Fix streaming pull.

Author: ioquatix

bake/async/ollama.rb

@@ -17,11 +17,7 @@ def models
 
 # Pulls the specified models from the Ollama API. If no models are specified, but there is a default model, it will pull that one.
 # @parameter models [Array(String)] The names of the models to pull.
-def pull(models)
-	if models.empty?
-		models = [Async::Ollama::Client.default_model]
-	end
-	
+def pull(models: [Async::Ollama::MODEL])
 	Async::Ollama::Client.open do |client|
 		models.each do |model|
 			client.pull(model) do |response|

lib/async/ollama/generate.rb

@@ -9,7 +9,7 @@
 module Async
 	module Ollama
 		# Represents a generated response from the Ollama API.
-		class Generate < Async::REST::Representation[Wrapper]
+		class Generate < Async::REST::Representation[GenerateWrapper]
 			# @returns [String | nil] The generated response, or nil if not present.
 			def response
 				self.value[:response]

lib/async/ollama/wrapper.rb

@@ -139,6 +139,24 @@ def parser_for(response)
 				content_type = response.headers["content-type"]
 				media_type = content_type.split(";").first
 
+				case media_type
+				when APPLICATION_JSON
+					return Async::REST::Wrapper::JSON::Parser
+				when APPLICATION_JSON_STREAM
+					return StreamingParser
+				end
+			end
+		end
+		
+		# Wraps generate-specific HTTP responses for the Ollama API, selecting the appropriate parser.
+		class GenerateWrapper < Wrapper
+			# Selects the appropriate parser for the generate HTTP response.
+			# @parameter response [Protocol::HTTP::Response] The HTTP response object.
+			# @returns [Class] The parser class to use.
+			def parser_for(response)
+				content_type = response.headers["content-type"]
+				media_type = content_type.split(";").first
+				
 				case media_type
 				when APPLICATION_JSON
 					return Async::REST::Wrapper::JSON::Parser