chat___asyncchat

Chapter 4: Chat / AsyncChat

In Chapter 3: API Modules (Models, Chats, Files, Tunings, Caches, Batches, Operations, Live), we saw how the SDK organizes functionalities into modules like Models. While client.models.generate_content is powerful, using it directly for conversations requires manually managing the history: appending the user's message, sending the full history, and then appending the model's response. This chapter introduces Chat and AsyncChat, abstractions designed to simplify this common pattern.

Motivation and Use Case

Building conversational AI requires maintaining the context of the interaction over multiple turns. Sending only the latest user message to the model results in stateless, disconnected responses. The model needs the preceding dialogue to understand the context and respond coherently. Manually constructing the list[Content] history for every API call can be repetitive and error-prone, especially when handling potential errors or invalid responses from the model.

The Chat (and its asynchronous counterpart AsyncChat) provides a stateful object that automatically manages this conversation history. You send a message, and the Chat object handles appending it to the history, calling the underlying Models module with the full context, and storing the model's response, ready for the next turn.

Central Use Case: Imagine creating a simple chatbot that remembers previous interactions.

# Assuming 'client' is configured (e.g., using genai.Client(api_key=...))
from google import genai
from google.genai import types

# 1. Create a Chat session using the 'chats' module factory
# Note: client.chats is a factory for Chat instances
chat = client.chats.create(model='gemini-1.5-flash') # Or your preferred model

# 2. Send the first message
response1 = chat.send_message("Hello! My name is Alex.")
print(f"AI: {response1.text}")

# 3. Send the second message - the Chat object remembers the first turn
response2 = chat.send_message("What is my name?")
print(f"AI: {response2.text}") # The AI should know your name is Alex

# 4. Inspect the history managed by the Chat object
print("\n--- Chat History ---")
# Use get_history(curated=True) for only valid turns sent to the model
for content in chat.get_history(curated=False): # Shows all turns
    print(f"{content.role.capitalize()}: {content.parts[0].text}")

This example shows how chat.send_message handles the history. The second call automatically includes the context ("Hello! My name is Alex." and the AI's first response) when generating the answer to "What is my name?".