#ive added the system prompt to my agent but it is not calling the custom functions ive set up in my
obtuse juniper
okay dont respons yet let me show you what i used as system prompt
silver hatchBOT
obtuse juniper
You are a conversational agent. If the user mentions a city or asks to search for a city, extract the city name and call the function perform_city_search(city_name). Respond with the results of the function call.
silver hatchBOT
obtuse juniper
im using python sdk
silver hatchBOT
obtuse juniper
let me show you what my script looks like so far. i have to share it in two parts so wait tilll i say ive provided all relevant text.
silver hatchBOT
obtuse juniper
def handle_agent_response(response):
global driver # Ensure the driver instance is accessible
# Check if the LLM response indicates a city search
if "search city" in response.lower():
# Extract the city name from the response
city_name = response.lower().replace("search city", "").strip()
if not city_name:
print("Agent: Please specify a city name to search.")
return
# Call the perform_city_search function
result = perform_city_search(city_name)
if "error" in result:
print(f"Agent: Error: {result['error']}")
else:
# Format the results for user readability
rates = result.get("rates", [])
if not rates:
print(f"Agent: No rates found for {city_name}.")
else:
response = f"Found {len(rates)} rates for {city_name}:\n"
for rate in rates:
response += f"Hotel ID: {rate['hotel_id']}, Description: {rate['description']}, Price: {rate['price']}\n"
print(f"Agent: {response}")
else:
# Handle other responses
print(f"Agent: {response}")
okay, ill proved the rest
silver hatchBOT
obtuse juniper
conversation = Conversation(
client=client,
agent_id=AGENT_ID,
requires_auth=bool(API_KEY),
audio_interface=DefaultAudioInterface(),
callback_agent_response=lambda response: handle_agent_response(response),
callback_agent_response_correction=lambda original, corrected: print(f"Agent: {original} -> {corrected}"),
callback_user_transcript=lambda transcript: print(f"User: {transcript}"),
callback_latency_measurement=lambda latency: print(f"Latency: {latency}ms"),
)
Start the conversation session
def start_conversation():
print("Starting conversation...")
conversation.start_session()
try:
while True:
# Capture and send audio data
print("Listening for user input...")
audio_data = conversation.audio_interface.capture_audio()
conversation.send_audio(audio_data)
except KeyboardInterrupt:
pass
finally:
# End the session gracefully
conversation.end_session()
print("Conversation ended.")
in System Prompt Configuration:
i want to define how the LLM should interpret user inputs and decide when to call specific functions.
silver hatchBOT
obtuse juniper
ill make the llm prompt in the agent dashboard
silver hatchBOT
obtuse juniper
from this example how should I structure my following code to match that prompt function call. wait till ive provided all relevant code.
silver hatchBOT
obtuse juniper
def perform_city_search(city_name):
global driver
try:
# Wait for the search input field and enter the city name
search = WebDriverWait(driver, 10).until(
EC.presence_of_element_located((By.NAME, "City")) # Name of the input field
)
search.send_keys(city_name)
search.send_keys(Keys.RETURN)
# Wait for the page to load after searching
time.sleep(15)
# Wait for the 'Rates' div to be present
rates_div = WebDriverWait(driver, 60).until(
EC.presence_of_element_located((By.ID, "Rates"))
)
print("Found the 'Rates' div")
# Wait for the top 8 child divs within 'Rates' to be present
top_8_divs = WebDriverWait(driver, 60).until(
EC.presence_of_all_elements_located((By.XPATH, "//div[@id='Rates']/div[position() <= 8]"))
)
print("Found the top 8 divs within 'Rates'")
# Extract data from the top 8 divs
rates_data = []
for div in top_8_divs:
try:
# Extract price and description using a single XPath expression
elements = div.find_elements(
By.XPATH,
".//div[@class='price']/span | .//div[@class='description-column']"
)
if len(elements) >= 2:
description = elements[0].text if elements[0] else None
price = elements[1].text if elements[1] else None
else:
price = None
description = None
rates_data.append({
"hotel_id": div.get_attribute('id'),
"description": description,
"price": price
})
except Exception as e:
print(f"Error extracting data from div: {e}")
return {"error": f"Error extracting data from div: {e}"}
i am just showing you what i want to call the function to do so that you can assist in making sure i call the functions correctly. ill provide the rest of my code.
