6a6f0e09b99b042286992f6b27299502f0d7664a92c62e520bf99a7f8aa0836ba0973292c74ab5d80
-  data.tar.gz: 57e5110fc829a790ae064c5a52b2ae856fd5566b5256444148008dd9f3dc170b3cbf57fda0c09f7ecc5630215c0c7e236375d063943d9b791a034359569c7ecb
+  metadata.gz: e5043707425445ea5709f4eed33b3feab33a376a66569abe46b6ad93274ffd3044eef1099625ec75f24f5ee5fb5387493adc97332c7be1d6fa2e085d6e33281c
+  data.tar.gz: 6f60c60130904dd23c67d65d50a04614c32659d5fe7af5e37ec2f9d6f25ccda1712a22bf872a7ae4c4aae49c2755cb1adff593fe1435c4231f445c40761a6477

data/README.md

@@ -5,10 +5,6 @@ includes OpenAI, Gemini, Anthropic, xAI (grok), DeepSeek, Ollama, and
 LlamaCpp. The toolkit includes full support for chat, streaming, tool calling,
 audio, images, files, and structured outputs (JSON Schema).
 
-The library provides a common, uniform interface for all the providers and
-features it supports, in addition to provider-specific features as well. Keep
-reading to find out more.
-
 ## Quick start
 
 #### Demo
@@ -29,8 +25,8 @@ reading to find out more.
 
 #### Ecosystem
 
-* [llm-shell](https://github.com/llmrb/llm-shell) – is a developer-oriented console for Large Language Model communication
-* [llm-spell](https://github.com/llmrb/llm-spell) – is a utility that can correct spelling mistakes with a Large Language Model
+* [llm-shell](https://github.com/llmrb/llm-shell) – a developer-oriented console for Large Language Model communication
+* [llm-spell](https://github.com/llmrb/llm-spell) – a utility that can correct spelling mistakes with a Large Language Model
 
 ## Features
 
@@ -138,15 +134,14 @@ require "llm"
 llm  = LLM.openai(key: ENV["KEY"])
 bot  = LLM::Bot.new(llm)
 url  = "https://en.wikipedia.org/wiki/Special:FilePath/Cognac_glass.jpg"
-msgs = bot.chat do |prompt|
-  prompt.system "Your task is to answer all user queries"
-  prompt.user ["Tell me about this URL", URI(url)]
-  prompt.user ["Tell me about this PDF", File.open("handbook.pdf", "rb")]
-  prompt.user "Are the URL and PDF similar to each other?"
-end
+
+bot.chat "Your task is to answer all user queries", role: :system
+bot.chat ["Tell me about this URL", URI(url)], role: :user
+bot.chat ["Tell me about this PDF", File.open("handbook.pdf", "rb")], role: :user
+bot.chat "Are the URL and PDF similar to each other?", role: :user
 
