upload agentlite example

examples/agentlite/example/ExecAction.py

@@ -0,0 +1,45 @@
+import os
+
+import wikipedia
+import duckduckgo_search
+
+from agentlite.actions.BaseAction import BaseAction
+import requests
+import json
+
+class ExecAction(BaseAction):
+    def __init__(self) -> None:
+        #action_name = "request"
+        #action_desc = "Using this action to request servers."
+        #params_doc = {"url": "the request url.","query":"the request content, json format."}
+        super().__init__(
+            action_name="request", action_desc="Using this action to request servers.", params_doc={"url": "the request url.","query":"if news or weather or qa or stock or currency server, then empty query. json format, like this{\"intent\": \"calendar_remove\", \"iid\": \"7890\", \"event_name\": \"haircut appointment\", \"date\": \"2024-11-20\"}"},
+        )
+        print(self.action_name)
+
+    def __call__(self, url,query):
+      if url.find("mediastack")!=-1 or url.find("meteo")!=-1 or url.find("marketstack")!=-1:
+        response=requests.get(url)
+        print(response.text)
+      else:
+        response = requests.post(url, data=query)
+
+        # 打印返回结果
+        print(response.text)
+        result=""
+        if response.status_code >= 400:
+            print(query)#["iid"])
+            result="code "+str(response.status_code)+"."
+        else:
+            pass
+            '''
+            res=json.loads(response.text)
+            if res["data"]["response"]=="true":
+                count+=1
+                result="true"
+            else:
+                print(query)#["iid"])
+                result="false"
+            print(res["data"]["response"])
+            '''
+      return response.text

examples/agentlite/example/ExecAgent.py

@@ -0,0 +1,128 @@
+import os 
+
+from typing import List
+from ExecAction import ExecAction
+from agentlite.actions import BaseAction, FinishAct, ThinkAct
+from agentlite.agents import BaseAgent
+from agentlite.commons import TaskPackage
+from agentlite.commons import AgentAct
+from agentlite.llm.agent_llms import get_llm_backend, LLMConfig
+from agentlite.llm.agent_llms import BaseLLM
+from agentlite.logging.terminal_logger import AgentLogger
+agent_logger = AgentLogger(PROMPT_DEBUG_FLAG=False)
+# LAM_URL = os.environ["LAM_URL"]
+# print(LAM_URL)
+# llm_config = LLMConfig(
+#         {
+#             "llm_name": "xlam_v2", 
+#             "temperature": 0.0, 
+#             "base_url": LAM_URL,
+#             "api_key": "EMPTY"
+#         }
+#     )
+llm_name = "gpt-4"#3.5-turbo"#-16k"#-0613"#gpt-4"
+llm_config = LLMConfig({"llm_name": llm_name, "temperature": 0.0,
+        "api_key": "",})
+llm = get_llm_backend(llm_config)
+class ExecAgent(BaseAgent):
+    def __init__(
+        self, llm: BaseLLM, actions: List[BaseAction] = [ExecAction()], **kwargs
+    ):
+        name = "ExecAgent"
+        role = "read the params, and send and receive the requests. iid should also be string. choose the url from the servers' url list:\
+                qa server is http://api.serpstack.com/search?access_key={key}&query={query}\n\
+                news query server is http://api.mediastack.com/v1/news?access_key={key}&keywords={keyword}&date={date}&sort=published_desc\n\
+                news subscription server http://127.0.0.1:3020/news,intent(news_subscription),iid,news_topic,\
+                weather server first request https://geocoding-api.open-meteo.com/v1/search?name={place_name}&count=10&language=en&format=json to get latitude and latitude, then request https://api.open-meteo.com/v1/forecast?latitude={latitude}&longitude={longitude}&daily=temperature_2m_max,temperature_2m_min,sunrise,sunset,uv_index_max,rain_sum,showers_sum,snowfall_sum,wind_speed_10m_max\n\
+                stock server is first to get the stock symbol http://api.serpstack.com/search? access_key = {key}& query = {name} stock symbol , then request to this server http://api.marketstack.com/v1/eod? access_key = {key}& symbols = {symbol}&limit=5\n\
+                currency server is https://www.amdoren.com/api/currency.php?api_key={key}&from={currency}&to={currency2}&amount={amount}\n\
