wehub-resource-sync
56 lines
2.8 KiB
Python
56 lines
2.8 KiB
Python
"""Demo comparison between Traditional LLM and Parlant agent responses."""
|
|
import asyncio
|
|
from rich_table_formatter import print_comparison_rich
|
|
from traditional_llm_prompt import call_traditional_llm as traditional_call, TRADITIONAL_HUGE_PROMPT
|
|
from parlant_client_utils import (
|
|
create_client as create_parlant_client,
|
|
create_session as create_parlant_session,
|
|
send_user_message as send_parlant_user_message,
|
|
await_ai_reply as await_parlant_ai_reply,
|
|
get_session_reasoning as get_parlant_reasoning,
|
|
)
|
|
|
|
|
|
async def main() -> None:
|
|
"""Compare Traditional LLM vs Parlant agent responses."""
|
|
demo_queries = [
|
|
"I want to replace my existing $500k term policy with a whole life policy. What should I do?",
|
|
"I'm 35 years old, make $80,000 a year, and have 2 kids. How much life insurance coverage should I get?",
|
|
"I have diabetes. Will this affect my life insurance rates?",
|
|
"I'm really confused about insurance. My car got totaled last week and I need to file a claim, but I also want to know about life insurance for my business, and my wife is asking about health insurance options. Can you help me with all of this?",
|
|
"I'm thinking about getting life insurance but I'm not sure if I should. I'm 30 years old, healthy, and make $60,000 a year. I don't really want to spend a lot on premiums, but I also want to make sure my family is protected. What do you think I should do?",
|
|
]
|
|
|
|
import os
|
|
agent_id_path = os.path.join("parlant-data", "agent_id.txt")
|
|
if not os.path.exists(agent_id_path):
|
|
raise RuntimeError("agent_id.txt not found. Please start parlant_agent_server.py first.")
|
|
with open(agent_id_path, "r", encoding="utf-8") as f:
|
|
agent_id = f.read().strip()
|
|
|
|
client = await create_parlant_client()
|
|
rows = []
|
|
|
|
for i, query in enumerate(demo_queries, 1):
|
|
print(f"🔄 Processing query {i}/{len(demo_queries)}: {query[:50]}...")
|
|
|
|
session_id = await create_parlant_session(client, agent_id)
|
|
|
|
print(" 📝 Getting traditional LLM response...")
|
|
traditional_response = await traditional_call(query, TRADITIONAL_HUGE_PROMPT)
|
|
|
|
print(" 🤖 Getting Parlant agent response...")
|
|
customer_event_offset = await send_parlant_user_message(client, session_id, query)
|
|
min_offset = customer_event_offset + 1
|
|
parlant_response = await await_parlant_ai_reply(client, session_id, min_offset) or "Error: No AI reply received from Parlant session."
|
|
reasoning = await get_parlant_reasoning(client, session_id, min_offset)
|
|
|
|
print(f" ✅ Query {i} complete")
|
|
rows.append([query, traditional_response, parlant_response, reasoning])
|
|
|
|
print_comparison_rich([], rows)
|
|
|
|
|
|
if __name__ == "__main__":
|
|
asyncio.run(main())
|