silver hatchBOT
obtuse juniper
okay ill do that. dont respond until ive stated that ive provided all relevant code
silver hatchBOT
obtuse juniper
def perform_city_search(city_name):
global driver
try:
# Wait for the search input field and enter the city name
search = WebDriverWait(driver, 10).until(
EC.presence_of_element_located((By.NAME, "City")) # Name of the input field
)
search.send_keys(city_name)
search.send_keys(Keys.RETURN)
# Wait for the page to load after searching
time.sleep(15)
# Wait for the 'Rates' div to be present
rates_div = WebDriverWait(driver, 60).until(
EC.presence_of_element_located((By.ID, "Rates"))
)
print("Found the 'Rates' div")
# Wait for the top 8 child divs within 'Rates' to be present
top_8_divs = WebDriverWait(driver, 60).until(
EC.presence_of_all_elements_located((By.XPATH, "//div[@id='Rates']/div[position() <= 8]"))
)
print("Found the top 8 divs within 'Rates'")
# Extract data from the top 8 divs
rates_data = []
for div in top_8_divs:
try:
# Extract price and description using a single XPath expression
elements = div.find_elements(
By.XPATH,
".//div[@class='price']/span | .//div[@class='description-column']"
)
if len(elements) >= 2:
description = elements[0].text if elements[0] else None
price = elements[1].text if elements[1] else None
else:
price = None
description = None
rates_data.append({
"hotel_id": div.get_attribute('id'),
"description": description,
"price": price
})
except Exception as e:
print(f"Error extracting data from div: {e}")
return {"error": f"Error extracting data from div: {e}"}
def handle_agent_response(response):
global driver # Ensure the driver instance is accessible
# Check if the LLM response indicates a city search
if "search city" in response.lower():
# Extract the city name from the response
city_name = response.lower().replace("search city", "").strip()
if not city_name:
print("Agent: Please specify a city name to search.")
return
# Call the perform_city_search function
result = perform_city_search(city_name)
if "error" in result:
print(f"Agent: Error: {result['error']}")
else:
# Format the results for user readability
rates = result.get("rates", [])
if not rates:
print(f"Agent: No rates found for {city_name}.")
else:
response = f"Found {len(rates)} rates for {city_name}:\n"
for rate in rates:
response += f"Hotel ID: {rate['hotel_id']}, Description: {rate['description']}, Price: {rate['price']}\n"
print(f"Agent: {response}")
else:
# Handle other responses
print(f"Agent: {response}")
what if i want the fuction calls to be determined by the llm based on the user input put not for the agent to have to communicate internal fuction calling or pprocess. I just watnt teh user to send input via the audio and for it to hear the response from the agent of what is extracted from the search or that it is currentyl working on it.
silver hatchBOT
obtuse juniper
the function callling is still not working. let me share my code. Dont answer until ive provided all relevant code and have stated so.
