feat: command interpreter as a service

frida_interfaces/hri/srv/CommandInterpreter.srv

@@ -1,3 +1,3 @@
-string text_heard
+string text
 ---
-frida_interfaces/CommandList commands
+frida_interfaces/Command[] commands

hri/packages/nlp/scripts/command_interpreter.py

@@ -17,6 +17,7 @@
 from std_msgs.msg import String
 
 from frida_interfaces.msg import Command, CommandList
+from frida_interfaces.srv import CommandInterpreter as CommandInterpreterSrv
 
 
 class CommandShape(BaseModel):
@@ -41,6 +42,10 @@ def __init__(self):
         self.declare_parameter("model", "gpt-4o-2024-08-06")
         self.declare_parameter("speech_command_topic", "/speech/raw_command")
         self.declare_parameter("publish_command_topic", "/task_manager/commands")
+        self.declare_parameter(
+            "COMMAND_INTERPRETER_SERVICE", "/nlp/command_interpreter"
+        )
+
         self.declare_parameter("temperature", 0.5)
 
         base_url = self.get_parameter("base_url").get_parameter_value().string_value
@@ -55,6 +60,11 @@ def __init__(self):
             .get_parameter_value()
             .string_value
         )
+        command_interpreter_service = (
+            self.get_parameter("COMMAND_INTERPRETER_SERVICE")
+            .get_parameter_value()
+            .string_value
+        )
 
         self.temperature = (
             self.get_parameter("temperature").get_parameter_value().double_value
@@ -73,6 +83,9 @@ def __init__(self):
             String, speech_command_topic, self._callback, 10
         )
         self.publisher = self.create_publisher(CommandList, publish_command_topic, 10)
+        self.create_service(
+            CommandInterpreterSrv, command_interpreter_service, self.command_service
+        )
 
         self.get_logger().info("Initialized Command Interpreter")
 
@@ -83,6 +96,15 @@ def _callback(self, data: String) -> None:
 
     def run(self, raw_command: str) -> None:
         """Method for running the interpretation of the commands"""
+        commands = self.get_commands(raw_command)
+        self.publisher.publish(commands)
+
+    def command_service(self, req, res):
+        commands = self.get_commands(req.text)
+        res.commands = commands.commands
+        return res
+
+    def get_commands(self, raw_command: str):
         response = (
             self.client.beta.chat.completions.parse(
                 model=self.model,
@@ -96,7 +118,6 @@ def run(self, raw_command: str) -> None:
             .choices[0]
             .message.content
         )
-
         try:
             response_data = json.loads(response)
             result = CommandListShape(**response_data)
@@ -105,7 +126,6 @@ def run(self, raw_command: str) -> None:
             return
 
         self.get_logger().debug(f"Commands interpreted: {result.commands}")
-
         command_list = CommandList()
         command_list.commands = [
             Command(
@@ -115,7 +135,7 @@ def run(self, raw_command: str) -> None:
             )
             for command in result.commands
         ]
-        self.publisher.publish(command_list)
+        return command_list
 
 
 def main(args=None):

hri/packages/speech/assets/.gitignore

@@ -0,0 +1,3 @@
+
+# Folder to include all the downloaded and asset files
+download/

task_manager/scripts/subtask_managers/hri_tasks.py

@@ -31,13 +31,13 @@ class HRITasks(metaclass=SubtaskMeta):
     def __init__(self, task_manager, config=None) -> None:
         self.node = task_manager
 
-        self.speak_service = self.node.create_client(Speak, SPEAK_SERVICE)
-        self.hear_service = self.node.create_client(STT, HEAR_SERVICE)
-        self.extract_data_service = self.node.create_client(
+        self.speak_client = self.node.create_client(Speak, SPEAK_SERVICE)
+        self.hear_client = self.node.create_client(STT, HEAR_SERVICE)
+        self.extract_data_client = self.node.create_client(
             ExtractInfo, EXTRACT_DATA_SERVICE
         )
 
-        self.command_interpreter_service = self.node.create_client(
+        self.command_interpreter_client = self.node.create_client(
             CommandInterpreter, COMMAND_INTERPRETER_SERVICE
         )
         self.grammar_service = self.node.create_client(Grammar, GRAMMAR_SERVICE)
@@ -47,7 +47,7 @@ def say(self, text: str, wait: bool = False) -> None:
         self.node.get_logger().info(f"Sending to saying service: {text}")
         request = Speak.Request(text=text)
 
-        future = self.speak_service.call_async(request)
+        future = self.speak_client.call_async(request)
 
         if wait:
             rclpy.spin_until_future_complete(self.node, future)
@@ -60,13 +60,15 @@ def say(self, text: str, wait: bool = False) -> None:
 
     def extract_data(self, query, complete_text) -> str:
         request = ExtractInfo.Request(data=query, full_text=complete_text)
-        future = self.extract_data_service.call_async(request)
+        future = self.extract_data_client.call_async(request)
         rclpy.spin_until_future_complete(self.node, future)
         return future.result().result
 
+    # TODO
     def hear(self, timeout: float = 15.0) -> str:
         pass
 
+    # TODO
     def interpret_keyword(self, keyword: str, timeout: float) -> str:
         pass
 
@@ -76,15 +78,20 @@ def refactor_text(self, text: str) -> str:
         rclpy.spin_until_future_complete(self.node, future)
         return future.result().corrected_text
 
+    # TODO
     def find_closest(self, query: str, options: Union[list[str], str]) -> str:
         pass
 
+    # TODO
     def ask(self, question: str) -> str:
-        """Method to publish directly text to the speech node"""
         pass
 
-    def command_interpreter(self, text: str) -> str:
-        pass
+    def command_interpreter(self, text: str) -> CommandInterpreter.Response:
+        request = CommandInterpreter.Request(text=text)
+        future = self.command_interpreter_client.call_async(request)
+        rclpy.spin_until_future_complete(self.node, future)
+
+        return future.result().commands
 
 
 if __name__ == "__main__":

task_manager/scripts/test_manager.py

@@ -23,18 +23,23 @@ def run(self):
         """testing vision tasks"""
 
         user_request = self.subtask_manager["hri"].hear()
-        say_res = self.subtask_manager["hri"].say("Hi, my name is frida", wait=True)
-
-        print("user_request:", user_request)
-        print("say_res:", say_res)
+        self.subtask_manager["hri"].say("Hi, my name is frida", wait=True)
 
         drink = self.subtask_manager["hri"].extract_data("Drink", user_request)
 
         self.get_logger().info(f"Extracted data: {drink}")
 
-        fixed_text = self.subtask_manager["hri"].refactor_text(
-            "I will do: (action: go), (complement: kitchen)"
-        )
+        commands = self.subtask_manager["hri"].command_interpreter(user_request)
+
+        self.get_logger().info(f"Interpreted commands: {commands}")
+
+        command_strs = [
+            f"I will do action:{command.action}, ({command.complement}), ({command.characteristic})"
+            for command in commands
+        ]
+        command_str = " and ".join(command_strs)
+
+        fixed_text = self.subtask_manager["hri"].refactor_text(command_str)
         self.subtask_manager["hri"].say(fixed_text)