Simple, stateful conversations with OpenAI's chat models, including truncation and token metering.
gptsession is a lightweight Python module for managing structured, multi-turn conversations with OpenAI’s ChatGPT models. It supports multiple concurrent sessions, automatic token tracking, message truncation, and full conversation recording.
pip install gptsession- Supports multiple parallel chat sessions
- Token-aware truncation based on model limits
- Tracks actual tokens sent and received
- Maintains full untruncated conversation history
- Clean, minimal API:
setup(),say(),system(),model(), etc.
import gptsession
gptsession.setup(open("mykey.txt").read().strip())
gptsession.new()
gptsession.model("gpt-3.5-turbo")
gptsession.system("You are a helpful assistant.")
reply = gptsession.say("Hello! Can you explain recursion?")
print("ChatGPT:", reply)
if gptsession.truncated():
print("[Note: Earlier messages were removed to stay within token limits.]")import gptsession as gpt
gpt.setup(open("mykey.txt").read().strip())
s1 = gpt.new()
gpt.system("You are a bird.")
s2 = gpt.new()
gpt.system("You are a robot.")
gpt.sel(s1)
print(gpt.say("Where do you live?"))
gpt.sel(s2)
print(gpt.say("Where do you live?"))MIT License
© 2025 Lion Kimbro