+                http://127.0.0.1:3000/alarm, intent(alarm_query,alarm_set),iid,event_name,descriptor,time,from_time,to_time,\
+                http://127.0.0.1:3001/audiobook,intent(play_audiobook), iid,player_setting,house_place,media_type,descriptor,audiobook_name,author_name,\
+                http://127.0.0.1:3002/calendar,intent(calendar_query,calendar_remove,calendar_set),iid,event_name,descriptor,person,relation,date,time,from_time,to_time,\
+                http://127.0.0.1:3003/cooking,intent(cooking_recipe),iid,food_type,descriptor,\
+                http://127.0.0.1:3004/datetime,intent(datetime_convert,datetime_query),iid,place_name,descriptor,time_zone,time_zone2,date,time,time2,\
+                http://127.0.0.1:3005/email,intent(email_query,email_sendemail),iid,setting,person,to_person,from_person,relation,to_relation,from_relation,email_folder,time,date,email_address,app_name,query,content,personal_info,\
+                http://127.0.0.1:3006/game,intent(play_game),iid,game_name,\
+                http://127.0.0.1:3007/iot,intent(iot_coffee,iot_hue_lightcolor,iot_hue_lightother,iot_hue_lightdim,iot_hue_lightup,audio_volume_mute,iot_hue_lightoff,audio_volume_up,iot_wemo_off,audio_volume_other,iot_cleaning,iot_wemo_on,audio_volume_down),iid,device_type,house_place,time,color_type,change_amount,change_to,item_name,setting,\
+                http://127.0.0.1:3008/lists,intent(lists_query,lists_remove,lists_createoradd),iid,list_name,item_name,descriptor,time,date,\
+                http://127.0.0.1:3009/music,intent(play_music,music_likeness,playlists_createoradd,music_settings,music_dislikeness,music_query),iid,player_setting,descriptor,artist_name,song_name,playlist_name,music_genre,query,\
+                http://127.0.0.1:3010/phone,intent(phone_text,phone_notification),iid,device_type,event_name,text,\
+                http://127.0.0.1:3011/podcasts,intent(play_podcasts),iid,podcast_name,player_setting,podcast_descriptor,\
+                http://127.0.0.1:3013/radio,intent(play_radio,radio_query),iid,radio_name,app_name,person_name,music_genre,device_type,house_place,player_setting,descriptor,query,time,\
+                http://127.0.0.1:3014/recommendation,intent(recommendation_events,recommendation_movies,recommendation_locations),iid,business_type,food_type,movie_type,movie_name,date,place_name,event_name,descriptor,\
+                http://127.0.0.1:3015/social,intent(social_query,social_post),iid,media_type,person,business_name,content,date,descriptor,\
+                http://127.0.0.1:3017/takeaway,intent(takeaway_query,takeaway_order),iid,food_type,order_type,business_type,business_name,place_name,date,time,descriptor,\
+                http://127.0.0.1:3018/transport,intent(transport_taxi,transport_ticket,transport_query,transport_traffic),iid,transport_agency,transport_type,business_type,business_name,place_name,to_place_name,from_place_name,query,date,time,descriptor,\
+"
+        super().__init__(
+            name=name, role=role, llm=llm, actions=actions, logger=agent_logger
+        )
+        self.__build_examples__()
+
+    def __build_examples__(self):
+        """
+        constructing the examples for agent working.
+        Each example is a successful action-obs chain of an agent.
+        those examples should cover all those api calls
+        """
+        # an example of search agent with wikipedia api call
+        # task
+        task = "intent:set_notification"
+
+
+                # 2. api call action and obs
+        act_params = {}
+        act_1 = AgentAct(name=ExecAction().action_name, params=act_params)
+        obs_1 = "OK"
+        '''
+        # 3. think action and obs
+        thought = "I find salesforce is Founded by former Oracle executive Marc Benioff in February 1999"
+        act_3 = AgentAct(name=ThinkAct.action_name, params={INNER_ACT_KEY: thought})
+        obs_3 = "OK"
+
+        # 4. finish action
+        answer = "February 1999"
+        act_4 = AgentAct(name=FinishAct.action_name, params={INNER_ACT_KEY: answer})
+        obs_4 = "Task Completed."
