	In Browser
	StumbleUpon
	del.icio.us
	Google
	Google Buzz
	reddit
	LinkedIn

	Facebook
	Twitter
	Linkedin
	E-Mail

Generative AI > Ollama > Multi-Turn Conversations with Ollama in Python

Multi-Turn Conversations with Ollama in Python

Author: Venkata Sudhakar

Ollama supports multi-turn conversations where the model remembers previous messages in the same session. This is essential for building chatbots that can ask clarifying questions, remember context from earlier in the conversation, and provide coherent follow-up responses. The messages list in the ollama.chat() call serves as the conversation history - you append each user and assistant message to maintain context. At ShopMax India, multi-turn chat powers the in-store kiosk assistant that helps customers compare products across multiple questions.

Managing conversation history requires keeping track of the messages list across turns. Each exchange adds a user message and an assistant message to the list. The model receives the full history on every call so it can reference earlier statements. For long conversations, you may want to trim old messages to stay within the model context window.

The below example shows how to build a multi-turn product advisor chatbot using Ollama.

import ollama

# ShopMax India - multi-turn product advisor chatbot
system_prompt = (
    "You are a helpful product advisor for ShopMax India, "
    "an electronics retailer. Help customers choose products. "
    "Keep each response under 50 words."
)

# Initialize conversation with system message
conversation = [
    {"role": "system", "content": system_prompt}
]

def chat(user_message):
    # Add user message to history
    conversation.append({"role": "user", "content": user_message})

# Get model response
    response = ollama.chat(model="llama3.2", messages=conversation)
    assistant_reply = response["message"]["content"]

# Add assistant reply to history
    conversation.append({"role": "assistant", "content": assistant_reply})
    return assistant_reply

# Simulate a multi-turn conversation
print("User: I need a laptop for video editing.")
print("Assistant:", chat("I need a laptop for video editing."))
print()
print("User: What is my budget range for that?")
print("Assistant:", chat("What is my budget range for that?"))
print()
print("User: Do you have anything with a dedicated GPU?")
print("Assistant:", chat("Do you have anything with a dedicated GPU?"))

It gives the following output,

User: I need a laptop for video editing.
Assistant: For video editing, the ShopMax CreatorBook Pro with
an i9 processor, 32GB RAM, and NVIDIA RTX 4060 GPU is ideal.
Priced from Rs 1,10,000 at ShopMax stores in Mumbai and Bangalore.

User: What is my budget range for that?
Assistant: For professional video editing, plan a budget between
Rs 80,000 and Rs 1,50,000 depending on the resolution and
complexity of your projects.

User: Do you have anything with a dedicated GPU?
Assistant: Yes, the ShopMax CreatorBook Pro I mentioned includes
a dedicated NVIDIA RTX 4060 GPU with 8GB VRAM, which handles
4K editing smoothly in Premiere Pro and DaVinci Resolve.

Notice that in the third response the model correctly refers back to the product it mentioned in the first response - this is the multi-turn context at work. For production deployments at ShopMax India, store the conversation list in a session object tied to each customer interaction so that different customers have independent conversation histories. Clear the history when the session ends to free memory and protect privacy.

Send your comments, suggestions or queries regarding this site to [email protected].