Personalization Engine

Purpose

The Personalization Engine ensures the agent adapts to the specific user over time. It infers preferences from interactions and explicitly reconciles conflicting information to build a coherent user model.

Key Features

• Inference: Automatically infers preferences from chat hints (e.g., "I hate Python" → lang: python, sentiment: negative).
• Reconciliation: Resolves conflicting preferences via LLM strategies (e.g., changing taste over time).

Functional Deep Dive

Inference

The engine analyzes chat history to find patterns. If a user repeatedly asks for Python code, it infers prefers_language: python without the user explicitly stating it.

Reconciliation

Uses an LLM to merge conflicting preferences. If an old preference says "Likes Java" and a recent one says "Hates Java", the engine updates the graph to "Hates Java" with high confidence, treating it as a preference update.

Inference Diagram

API Reference

Personalization Engine provides endpoints for managing user preferences.

SDK Reference

The Functor SDK provides a Python interface for personalization operations.

Managing Preferences

from functor_sdk import FunctorClient

client = FunctorClient(api_key="your-api-key")

# Add preference
client.memory.personalization.add_preference(
    user_id="user_123",
    preference_key="language",
    preference_value="python",
    source="explicit",
    confidence=1.0
)

# Get preferences
prefs = client.memory.personalization.get_preferences(user_id="user_123")

# Get personalized context
context = client.memory.personalization.get_context(
    user_id="user_123",
    query="show me examples",
    limit=10
)

# Infer preferences from history
inferred = client.memory.personalization.infer_preferences(
    user_id="user_123",
    interaction_history=[...]
)

# Reconcile conflicting preferences
reconciled = client.memory.personalization.reconcile(
    user_id="user_123",
    new_entities=[...],
    strategy="llm_guided"
)