+
+        task_pack = TaskPackage(instruction=task)
+        act_obs = [(act_1, obs_1), (act_2, obs_2), (act_3, obs_3), (act_4, obs_4)]
+        self.add_example(task=task_pack, action_chain=act_obs)
+        '''
+def test_search_agent():
+    llm_config_dict = {"llm_name": "gpt-4", "temperature": 0.1}
+    actions = [ExecAction()]
+    llm_config = LLMConfig(llm_config_dict)
+    # print(llm_config.__dict__)
+    llm = get_llm_backend(llm_config)
+    ## test the one-shot wikipedia search agent
+    labor_agent = ExecAgent(llm=llm)
+    # labor_agent = DuckSearchAgent(llm=llm)
+
+    test_task = "intent:set_notification, news_category:world news"
+    test_task_pack = TaskPackage(instruction=test_task)
+    response = labor_agent(test_task_pack)
+    print("response:", response)
+
+
+if __name__ == "__main__":
+    test_search_agent()
+
+'''
+agent_actions = []#get_intent(),get_user_current_date(), get_user_current_location(), get_latitude_longitude()]#, get_weather_forcast()]
+agent = BaseAgent(
+    name=agent_info["name"],
+    role=agent_info["role"],
+    llm=llm,
+    actions=agent_actions,
+    #reasoning_type="react",
+)
+prompt="the examples and results: turn up	intent:audio_volume_up\nwhat's the current weather	intent:weather_query\ni want you to remind me the next meeting with my girlfriend it will be at eight pm next sunday	intent:calendar_set,event_name : meeting,relation:girlfriend, time : eight pm, date : next Sunday\ncancel alarm for tenth of march two thousand seventeen	intent:alarm_remove,date : tenth of march two thousand seventeen\nhave you heard any good jokes lately	intent:general_joke\nget it fast resolved	intent:social_post\nalexa book me a train ticket for this afternoon to chicago	intent:transport_ticket,transport_type : train, timeofday : this afternoon, place_name : chicago\ni did not want you to send that text yet wait until i say send	intent:email_sendemail,setting:save\nwhat causes if i had junk food and alcohols	intent:general_quirky,content:what causes if i had junk food and alcohols\nfind a recipe for a romantic dinner for two	intent:cooking_recipe,food_descriptor: romantic dinner for two\nplease turn up the lights in this room	intent:iot_hue_lightup,house_place:this room\nwhats happening in pop industry	intent:news_query	,news_topic : pop industry\nis today the fourth or the fifth	intent:datetime_query,date : today\nthe wemo plug should be turned off on	intent:iot_wemo_off,device_type : wemo plug\nwhen i get home can you please order a pizza	intent:takeaway_order,food_type : pizza\nfind the events	intent:recommendation_events\nstart radio and go to frequency on one thousand and forty eight	intent:play_radio,radio_name : one thousand and forty eight\ni want to listen arijit singh song once again	intent:play_music,artist_name : arijit singh\nplease arrange to wake me up at three am alarm	intent:alarm_set,time : three am\nstart dune from where i left off	intent:play_audiobook,player_setting : resume, audiobook_name : dune\n please return the result of this sentence: "
+FLAG_CONTINUE = True
+while FLAG_CONTINUE:
+    input_text = input("Ask Intent Agent question:\n")