 # At this point, we execute a single request
-msgs.each { print "[#{_1.role}] ", _1.content, "\n" }
+bot.messages.each { print "[#{_1.role}] ", _1.content, "\n" }
 ```
 
 #### Streaming
@@ -528,12 +523,6 @@ over or doesn't cover at all. The API reference is available at
   a blog post that implements image editing with Gemini
 * [docs/](docs/) – the docs directory contains additional guides
 
-
-## See also
-
-* [llm-shell](https://github.com/llmrb/llm-shell) – a shell that uses llm.rb to
-  provide a command-line interface to LLMs.
-
 ## Install
 
 llm.rb can be installed via rubygems.org:

data/lib/llm/buffer.rb

@@ -135,7 +135,7 @@ module LLM
       params = [
         *oldparams,
         params.merge(input: messages),
-        @response ? {previous_response_id: @response.id} : {}
+        @response ? {previous_response_id: @response.response_id} : {}
       ].inject({}, &:merge!)
       @response = @provider.responses.create(message.content, params.merge(role:))
       @completed.concat([*pendings, message, *@response.outputs[0]])

data/lib/llm/message.rb

@@ -127,6 +127,25 @@ module LLM
       extra[:response]
     end
 
+    ##
+    # @note
+    #  This method returns token usage for assistant messages,
+    #  and it returns an empty object for non-assistant messages
+    # Returns token usage statistics
+    # @return [LLM::Object]
+    def usage
+      @usage ||= if response
+        LLM::Object.from_hash({
+          input_tokens: response.prompt_tokens || 0,
+          output_tokens: response.completion_tokens || 0,
+          total_tokens: response.total_tokens || 0
+        })
+      else
+        LLM::Object.from_hash({})
+      end
+    end
+    alias_method :token_usage, :usage
+
     ##
     # Returns a string representation of the message
     # @return [String]

data/lib/llm/provider.rb

@@ -268,7 +268,7 @@ class LLM::Provider
   # @raise [SystemCallError]
   #  When there is a network error at the operating system level
   # @return [Net::HTTPResponse]
-  def execute(request:, stream: nil, &b)
+  def execute(request:, stream: nil, stream_parser: self.stream_parser, &b)
     res = if stream
       client.request(request) do |res|
         handler = event_handler.new stream_parser.new(stream)

data/lib/llm/providers/openai/response/responds.rb

@@ -2,6 +2,7 @@
 
 module LLM::OpenAI::Response
   module Responds
+    def response_id = respond_to?(:response) ? response["id"] : id
     def outputs = [format_message]
     def choices = body.output
     def tools = output.select { _1.type == "function_call" }

data/lib/llm/providers/openai/responses/stream_parser.rb

@@ -0,0 +1,76 @@
+# frozen_string_literal: true
+
+class LLM::OpenAI
+  ##
+  # @private
+  class Responses::StreamParser
+    ##
+    # Returns the fully constructed response body
+    # @return [LLM::Object]
+    attr_reader :body
+
+    ##
+    # @param [#<<] io An IO-like object
+    # @return [LLM::OpenAI::Responses::StreamParser]
+    def initialize(io)
+      @body = LLM::Object.new(output: []) # Initialize with an empty output array
+      @io = io
+    end
+
+    ##
+    # @param [Hash] chunk
+    # @return [LLM::OpenAI::Responses::StreamParser]
+    def parse!(chunk)
+      tap { handle_event(chunk) }
+    end
+
+    private
+
+    def handle_event(chunk)
+      case chunk["type"]
+      when "response.created"
+        chunk.each do |k, v|
+          next if k == "type"
+          @body[k] = v
+        end
+        @body.output ||= []
+      when "response.output_item.added"
+        output_index = chunk["output_index"]
+        item = LLM::Object.from_hash(chunk["item"])
+        @body.output[output_index] = item
+        @body.output[output_index].content ||= []
+      when "response.content_part.added"
+        output_index = chunk["output_index"]
+        content_index = chunk["content_index"]
+        part = LLM::Object.from_hash(chunk["part"])
+        @body.output[output_index] ||= LLM::Object.new(content: [])
+        @body.output[output_index].content ||= []
+        @body.output[output_index].content[content_index] = part
+      when "response.output_text.delta"
+        output_index = chunk["output_index"]
+        content_index = chunk["content_index"]
+        delta_text = chunk["delta"]
+        output_item = @body.output[output_index]
+        if output_item&.content
+          content_part = output_item.content[content_index]
+          if content_part && content_part.type == "output_text"
+            content_part.text ||= ""
+            content_part.text << delta_text
+            @io << delta_text if @io.respond_to?(:<<)
+          end
+        end
+      when "response.output_item.done"
+        output_index = chunk["output_index"]
+        item = LLM::Object.from_hash(chunk["item"])
+        @body.output[output_index] = item
+      when "response.content_part.done"
+        output_index = chunk["output_index"]
+        content_index = chunk["content_index"]
+        part = LLM::Object.from_hash(chunk["part"])
+        @body.output[output_index] ||= LLM::Object.new(content: [])
+        @body.output[output_index].content ||= []
+        @body.output[output_index].content[content_index] = part
+      end
+    end
+  end
+end

data/lib/llm/providers/openai/responses.rb

@@ -15,6 +15,7 @@ class LLM::OpenAI
   #   [res1, res2].each { llm.responses.delete(_1) }
   class Responses
     require_relative "response/responds"
+    require_relative "responses/stream_parser"
     include Format
 
     ##
@@ -37,12 +38,13 @@ class LLM::OpenAI
     def create(prompt, params = {})
       params = {role: :user, model: @provider.default_model}.merge!(params)
       params = [params, format_schema(params), format_tools(params)].inject({}, &:merge!).compact
-      role = params.delete(:role)
+      role, stream = params.delete(:role), params.delete(:stream)
+      params[:stream] = true if stream.respond_to?(:<<) || stream == true
       req = Net::HTTP::Post.new("/v1/responses", headers)
       messages = [*(params.delete(:input) || []), LLM::Message.new(role, prompt)]
       body = JSON.dump({input: [format(messages, :response)].flatten}.merge!(params))
       set_body_stream(req, StringIO.new(body))
-      res = execute(request: req)
+      res = execute(request: req, stream:, stream_parser:)
       LLM::Response.new(res).extend(LLM::OpenAI::Response::Responds)
     end
 
@@ -86,5 +88,9 @@ class LLM::OpenAI
       name = "JSONSchema"
       {text: {format: {type: "json_schema", name:, schema:}}}
     end
+
+    def stream_parser
+      LLM::OpenAI::Responses::StreamParser
+    end
   end
 end

data/lib/llm/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module LLM
-  VERSION = "0.14.2"
+  VERSION = "0.15.0"
 end

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: llm.rb
 version: !ruby/object:Gem::Version
-  version: 0.14.2
+  version: 0.15.0
 platform: ruby
 authors:
 - Antar Azri
@@ -241,6 +241,7 @@ files:
 - lib/llm/providers/openai/response/moderations.rb
 - lib/llm/providers/openai/response/responds.rb
 - lib/llm/providers/openai/responses.rb
+- lib/llm/providers/openai/responses/stream_parser.rb
 - lib/llm/providers/openai/stream_parser.rb
 - lib/llm/providers/openai/vector_stores.rb
 - lib/llm/providers/xai.rb