feat: implemented basic belief-from-text extractor

The communication with other agents has been tested with mock data as the other agents (transcriber and belief collector) are not yet implemented.

ref: N25B-208

src/control_backend/agents/bdi/behaviours/text_belief_extractor.py

@@ -0,0 +1,76 @@
+import asyncio
+from spade.behaviour import CyclicBehaviour
+import logging
+from spade.message import Message
+import json
+from control_backend.core.config import settings
+
+
+class BeliefFromText(CyclicBehaviour):
+    logger = logging.getLogger("Belief From Text")
+
+    # TODO: LLM prompt nog hardcoded
+    llm_instruction_prompt = """
+    You are an information extraction assistent for a BDI agent. Your task is to extract values from 
+    a user's text to bind a list of ungrounded beliefs. Rules:
+    You will receive a JSON object with "beliefs" (a list of ungrounded AgentSpeak beliefs) and "text" (user's transcript).
+    Analyze the text to find values that sematically match the variables (X,Y,Z) in the beliefs.
+    A single piece of text might contain multiple instances that match a belief.
+    Respond ONLY with a single JSON object.
+    The JSON object's keys should be the belief functors (e.g., "weather").
+    The value for each key must be a list of lists.
+    Each inner list must contain the extracted arguments (as strings) for one instance of that belief.
+    CRITICAL: If no information in the text matches a belief, DO NOT include that key in your response.
+    """
+    
+    # on_start agent receives message containing the beliefs to look out for and sets up the LLM with instruction prompt
+    #async def on_start(self):
+    #    msg = await self.receive(timeout=0.1)
+    #    self.beliefs = dict uit message
+    #    send instruction prompt to LLM
+
+    beliefs: dict[str,list[str]]
+    beliefs = {
+        "mood": ["X"],
+        "car": ["Y"]
+    }
+
+    async def run(self):
+        msg = await self.receive(timeout=0.1)
+        if msg:
+            sender = msg.sender.node
+            match sender:
+                # TODO: Change to Transcriber agent name once implemented
+                case settings.agent_settings.test_agent_name:
+                    self.logger.info("Received text from transcriber.")
+                    await self._process_transcription(msg.body)
+                case _:
+                    self.logger.info("Received message from other agent.")
+                    pass
+        await asyncio.sleep(1)
+    
+    async def _process_transcription(self,text: str):
+        text_prompt = f"Text: {text}"
+        
+        beliefs_prompt = "These are the beliefs to be bound:\n"
+        for belief, values in self.beliefs.items():
+            beliefs_prompt += f"{belief}({', '.join(values)})\n"
+        
+        prompt = text_prompt + beliefs_prompt
+        self.logger.info(prompt)
+        #prompt_msg = Message(to="LLMAgent@whatever")
+        #response = self.send(prompt_msg)
+
+        # Mock response; response is beliefs in JSON format, it parses do dict[str,list[list[str]]]
+        response = '{"mood": [["happy"]]}'
+        # Verify by trying to parse
+        try:
+            json.loads(response)
+            belief_message = Message(to=settings.agent_settings.bdi_core_agent_name + '@' + settings.agent_settings.host, body=response)
+            belief_message.thread = "beliefs"
+
+            await self.send(belief_message)
+            self.logger.info("Sent beliefs to BDI.")
+        except:
+            #loading failed so the response is in wrong format, throw warning (let LLM respond to ask again?)
+            self.logger.warning("Received LLM response in incorrect format.")

src/control_backend/agents/bdi/test_agent.py

@@ -0,0 +1,26 @@
+import spade
+from spade.agent import Agent
+from spade.behaviour import OneShotBehaviour
+from spade.message import Message
+from spade.template import Template
+from control_backend.core.config import AgentSettings, settings
+
+class SenderAgent(Agent):
+    class InformBehav(OneShotBehaviour):
+        async def run(self):
+            msg = Message(to=settings.agent_settings.text_belief_extractor_agent_name + '@' + settings.agent_settings.host)     # Instantiate the message
+            msg.body = "This is a test input to extract beliefs from.\n"                    # Set the message content
+
+            await self.send(msg)
+            print("Message sent!")
+
+            # set exit_code for the behaviour
+            self.exit_code = "Job Finished!"
+
+            # stop agent from behaviour
+            await self.agent.stop()
+
+    async def setup(self):
+        print("SenderAgent started")
+        self.b = self.InformBehav()
+        self.add_behaviour(self.b)

src/control_backend/agents/bdi/text_extractor.py

@@ -0,0 +1,10 @@
+import spade
+from spade.agent import Agent
+import logging
+
+from control_backend.agents.bdi.behaviours.text_belief_extractor import BeliefFromText
+
+class TBeliefExtractor(Agent):
+    async def setup(self):
+        self.b = BeliefFromText()
+        self.add_behaviour(self.b)

src/control_backend/main.py

@@ -13,6 +13,8 @@ import zmq
 
 # Internal imports
 from control_backend.agents.bdi.bdi_core import BDICoreAgent
+from control_backend.agents.bdi.text_extractor import TBeliefExtractor
+from control_backend.agents.bdi.test_agent import SenderAgent
 from control_backend.api.v1.router import api_router
 from control_backend.core.config import AgentSettings, settings
 from control_backend.core.zmq_context import context
@@ -32,8 +34,12 @@ async def lifespan(app: FastAPI):
     logger.info("Internal publishing socket bound to %s", internal_comm_socket)
 
     # Initiate agents
-    bdi_core = BDICoreAgent(settings.agent_settings.bdi_core_agent_name + '@' + settings.agent_settings.host, settings.agent_settings.bdi_core_agent_name, "src/control_backend/agents/bdi/rules.asl")
+    bdi_core = BDICoreAgent(settings.agent_settings.bdi_core_agent_name + '@' + settings.agent_settings.host, "pohpu7-huqsyH-qutduk", "src/control_backend/agents/bdi/rules.asl")
     await bdi_core.start()
+    text_belief_extractor = TBeliefExtractor(settings.agent_settings.text_belief_extractor_agent_name + '@' + settings.agent_settings.host, "pohpu7-huqsyH-qutduk")
+    await text_belief_extractor.start()
+    test_agent = SenderAgent(settings.agent_settings.test_agent_name + '@' + settings.agent_settings.host, "pohpu7-huqsyH-qutduk")
+    await test_agent.start()
     
     yield