+    task = TaskPackage(instruction=prompt+"\n"+input_text)
+    agent(task)
+    if input("Do you want to continue? (y/n): ") == "n":
+        FLAG_CONTINUE = False
+'''

examples/agentlite/example/IntentAgent.py

@@ -0,0 +1,91 @@
+import os 
+
+from typing import List
+
+from agentlite.actions import BaseAction, FinishAct, ThinkAct
+from agentlite.agents import BaseAgent
+from agentlite.commons import TaskPackage
+from agentlite.llm.agent_llms import get_llm_backend, LLMConfig
+from agentlite.llm.agent_llms import BaseLLM
+from agentlite.logging.terminal_logger import AgentLogger
+agent_logger = AgentLogger(PROMPT_DEBUG_FLAG=False)
+# LAM_URL = os.environ["LAM_URL"]
+# print(LAM_URL)
+# llm_config = LLMConfig(
+#         {
+#             "llm_name": "xlam_v2", 
+#             "temperature": 0.0, 
+#             "base_url": LAM_URL,
+#             "api_key": "EMPTY"
+#         }
+#     )
+llm_name = "gpt-4"#3.5-turbo"#-16k"#-0613"#gpt-4"
+llm_config = LLMConfig({"llm_name": llm_name, "temperature": 0.0,
+        "api_key": "",})
+llm = get_llm_backend(llm_config)
+class IntentAgent(BaseAgent):
+    def __init__(
+        self, llm: BaseLLM, actions: List[BaseAction] = [], **kwargs
+    ):
+        name = "IntentAgent"
+        role = "read the examples and results, copy iid and predict intent for the sentence. for 'iid:7199,query:set the alarm to two pm' first predict the domain, as domain:alarm, then copy the iid from query,iid:7199, then the intent and slots, as the format: intent:alarm_set,time:two pm. the intents are calendar:calendar_set,calendar_remove,calendar_query\n\
+lists:lists_query,lists_remove,lists_createoradd\n\
+music:play_music,music_likeness,playlists_createoradd,music_settings,music_dislikeness,music_query\n\
+news:news_query,news_subscription\n\
+alarm:alarm_set,alarm_query,alarm_remove,alarm_change\n\
+email:email_sendemail,email_query,email_querycontact,email_subscription,email_addcontact,email_remove\n\
+iot:iot_hue_lightcolor,iot_hue_lightother,iot_coffee,iot_hue_lightdim,iot_hue_lightup,audio_volume_mute,iot_hue_lightoff,audio_volume_up,iot_wemo_off,audio_volume_other,iot_cleaning,iot_wemo_on,audio_volume_down\n\
+weather:weather_query\n\
+datetime:datetime_query,datetime_convert\n\
+stock:qa_stock\n\
+qa:qa_factoid,general_quirky,qa_definition,general_joke,qa_maths\n\
+greet:general_greet\n\
+currency:qa_currency\n\
+transport:transport_taxi,transport_ticket,transport_query,transport_traffic\n\
+recommendation:recommendation_events,recommendation_movies,recommendation_locations\n\
+podcast:play_podcasts\n\
+audiobook:play_audiobook\n\
+radio:play_radio,radio_query\n\
+takeaway:takeaway_query,takeaway_order\n\
+social:social_query,social_post\n\
+cooking:cooking_recipe\n\
+phone:phone_text,phone_notification\n\
+game:play_game\
+"
+        super().__init__(
+            name=name, role=role, llm=llm, actions=actions, logger=agent_logger
+        )
+        #self.__build_examples__()
+
+    def __build_examples__(self):
+        """
+        constructing the examples for agent working.
+        Each example is a successful action-obs chain of an agent.
