Skip to content

06: Temporal Memory Consolidation

Overview

Turn a stream of fragmented events into a single "Daily Summary" record using Composite Keys. Demonstrates time-slicing aggregation where events throughout a day are consolidated into one coherent record with LLM-synthesized summaries.

Theme

Time-Series Data Aggregation

Strategy

Time-Slicing with LLM.BALANCED merge

Key Features

  • ✅ Daily aggregation of fragmented events
  • ✅ Temporal bucketing using composite keys
  • ✅ Time-aware consolidation
  • ✅ LLM-synthesized daily summaries
  • ✅ Trend analysis across time

Data Structure

DailyTrace

user: str                          # User identifier
date: str                          # Date (YYYY-MM-DD)
actions: list[str]                 # Accumulated actions
summary: str                       # LLM-synthesized summary
mood: str | None                   # Inferred mood/sentiment
key_events: list[str]              # Important events
productivity: str | None           # Productivity assessment
notes: list[str]                   # Additional notes

Use Case

Analytics & Logging: Consolidate streaming logs, user events, or telemetry into daily summaries. Useful for dashboards, analytics, and trend analysis across time periods.

Benefits: - Reduced data volume (1 summary per day instead of 1000s of events) - Better pattern recognition over time - Efficient storage and retrieval - Time-aware insights

Running the Example

cd examples/

# Set your OpenAI API key for LLM synthesis (optional)
export OPENAI_API_KEY="your-key-here"

python 06_temporal_memory_consolidation.py

Output

Results are stored in temp/temporal_memory/: - memory.json: Daily summary records - metadata.json: Schema and statistics - faiss_index/: Vector index for temporal search

API Requirements 🔄 Optional

This example works better with an OpenAI API key for LLM-powered synthesis, but can work without it for basic consolidation.

What You'll Learn

  1. Composite Keys: Using composite keys for time-aware grouping
  2. Time-Slicing: Aggregating events into time buckets
  3. LLM Synthesis: Generating summaries of events
  4. Temporal Patterns: Discovering patterns across time
  5. Stream Processing: Handling continuous event streams

Complexity

⭐⭐⭐ Intermediate: Shows advanced time-aware consolidation patterns.

Real-World Applications

  • Daily Activity Summaries: Convert 1000s of user actions into daily digests
  • System Monitoring: Aggregate logs into daily health reports
  • Analytics: Convert event streams into analyzable daily metrics
  • Trend Analysis: Identify patterns across days/weeks/months

Next Steps