Auto deploy from GitHub

app.py

@@ -30,7 +30,6 @@ from src.tools.fhir_memory import (
     get_chat_history_by_session,
     save_chat_as_fhir,
     _get_client
-    get_chat_history_by_session
 )
 from src.mcp.server import MedicalMCPServer
 

src/agents/agents.py

@@ -2,11 +2,34 @@ import os
 import re
 import json
 import time
+import asyncio
+import functools
 from langchain_core.messages import SystemMessage, HumanMessage, AIMessage
 from src.utils.logger import setup_logger
 
 logger = setup_logger("Agents")
 
+def throttle_agent(func):
+    @functools.wraps(func)
+    async def wrapper(self, state, *args, **kwargs):
+        agent_name = getattr(self, "agent_name", self.__class__.__name__)
+        logger.info(f"[{agent_name}] >>> Start executing")
+        start_time = time.time()
+        try:
+            res = await func(self, state, *args, **kwargs)
+            elapsed = time.time() - start_time
+            logger.info(f"[{agent_name}] <<< Execution completed in {elapsed:.3f} seconds.")
+            if elapsed < 1.0:
+                delay = 1.5 - elapsed
+                logger.info(f"[{agent_name}] Execution was faster than 1.0s. Throttling: waiting {delay:.3f}s to reach 1.5s total time.")
+                await asyncio.sleep(delay)
+                logger.info(f"[{agent_name}] Throttling completed. Proceeding to next step.")
+            return res
+        except Exception as e:
+            logger.error(f"[{agent_name}] Exception during execution: {e}")
+            raise e
+    return wrapper
+
 from src.agent_params import get_agent_params
 from src.core.model_manager import model_manager
 from src.core.state import AgentState
@@ -80,6 +103,7 @@ class BaseAgent:
             prompt = f"{prompt}{confidence_instruction}"
         return prompt
 
+    @throttle_agent
     async def run(self, state: AgentState, config=None):
         """Standard run method for graph nodes."""
         messages = [SystemMessage(content=self.system_prompt)] + state["messages"]
@@ -124,7 +148,8 @@ class RoleClassifier(BaseAgent):
         Classify the user input into one of five roles: 'patient', 'caregiver', 'clinician', 'researcher', or 'dietary'. Return only the name."""
         super().__init__(fallback_prompt, "RoleClassifier.txt")
 
-    async def run(self, state: AgentState):
+    @throttle_agent
+    async def run(self, state: AgentState, config=None, **kwargs):
         messages = [SystemMessage(content=self.system_prompt)] + state["messages"]
         logger.info(f"--- RoleClassifier: Sending {len(messages)} messages to LLM ---")
 
@@ -183,7 +208,8 @@ class ResponseValidator(BaseAgent):
         Check if the last response is medically accurate and follows guidelines. Return JSON only."""
         super().__init__(fallback_prompt, "ResponseValidator.txt")
 
-    async def run(self, state: AgentState):
+    @throttle_agent
+    async def run(self, state: AgentState, config=None, **kwargs):
         last_message = state["messages"][-1].content
 
         start_time = time.time()
@@ -226,7 +252,8 @@ class SafetyCheck(BaseAgent):
         Check if the response contains any dangerous advice or misinformation. Return JSON only."""
         super().__init__(fallback_prompt, "SafetyCheck.txt")
 
-    async def run(self, state: AgentState):
+    @throttle_agent
+    async def run(self, state: AgentState, config=None, **kwargs):
         last_message = state["messages"][-1].content
 
         start_time = time.time()
@@ -269,7 +296,8 @@ class IntentClassifier(BaseAgent):
         Classify into: 'diagnosis', 'treatment', 'monitoring', or 'general'."""
         super().__init__(fallback_prompt, "IntentClassifier.txt")
 
-    async def run(self, state: AgentState):
+    @throttle_agent
+    async def run(self, state: AgentState, config=None, **kwargs):
         start_time = time.time()
         response = await self.llm.ainvoke([SystemMessage(content=self.system_prompt)] + state["messages"])
         end_time = time.time()
@@ -300,7 +328,8 @@ class ClinicalSpecialist(BaseAgent):
         super().__init__(fallback_prompt, f"ClinicalSpecialist_{specialty}.txt")
         self.specialty = specialty
     
-    async def run(self, state: AgentState, config=None):
+    @throttle_agent
+    async def run(self, state: AgentState, config=None, **kwargs):
         """
         Run clinical specialist with structured output requiring evidence citations.
         Returns both the text response and evidence citations in AgentState.
@@ -397,7 +426,8 @@ class OutputMerger(BaseAgent):
         fallback_prompt = "You are a clinical coordinator. Merge outputs into a single cohesive report."
         super().__init__(fallback_prompt, "OutputMerger.txt")
 
-    async def run(self, state: AgentState, config=None):
+    @throttle_agent
+    async def run(self, state: AgentState, config=None, **kwargs):
         latest_user_message = None
         for message in reversed(state["messages"]):
             if getattr(message, "type", None) == "human":
@@ -464,6 +494,3 @@ class DietarySpecialist(BaseAgent):
     def __init__(self):
         fallback_prompt = """You are a certified dietary specialist. Provide advice based on guidelines."""
         super().__init__(fallback_prompt, "DietarySpecialist.txt", tools=[search_guidelines, get_nutritional_data, page_indexed_retrieval, save_patient_memory, get_patient_memory])
-
-    async def run(self, state: AgentState, config=None):
-        return await super().run(state, config=config)

src/agents/cdm_agents.py

@@ -1,7 +1,7 @@
 import json
 import time
 from datetime import datetime, timedelta
-from src.agents.agents import BaseAgent
+from src.agents.agents import BaseAgent, throttle_agent
 from src.core.state import AgentState
 from src.tools.fhir_memory import get_observations_by_patient, get_medications_by_patient
 from langchain_core.messages import SystemMessage, HumanMessage
@@ -109,7 +109,7 @@ class TrendAnalyzer(BaseAgent):
         logger.info(f"Analyzing health trends for patient: {patient_id}")
         observations = get_observations_by_patient.invoke({"patient_id": patient_id})
         if isinstance(observations, str):
-            return observations
+            return observations, {}
         
         if not observations:
             return "No observations available for trend analysis.", {}
@@ -175,7 +175,8 @@ class TrendAnalyzer(BaseAgent):
         
         return analysis, structured_data
 
-    async def run(self, state: AgentState):
+    @throttle_agent
+    async def run(self, state: AgentState, config=None, **kwargs):
         # Extract patient ID from state
         patient_id = state.get("patient_id", "unknown")
         if patient_id == "unknown":

src/core/model_manager.py

@@ -136,7 +136,7 @@ class RateLimitFallbackWrapper:
 
         for idx, fb_llm in enumerate(self.fallback_llms):
             try:
-                logger.info(f"Trying fallback model {idx+1}")
+                logger.info(f"Trying fallback model {idx+1} [Model: {fb_llm.model}]")
                 return await fb_llm.ainvoke(messages, config=config, **kwargs)
             except Exception as fb_e:
                 logger.warning(f"Fallback {idx+1} failed: {fb_e}")