Personalization

When to Use

Use this module to make your agent "feel" like it knows the user. It explicitly manages preferences distinct from general facts. Perfect for:

• Communication style (Formal vs Casual)
• User interests and hobbies
• Settings and defaults (Dark mode, Metric system)

Key Operations

Code Example

We can add preferences manually or let the system infer them from a conversation transcript.

# 1. Explicitly add a preference
client.memory.personalization.add_preference(
    user_id="emma_2024",
    preference_key="communication_style",
    preference_value="Warm, supportive, and concise",
    source="onboarding_form"
)

# 2. Infer from interaction
# The system reads the chat and realizes she likes running
client.memory.personalization.infer_preferences(
    user_id="emma_2024",
    interaction_history=[
        {"role": "user", "content": "I really love my morning runs, they clear my head."}
    ]
)
# Result: Adds preference "interest: running" with high confidence

Key Takeaways

• Explicit vs Implicit: supports both "User told me" and "I guessed it".
• Conflict Resolution: Use reconcile() (not shown) to handle conflicting preferences (e.g., User liked running, now hates it).
• Personalization context is automatically injected into get_context() calls for relevant users.