Fix streaming response token count report

Author: ldmosquera

lib/ruby_llm/providers/ollama/chat.rb

@@ -28,9 +28,9 @@ def parse_completion_response(response)
           Message.new(
             role: :assistant,
             content: data.dig('message', 'content'),
+            model_id: data['model'],
             input_tokens: data['prompt_eval_count'].to_i,
             output_tokens: data['eval_count'].to_i,
-            model_id: data['model'],
             tool_calls: parse_tool_calls(data.dig('message', 'tool_calls'))
           )
         end

lib/ruby_llm/providers/ollama/streaming.rb

@@ -31,11 +31,9 @@ def build_chunk(data)
             role: :assistant,
             content: data.dig('message', 'content'),
             model_id: data['model'],
-            tool_calls: parse_tool_calls(data.dig('message', 'tool_calls')),
-
-            # NOTE: unavailable in the response - https://ollama.readthedocs.io/en/api/#streaming-responses
-            input_tokens: nil,
-            output_tokens: nil
+            input_tokens: data['prompt_eval_count'],
+            output_tokens: data['eval_count'],
+            tool_calls: parse_tool_calls(data.dig('message', 'tool_calls'))
           )
         end