improvements(ai): Improve AI streaming UI/UX interactions + better separation of AI provider responsibilities (#2039)

* Start refactor

* Interface updates

* Rework Assistant, Provider, and tests for better domain boundaries

* Consolidate and simplify OpenAI provider and provider concepts

* Clean up assistant streaming

* Improve assistant message orchestration logic

* Clean up "thinking" UI interactions

* Remove stale class

* Regenerate VCR test responses

app/models/assistant/responder.rb

@@ -0,0 +1,87 @@
+class Assistant::Responder
+  def initialize(message:, instructions:, function_tool_caller:, llm:)
+    @message = message
+    @instructions = instructions
+    @function_tool_caller = function_tool_caller
+    @llm = llm
+  end
+
+  def on(event_name, &block)
+    listeners[event_name.to_sym] << block
+  end
+
+  def respond(previous_response_id: nil)
+    # For the first response
+    streamer = proc do |chunk|
+      case chunk.type
+      when "output_text"
+        emit(:output_text, chunk.data)
+      when "response"
+        response = chunk.data
+
+        if response.function_requests.any?
+          handle_follow_up_response(response)
+        else
+          emit(:response, { id: response.id })
+        end
+      end
+    end
+
+    get_llm_response(streamer: streamer, previous_response_id: previous_response_id)
+  end
+
+  private
+    attr_reader :message, :instructions, :function_tool_caller, :llm
+
+    def handle_follow_up_response(response)
+      streamer = proc do |chunk|
+        case chunk.type
+        when "output_text"
+          emit(:output_text, chunk.data)
+        when "response"
+          # We do not currently support function executions for a follow-up response (avoid recursive LLM calls that could lead to high spend)
+          emit(:response, { id: chunk.data.id })
+        end
+      end
+
+      function_tool_calls = function_tool_caller.fulfill_requests(response.function_requests)
+
+      emit(:response, {
+        id: response.id,
+        function_tool_calls: function_tool_calls
+      })
+
+      # Get follow-up response with tool call results
+      get_llm_response(
+        streamer: streamer,
+        function_results: function_tool_calls.map(&:to_result),
+        previous_response_id: response.id
+      )
+    end
+
+    def get_llm_response(streamer:, function_results: [], previous_response_id: nil)
+      response = llm.chat_response(
+        message.content,
+        model: message.ai_model,
+        instructions: instructions,
+        functions: function_tool_caller.function_definitions,
+        function_results: function_results,
+        streamer: streamer,
+        previous_response_id: previous_response_id
+      )
+
+      unless response.success?
+        raise response.error
+      end
+
+      response.data
+    end
+
+    def emit(event_name, payload = nil)
+      listeners[event_name.to_sym].each { |block| block.call(payload) }
+    end
+
+    def listeners
+      @listeners ||= Hash.new { |h, k| h[k] = [] }
+    end
+end