+        those examples should cover all those api calls
+        """
+        # an example of search agent with wikipedia api call
+        # task
+        task = "send the email to this new email address"
+
+        # 1. think action and obs
+        thought = "I should first figure out the intent"
+        act_1 = AgentAct(name=ThinkAct.action_name, params={INNER_ACT_KEY: thought})
+        obs_1 = ""
+'''
+agent_actions = []
+agent = BaseAgent(
+    name=agent_info["name"],
+    role=agent_info["role"],
+    llm=llm,
+    actions=agent_actions,
+    #reasoning_type="react",
+)
+prompt="the examples and results:  {"iid": "7499", "query": "turn up"}	iid:7499,intent:audio_volume_up\n {"iid": "6811", "query": "what's the current weather"}	iid:6811,intent:weather_query\ni want you to remind me the next meeting with my girlfriend it will be at eight pm next sunday	intent:calendar_set,event_name : meeting,relation:girlfriend, time : eight pm, date : next Sunday\ncancel alarm for tenth of march two thousand seventeen	intent:alarm_remove,date : tenth of march two thousand seventeen\nhave you heard any good jokes lately	intent:general_joke\nget it fast resolved	intent:social_post\nalexa book me a train ticket for this afternoon to chicago	intent:transport_ticket,transport_type : train, timeofday : this afternoon, place_name : chicago\ni did not want you to send that text yet wait until i say send	intent:email_sendemail,setting:save\nwhat causes if i had junk food and alcohols	intent:general_quirky,content:what causes if i had junk food and alcohols\nfind a recipe for a romantic dinner for two	intent:cooking_recipe,food_descriptor: romantic dinner for two\nplease turn up the lights in this room	intent:iot_hue_lightup,house_place:this room\nwhats happening in pop industry	intent:news_query	,news_topic : pop industry\nis today the fourth or the fifth	intent:datetime_query,date : today\nthe wemo plug should be turned off on	intent:iot_wemo_off,device_type : wemo plug\nwhen i get home can you please order a pizza	intent:takeaway_order,food_type : pizza\nfind the events	intent:recommendation_events\nstart radio and go to frequency on one thousand and forty eight	intent:play_radio,radio_name : one thousand and forty eight\ni want to listen arijit singh song once again	intent:play_music,artist_name : arijit singh\nplease arrange to wake me up at three am alarm	intent:alarm_set,time : three am\nstart dune from where i left off	intent:play_audiobook,player_setting : resume, audiobook_name : dune\n please return the result of this sentence: "
+FLAG_CONTINUE = True
+while FLAG_CONTINUE:
+    input_text = input("Ask Intent Agent question:\n")
+    task = TaskPackage(instruction=prompt+"\n"+input_text)
+    agent(task)
+    if input("Do you want to continue? (y/n): ") == "n":
+        FLAG_CONTINUE = False
+'''

examples/agentlite/example/LocationAgent.py

@@ -0,0 +1,68 @@
+import os 
+
+from typing import List
+from intent_actions import get_intent,get_user_current_date, get_user_current_location, get_latitude_longitude
+from agentlite.actions import BaseAction, FinishAct, ThinkAct
+from agentlite.agents import BaseAgent
+from agentlite.commons import TaskPackage
+from agentlite.llm.agent_llms import get_llm_backend, LLMConfig
+from agentlite.llm.agent_llms import BaseLLM
+from agentlite.logging.terminal_logger import AgentLogger
+agent_logger = AgentLogger(PROMPT_DEBUG_FLAG=False)
+# LAM_URL = os.environ["LAM_URL"]
+# print(LAM_URL)
+# llm_config = LLMConfig(
+#         {
+#             "llm_name": "xlam_v2", 
+#             "temperature": 0.0, 
+#             "base_url": LAM_URL,
+#             "api_key": "EMPTY"
+#         }
+#     )
+llm_name = "gpt-4"#3.5-turbo"#-16k"#-0613"#gpt-4"
+llm_config = LLMConfig({"llm_name": llm_name, "temperature": 0.0,
+        "api_key": "",})
+llm = get_llm_backend(llm_config)
+class LocationAgent(BaseAgent):
+    def __init__(
+        self, llm: BaseLLM, actions: List[BaseAction] = [], **kwargs
+    ):
+        name = "LocationAgent"
+        role = "read the location params, and convert to formated location. if location is in house, do not need to resolve. current location is new york."
+        super().__init__(
+            name=name, role=role, llm=llm, actions=actions, logger=agent_logger
+        )
+        #self.__build_examples__()
+
+    def __build_examples__(self):
+        """
+        constructing the examples for agent working.
+        Each example is a successful action-obs chain of an agent.
