I’m building a Python app that handles job interviews using AI agents. Right now I have basic functionality working but I want to add LangGraph to make the conversation flow better.
Here’s what my current code looks like:
import os
import random
from openai import OpenAI
from personality_types import PERSONALITY_PROFILES
def store_chat_history(messages):
file_id = ''.join(random.choices('abcdefghijklmnopqrstuvwxyz0123456789', k=6)) + '.log'
with open(file_id, 'w') as output_file:
for msg in messages:
if len(msg) == 3:
output_file.write(f"Context: {msg[0]}\nQuery: {msg[1]}\nAnswer: {msg[2]}\n\n")
else:
output_file.write(f"Query: {msg[0]}\nAnswer: {msg[1]}\n\n")
def get_user_name():
client = OpenAI()
greeting_prompt = "Create a professional greeting to ask for the interviewee's name. Make it sound natural and welcoming."
result = client.chat.completions.create(
model="gpt-3.5-turbo",
messages=[{"role": "user", "content": greeting_prompt}]
)
greeting_text = result.choices[0].message.content
user_name = input(f"\nAI: {greeting_text}\nYou: ").strip()
print(f"\nAI: Nice to meet you, {user_name}. Let's start the interview.\n")
return user_name
def collect_response(question_text):
while True:
answer = input(f"\nAI: {question_text}\n\nYou: ").strip()
if answer:
return answer
else:
print("AI: Could you please answer the question?")
interviewee_name = get_user_name()
I tried using Crew AI before but it was too complicated for what I need. Now I’m looking at LangGraph but I’m not sure how to add it to my existing code.
What I need help with:
How do I actually implement LangGraph with my current setup
Are there simpler alternatives that work well for interview bots
Best way to structure the conversation flow for behavioral questions
I’m pretty new to these AI frameworks so any step by step guidance would be really helpful. Thanks!
honestly langgraph might be overkill for your usecase. ive seen people overthink this stuff when simple state tracking works fine. maybe try just adding a conversation_state dict to track where you are in the interview flow before jumping into complex frameworks? your existing code is actually pretty solid for basic interviews.
I implemented LangGraph for a similar recruitment tool last year and found it particularly useful for maintaining conversation context across multiple interview rounds. The key is starting with their StateGraph class which handles the conversation flow much better than basic OpenAI calls. For your existing code, you’d replace the simple input/output pattern with nodes that represent different interview stages - greeting, technical questions, behavioral assessment, wrap-up. Each node can access previous responses through the shared state, which makes follow-up questions much more natural. One thing I learned the hard way is to keep your initial graph simple. Start with maybe 3-4 nodes maximum and add complexity later. LangGraph’s conditional edges work well for branching based on candidate responses - like asking deeper technical questions if someone shows strong knowledge in their initial answer. Regarding alternatives, I actually found LangGraph easier than CrewAI once you get past the initial learning curve. The documentation has improved significantly and there are good examples for chatbot scenarios. The state management alone made it worth switching from my previous setup that was similar to yours.
I’ve been working with LangGraph for about 6 months now and had a similar integration challenge. The biggest difference from your current approach is that LangGraph operates on a graph-based state machine rather than linear function calls. What worked for me was gradually migrating one piece at a time instead of rewriting everything. Start by wrapping your existing get_user_name() and collect_response() functions as LangGraph nodes, then connect them with edges. Your chat history storage can become part of the graph state, which actually simplifies things since LangGraph handles state persistence automatically. The real advantage comes when you want to implement dynamic interview paths - like adjusting question difficulty based on previous answers or circling back to topics that need clarification. I found that behavioral questions especially benefit from LangGraph’s ability to maintain context and create natural follow-ups. For a beginner-friendly approach, consider using LangGraph’s prebuilt templates for conversational agents rather than building from scratch. They have interview-specific examples in their community repo that might save you significant development time.