How to maintain conversation history in PandasAI V3?

[muhammadhamzaazhar]

@gventuri could you please guide me on how to implement and maintain conversational memory or chat history within PandasAI? I'm looking to preserve context across multiple interactions to enable more coherent and context-aware responses.

[xXMrNidaXx]

PandasAI V3 has built-in memory support via the Agent class.

Basic conversation memory:

from pandasai import Agent
import pandas as pd

df = pd.read_csv("data.csv")
agent = Agent(dfs=[df])

# Each chat call maintains history
response1 = agent.chat("What is the average sales?")
response2 = agent.chat("Now filter to just Q4")  # Remembers context!
response3 = agent.chat("Show me the top 5")  # Still has context

Access conversation history:

# Get all messages
history = agent.conversation_history
for msg in history:
    print(f"{msg.role}: {msg.content}")

# Clear when needed
agent.clear_conversation()

Persist across sessions:

import json

# Save
with open("chat_history.json", "w") as f:
    json.dump([m.dict() for m in agent.conversation_history], f)

# Restore (on new agent)
agent = Agent(dfs=[df])
with open("chat_history.json") as f:
    history = json.load(f)
    for msg in history:
        agent.conversation_history.append(Message(**msg))

Tips:

• Memory grows token usage — clear periodically for long sessions
• Use agent.clarification_questions for ambiguous queries
• The agent tracks generated code + results, not just text

We build data analysis agents at Revolution AI — PandasAI V3 memory works well for multi-turn analysis. The key is managing context window size for longer conversations.