+        those examples should cover all those api calls
+        """
+        # an example of search agent with wikipedia api call
+        # task
+        task = "place_name:ny"
+
+        # 1. think action and obs
+        thought = "I should first figure out ny"
+        act_1 = AgentAct(name=ThinkAct.action_name, params={INNER_ACT_KEY: thought})
+        obs_1 = ""
+'''
+agent_actions = []#get_intent(),get_user_current_date(), get_user_current_location(), get_latitude_longitude()]#, get_weather_forcast()]
+agent = BaseAgent(
+    name=agent_info["name"],
+    role=agent_info["role"],
+    llm=llm,
+    actions=agent_actions,
+    #reasoning_type="react",
+)
+prompt="the examples and results: turn up	intent:audio_volume_up\nwhat's the current weather	intent:weather_query\ni want you to remind me the next meeting with my girlfriend it will be at eight pm next sunday	intent:calendar_set,event_name : meeting,relation:girlfriend, time : eight pm, date : next Sunday\ncancel alarm for tenth of march two thousand seventeen	intent:alarm_remove,date : tenth of march two thousand seventeen\nhave you heard any good jokes lately	intent:general_joke\nget it fast resolved	intent:social_post\nalexa book me a train ticket for this afternoon to chicago	intent:transport_ticket,transport_type : train, timeofday : this afternoon, place_name : chicago\ni did not want you to send that text yet wait until i say send	intent:email_sendemail,setting:save\nwhat causes if i had junk food and alcohols	intent:general_quirky,content:what causes if i had junk food and alcohols\nfind a recipe for a romantic dinner for two	intent:cooking_recipe,food_descriptor: romantic dinner for two\nplease turn up the lights in this room	intent:iot_hue_lightup,house_place:this room\nwhats happening in pop industry	intent:news_query	,news_topic : pop industry\nis today the fourth or the fifth	intent:datetime_query,date : today\nthe wemo plug should be turned off on	intent:iot_wemo_off,device_type : wemo plug\nwhen i get home can you please order a pizza	intent:takeaway_order,food_type : pizza\nfind the events	intent:recommendation_events\nstart radio and go to frequency on one thousand and forty eight	intent:play_radio,radio_name : one thousand and forty eight\ni want to listen arijit singh song once again	intent:play_music,artist_name : arijit singh\nplease arrange to wake me up at three am alarm	intent:alarm_set,time : three am\nstart dune from where i left off	intent:play_audiobook,player_setting : resume, audiobook_name : dune\n please return the result of this sentence: "
+FLAG_CONTINUE = True
+while FLAG_CONTINUE:
+    input_text = input("Ask Intent Agent question:\n")
+    task = TaskPackage(instruction=prompt+"\n"+input_text)
+    agent(task)
+    if input("Do you want to continue? (y/n): ") == "n":
+        FLAG_CONTINUE = False
+'''

examples/agentlite/example/TimeAction.py

@@ -0,0 +1,44 @@
+import os
+
+
+from agentlite.actions.BaseAction import BaseAction
+from datetime import date, datetime
+
+class TimeAction(BaseAction):
+    def __init__(self) -> None:
+        action_name = "Time_Act"
+        action_desc = "Using this action to get time."
+        params_doc = {"query": "the search string. be simple."}
+        super().__init__(
+            action_name=action_name, action_desc=action_desc, params_doc=params_doc,
+        )
+
+    def __call__(self, query):
+        results = query
+        return results
+
+class CurDateAction(BaseAction):
+    def __init__(self) -> None:
+        action_name = "CurDate_Act"
+        action_desc = "Using this action to get current date."
+        params_doc = {"query": "the search string. be simple."}
+        super().__init__(
+            action_name=action_name, action_desc=action_desc, params_doc=params_doc,
+        )
+
+    def __call__(self, query):
+        results = date.today()
+        return str(results)
+    
+class CurTimeAction(BaseAction):
+    def __init__(self) -> None:
+        action_name = "CurTime_Act"
+        action_desc = "Using this action to get current time."