silver hatchBOT
obtuse juniper
import os
import signal
import sys
from elevenlabs.client import ElevenLabs
from elevenlabs.conversational_ai.conversation import Conversation
from elevenlabs.conversational_ai.default_audio_interface import DefaultAudioInterface
from selenium import webdriver
from selenium.webdriver.chrome.service import Service
from selenium.webdriver.common.by import By
from selenium.webdriver.common.keys import Keys
from selenium.webdriver.support.ui import WebDriverWait
from selenium.webdriver.support import expected_conditions as EC
import time
import re # Import the regular expression module
Set environment variables directly in the script
os.environ["AGENT_ID"] = "*************"
os.environ["ELEVENLABS_API_KEY"] = "*****************"
Load environment variables for Agent ID and API Key
AGENT_ID = os.getenv("AGENT_ID")
API_KEY = os.getenv("ELEVENLABS_API_KEY")
Debugging: Print environment variables
print(f"AGENT_ID: {AGENT_ID}")
print(f"ELEVENLABS_API_KEY: {API_KEY}")
if not AGENT_ID:
sys.stderr.write("AGENT_ID environment variable must be set\n")
sys.exit(1)
if not API_KEY:
sys.stderr.write("ELEVENLABS_API_KEY not set, assuming the agent is public\n")
Initialize the ElevenLabs client
client = ElevenLabs(api_key=API_KEY)
silver hatchBOT
obtuse juniper
PATH = "/Users/selvatica/chromedriver-mac-arm64/chromedriver"
service = Service(PATH)
driver = webdriver.Chrome(service=service)
driver.get("https://www.hotelplanner.com/Search/#dir-bar") # URL of the website
Perform a city search using Selenium
def perform_city_search(city_name):
global driver
try:
# Wait for the search input field and enter the city name
search = WebDriverWait(driver, 10).until(
EC.presence_of_element_located((By.NAME, "City")) # Name of the input field
)
search.send_keys(city_name)
search.send_keys(Keys.RETURN)
# Wait for the page to load after searching
time.sleep(15)
# Wait for the 'Rates' div to be present
rates_div = WebDriverWait(driver, 60).until(
EC.presence_of_element_located((By.ID, "Rates"))
)
print("Found the 'Rates' div")
# Wait for the top 8 child divs within 'Rates' to be present
top_8_divs = WebDriverWait(driver, 60).until(
EC.presence_of_all_elements_located((By.XPATH, "//div[@id='Rates']/div[position() <= 8]"))
)
print("Found the top 8 divs within 'Rates'")
rates_data = []
for div in top_8_divs:
try:
# Extract price and description using a single XPath expression
elements = div.find_elements(
By.XPATH,
".//div[@class='price']/span | .//div[@class='description-column']"
)
if len(elements) >= 2:
description = elements[0].text if elements[0] else None
price = elements[1].text if elements[1] else None
else:
price = None
description = None
rates_data.append({
"hotel_id": div.get_attribute('id'),
"description": description,
"price": price
})
except Exception as e:
print(f"Error extracting data from div: {e}")
return {"error": f"Error extracting data from div: {e}"}
print("Successfully extracted rates data")
return {"rates": rates_data}
except Exception as e:
print(f"Error performing city search: {e}")
return {"error": f"Error performing city search: {e}"}
def handle_agent_response(response):
global driver # Ensure the driver instance is accessible
# Use a regular expression for more flexible matching
match = re.search(r"(?:search|find|look for)\s+city\s+(.+)", response.lower())
if match:
# Extract the city name from the response
city_name = match.group(1).strip()
if not city_name:
print("Agent: Please specify a city name to search.")
return
# Inform the user that the search is in progress
print(f"Agent: Searching for rates in {city_name}...")
# Call the perform_city_search function
result = perform_city_search(city_name)
if "error" in result:
print(f"Agent: Error: {result['error']}")
else:
# Format the results for user readability
rates = result.get("rates", [])
if not rates:
print(f"Agent: No rates found for {city_name}.")
else:
response = f"Found {len(rates)} rates for {city_name}:\n"
for rate in rates:
response += f"Hotel ID: {rate['hotel_id']}, Description: {rate['description']}, Price: {rate['price']}\n"
print(f"Agent: {response}")
else:
# Handle other responses
print(f"Agent: {response}")
conversation = Conversation(
client=client,
agent_id=AGENT_ID,
requires_auth=bool(API_KEY),
audio_interface=DefaultAudioInterface(),
callback_agent_response=lambda response: handle_agent_response(response),
callback_agent_response_correction=lambda original, corrected: print(f"Agent: {original} -> {corrected}"),
callback_user_transcript=lambda transcript: print(f"User: {transcript}"),
callback_latency_measurement=lambda latency: print(f"Latency: {latency}ms"),
)
Start the conversation session
def start_conversation():
print("Starting conversation...")
conversation.start_session()
try:
while True:
# Capture and send audio data
print("Listening for user input...")
audio_data = conversation.audio_interface.capture_audio()
conversation.send_audio(audio_data)
except KeyboardInterrupt:
pass
finally:
# End the session gracefully
conversation.end_session()
print("Conversation ended.")
Handle SIGINT (Ctrl+C) to end the conversation gracefully
signal.signal(signal.SIGINT, lambda sig, frame: conversation.end_session())
if name == "main":
start_conversation()