Restructure: agentic workflow code

multi-agentic-travel-advisor/places_agent/Dockerfile → agentic-workflows/multi-agentic-travel-advisor/places_agent/Dockerfile

 # Use a slim Python base image
-FROM python:3.10-slim
+FROM python:3.12-slim
 
 ENV PYTHONUNBUFFERED=1
 

multi-agentic-travel-advisor/weather_agent/agent.proto → agentic-workflows/multi-agentic-travel-advisor/places_agent/agent.proto

@@ -18,7 +18,6 @@ enum AgentType {
     WEATHER = 1;
     PLACES = 2;
     FOOD = 3;
-    MAP = 4;
 }
 
 message AgentResponse {

multi-agentic-travel-advisor/places_agent/requirements.txt → agentic-workflows/multi-agentic-travel-advisor/places_agent/requirements.txt

-grpcio==1.38.0
-grpcio-tools==1.38.0
+grpcio
+grpcio-tools
 grpc-interceptor
-protobuf==3.16.0
+protobuf
 multithreading
 openai
 requests

multi-agentic-travel-advisor/planner/Dockerfile → agentic-workflows/multi-agentic-travel-advisor/planner/Dockerfile

 # Use a slim Python base image
-FROM python:3.10-slim
+FROM python:3.12-slim
 
 ENV PYTHONUNBUFFERED=1
 

multi-agentic-travel-advisor/planner/planner.proto → agentic-workflows/multi-agentic-travel-advisor/planner/planner.proto

@@ -18,7 +18,6 @@ enum AgentType {
     WEATHER = 1;
     PLACES = 2;
     FOOD = 3;
-    MAP = 4;
 }
 
 message AgentResponse {

multi-agentic-travel-advisor/planner/requirements.txt → agentic-workflows/multi-agentic-travel-advisor/planner/requirements.txt

-grpcio==1.38.0
-grpcio-tools==1.38.0
+grpcio
+grpcio-tools
 grpc-interceptor
-protobuf==3.16.0
+protobuf
 multithreading
 Flask
 
 torch
-transformers
+transformers
\ No newline at end of file
+hf_xet
\ No newline at end of file

agentic-workflows/multi-agentic-travel-advisor/planner/server.py

+import grpc
+from concurrent import futures
+import logging
+import time
+
+import planner_pb2
+import planner_pb2_grpc
+from transformers import pipeline
+
+
+# ------------------ Logger Setup ------------------
+def setup_logger():
+    logger = logging.getLogger("PlannerService")
+    logging.basicConfig(level=logging.INFO, format="%(asctime)s - %(levelname)s - %(message)s")
+    return logger
+
+
+logger = setup_logger()
+
+# ------------------ Classifier Setup ------------------
+def initialize_classifier():
+    return pipeline(
+        "zero-shot-classification",
+        model="facebook/bart-large-mnli",
+        revision="d7645e1"
+    )
+
+
+AGENT_LABELS = ["weather", "places", "food", "map"]
+LABEL_TO_AGENT_TYPE = {
+    "weather": planner_pb2.WEATHER,
+    "places": planner_pb2.PLACES,
+    "food": planner_pb2.FOOD
+}
+
+
+def match_agent(text):
+    try:
+        result = classifier(text, AGENT_LABELS)
+        label = result["labels"][0]
+        logger.info(f"Classifier result: {label}")
+        return LABEL_TO_AGENT_TYPE.get(label, planner_pb2.UNKNOWN)
+    except Exception as e:
+        logger.error(f"Classifier error: {e}")
+        return planner_pb2.UNKNOWN
+
+# ------------------ gRPC Service Implementation ------------------
+class PlannerService(planner_pb2_grpc.PlannerServicer):
+    def __init__(self):
+        logger.info("PlannerService initialized.")
+        self.agent_type_stored = None
+
+    def evaluateUserQueriesFromChatbot(self, request_iterator, context):
+        logger.info("Evaluating user queries from chatbot...")
+        for user_query in request_iterator:
+            logger.info(f"User query: {user_query.text}")
+            agent_type = match_agent(user_query.text)
+
+            response = planner_pb2.AgentRequest(
+                text=user_query.text,
+                agent_type=agent_type
+            )
+
+            # Store agent type as a readable string
+            self.agent_type_stored = planner_pb2.AgentType.Name(agent_type)
+            logger.info(f"Matched agent type: {self.agent_type_stored}")
+
+            # Handle unknown queries
+            if agent_type == planner_pb2.UNKNOWN:
+                response.text = "Sorry, I couldn't understand the query."
+
+            yield response
+
+    def processAgentResponsesFromAgent(self, request_iterator, context):
+        logger.info("Processing responses from agent...")
+        for response in request_iterator:
+            logger.info(f"Agent response: {response.text}")
+
+            formatted_response = planner_pb2.AgentResponse(
+                text=f"[**Answering from {self.agent_type_stored} Agent**] \n{response.text}"
+            )
+            yield formatted_response
+
+        logger.info("Finished processing all agent responses.")
+
+
+# ------------------ Server Setup ------------------
+def serve():
+    server = grpc.server(futures.ThreadPoolExecutor(max_workers=10))
+    planner_pb2_grpc.add_PlannerServicer_to_server(PlannerService(), server)
+    port = 8061
+    server.add_insecure_port(f"[::]:{port}")
+    server.start()
+    logger.info(f"Planner Service is running on port {port}")
+    server.wait_for_termination()
+
+
+if __name__ == "__main__":
+    classifier = initialize_classifier()
+    serve()

multi-agentic-travel-advisor/weather_agent/Dockerfile → agentic-workflows/multi-agentic-travel-advisor/weather_agent/Dockerfile

 # Use a slim Python base image
-FROM python:3.10-slim
+FROM python:3.12-slim
 
 ENV PYTHONUNBUFFERED=1