Removed supervisor crew, added simple LLM supervisor

apps/ai_api/eda_ai_api/api/routes/supervisor.py

@@ -1,21 +1,42 @@
+import os
 from fastapi import APIRouter
+from langchain.prompts import PromptTemplate
+from langchain_groq import ChatGroq
+from langchain.chains import LLMChain
 from eda_ai_api.models.supervisor import SupervisorRequest, SupervisorResponse
-from supervisor.crew import SupervisorCrew
 from opportunity_finder.crew import OpportunityFinderCrew
 from proposal_writer.crew import ProposalWriterCrew
 
 router = APIRouter()
 
+print("API KEY:", os.environ.get("GROQ_API_KEY"))
+# Setup LLM and prompt
+llm = ChatGroq(
+    model_name="llama3-groq-70b-8192-tool-use-preview",
+    api_key=os.environ.get("GROQ_API_KEY"),
+    temperature=0.5,
+)
+
+
+ROUTER_TEMPLATE = """Given a user message, determine the appropriate service to handle the request.
+Choose between:
+- discovery: For finding grant opportunities
+- proposal: For writing grant proposals
+- heartbeat: For checking system health
+
+User message: {message}
+
+Return only one word (discovery/proposal/heartbeat):"""
+
+router_prompt = PromptTemplate(input_variables=["message"], template=ROUTER_TEMPLATE)
+
+router_chain = LLMChain(llm=llm, prompt=router_prompt)
+
 
 @router.post("/supervisor", response_model=SupervisorResponse)
 async def supervisor_route(request: SupervisorRequest) -> SupervisorResponse:
-    # Get crew decision
-    crew_response = (
-        SupervisorCrew().create_crew().kickoff(inputs={"message": request.message})
-    )
-
-    # Extract decision from crew response
-    decision = str(crew_response).lower()
+    # Get routing decision from LLM
+    decision = router_chain.run(message=request.message).strip().lower()
 
     # Print input message and decision for debugging
     print("\n==================================================")
@@ -26,16 +47,13 @@ async def supervisor_route(request: SupervisorRequest) -> SupervisorResponse:
 
     # Handle different decision paths
     if decision == "discovery":
-        # TODO: Extract topics from the actual message
         topics = ["AI", "Technology"]
         result = (
             OpportunityFinderCrew().crew().kickoff(inputs={"topics": ", ".join(topics)})
         )
     elif decision == "proposal":
-        # Initialize ProposalWriterCrew with context from the message
-        # TODO: Extract project and grant details from message
-        community_project = "sample_project"  # This should come from message parsing
-        grant_call = "sample_grant"  # This should come from message parsing
+        community_project = "sample_project"
+        grant_call = "sample_grant"
         result = (
             ProposalWriterCrew(
                 community_project=community_project, grant_call=grant_call