+        params_doc = {""}
+        super().__init__(
+            action_name=action_name, action_desc=action_desc, params_doc=params_doc,
+        )
+
+    def __call__(self, query):
+        results = datetime.now()
+        return results

examples/agentlite/example/TimeAgent.py

@@ -0,0 +1,71 @@
+import os 
+
+from typing import List
+from intent_actions import get_intent,get_user_current_date, get_user_current_location, get_latitude_longitude
+from agentlite.actions import BaseAction, FinishAct, ThinkAct
+from agentlite.agents import BaseAgent
+from agentlite.commons import TaskPackage
+from agentlite.llm.agent_llms import get_llm_backend, LLMConfig
+from agentlite.llm.agent_llms import BaseLLM
+from intent_actions import get_intent,get_user_current_date, get_user_current_location, get_latitude_longitude
+
+from TimeAction import TimeAction,CurDateAction
+from agentlite.logging.terminal_logger import AgentLogger
+agent_logger = AgentLogger(PROMPT_DEBUG_FLAG=False)
+# LAM_URL = os.environ["LAM_URL"]
+# print(LAM_URL)
+# llm_config = LLMConfig(
+#         {
+#             "llm_name": "xlam_v2", 
+#             "temperature": 0.0, 
+#             "base_url": LAM_URL,
+#             "api_key": "EMPTY"
+#         }
+#     )
+llm_name = "gpt-4"#3.5-turbo"#-16k"#-0613"#gpt-4"
+llm_config = LLMConfig({"llm_name": llm_name, "temperature": 0.0,
+        "api_key": "",})
+llm = get_llm_backend(llm_config)
+class TimeAgent(BaseAgent):
+    def __init__(
+        self, llm: BaseLLM, actions: List[BaseAction] = [TimeAction(),CurDateAction()], **kwargs
+    ):
+        name = "TimeAgent"
+        role = "read the time params, and convert to formated time with TimeAction. if has date, call the CurDateAction to get date, format should be 2024-11-20. the time is 10:00. if has time, the time format should be 10:00"
+        super().__init__(
+            name=name, role=role, llm=llm, actions=actions, logger=agent_logger
+        )
+        #self.__build_examples__()
+
+    def __build_examples__(self):
+        """
+        constructing the examples for agent working.
+        Each example is a successful action-obs chain of an agent.
+        those examples should cover all those api calls
+        """
+        # an example of search agent with wikipedia api call
+        # task
+        task = "send the email to this new email address"
+
+        # 1. think action and obs
+        thought = "I should first figure out the intent"
+        act_1 = AgentAct(name=ThinkAct.action_name, params={INNER_ACT_KEY: thought})
+        obs_1 = ""
+'''
+agent_actions = []#get_intent(),get_user_current_date(), get_user_current_location(), get_latitude_longitude()]#, get_weather_forcast()]
+agent = BaseAgent(
+    name=agent_info["name"],
+    role=agent_info["role"],
+    llm=llm,
+    actions=agent_actions,
+    #reasoning_type="react",
+)
+prompt="the examples and results: turn up	intent:audio_volume_up\nwhat's the current weather	intent:weather_query\ni want you to remind me the next meeting with my girlfriend it will be at eight pm next sunday	intent:calendar_set,event_name : meeting,relation:girlfriend, time : eight pm, date : next Sunday\ncancel alarm for tenth of march two thousand seventeen	intent:alarm_remove,date : tenth of march two thousand seventeen\nhave you heard any good jokes lately	intent:general_joke\nget it fast resolved	intent:social_post\nalexa book me a train ticket for this afternoon to chicago	intent:transport_ticket,transport_type : train, timeofday : this afternoon, place_name : chicago\ni did not want you to send that text yet wait until i say send	intent:email_sendemail,setting:save\nwhat causes if i had junk food and alcohols	intent:general_quirky,content:what causes if i had junk food and alcohols\nfind a recipe for a romantic dinner for two	intent:cooking_recipe,food_descriptor: romantic dinner for two\nplease turn up the lights in this room	intent:iot_hue_lightup,house_place:this room\nwhats happening in pop industry	intent:news_query	,news_topic : pop industry\nis today the fourth or the fifth	intent:datetime_query,date : today\nthe wemo plug should be turned off on	intent:iot_wemo_off,device_type : wemo plug\nwhen i get home can you please order a pizza	intent:takeaway_order,food_type : pizza\nfind the events	intent:recommendation_events\nstart radio and go to frequency on one thousand and forty eight	intent:play_radio,radio_name : one thousand and forty eight\ni want to listen arijit singh song once again	intent:play_music,artist_name : arijit singh\nplease arrange to wake me up at three am alarm	intent:alarm_set,time : three am\nstart dune from where i left off	intent:play_audiobook,player_setting : resume, audiobook_name : dune\n please return the result of this sentence: "
+FLAG_CONTINUE = True
+while FLAG_CONTINUE:
+    input_text = input("Ask Intent Agent question:\n")
+    task = TaskPackage(instruction=prompt+"\n"+input_text)
+    agent(task)
+    if input("Do you want to continue? (y/n): ") == "n":
+        FLAG_CONTINUE = False
+'''

examples/agentlite/example/iot_manager.py

@@ -0,0 +1,80 @@
+# get llm backend
+from agentlite.llm.agent_llms import get_llm_backend
+from agentlite.llm.LLMConfig import LLMConfig
+import csv
+import sys
+import json
+
+llm_config_dict = {
+    "llm_name": "gpt-4",#3.5-turbo-16k-0613",
+    "temperature": 0.0,
+     "api_key": "",
+}
+llm_config = LLMConfig(llm_config_dict)
+llm = get_llm_backend(llm_config)
+
+# define the individual agents
+from IntentAgent import IntentAgent
+from TimeAgent import TimeAgent
+
+
+from LocationAgent import LocationAgent
+from ExecAgent import ExecAgent
+
+
+intent_agent = IntentAgent(llm)
+time_agent = TimeAgent(llm)
+location_agent = LocationAgent(llm)
+excution_agent = ExecAgent(llm)
+
+
+# define the manager agent
+from agentlite.agents import ManagerAgent
+
+manager_agent_info = {
+    "name": "whole_manager",
+    "role": "you are controlling smart home system, you have intent_agent, time_agent, location_agent, and excution_agent to complete the user's task. You should first use intent_agent to complete the intent prediction. Then if the result has time or location params, please try to ask time_agent or location_agent to solve the time and location. If has currency name, you should convert the currency name to the name in the currency map. At last you should use excution_agent to send and receive request from other servers such as weather server and response to user to finalize the task. If the server's response is need further information, need to response to the user that need infomation, and your task is finished.",
+}
+iot_manager = ManagerAgent(
+    llm,
+    manager_agent_info["name"],
+    manager_agent_info["role"],
+    TeamAgents=[intent_agent, time_agent,location_agent,currency_agent,excution_agent],
+)
+
+# test the manager agent with TaskPackage
+from agentlite.commons import TaskPackage
+
+def read_data(file_path):
+    queries=[]
+    with open(file_path, newline='') as csvfile:
+      spamreader = csv.reader(csvfile, delimiter=',', quotechar='"')
+      count=0
+      for row in spamreader:
+        if count==0:
+            count+=1
+            continue
+        #print(row[2])
+        #print(len(row))
+        iid=row[0]
+        query={"iid":iid,"query":row[1]}#,"slots":slot,"domain":row[4]}
+        queries.append(query)
+    return queries
+
+if __name__=="__main__":
+  tasks=read_data("~/data/test.csv")
+  #print(tasks)
+  index = int(sys.argv[1])
+  tasks=tasks[index:index+1]
+  for task in tasks:
+  #for task in tasks:
+    test_task = json.dumps(task)
+    test_task_pack = TaskPackage(instruction=test_task, task_creator="User")
+    response = iot_manager(test_task_pack)
+    print(response)
+#test_task = "set notification from world news"
+'''
+test_task = "what is the weather today"
+test_task_pack = TaskPackage(instruction=test_task, task_creator="User")
+response = iot_manager(test_task_pack)
+print(response)