Skip to content

TokenPak Demo Data

Quickly showcase TokenPak with realistic demo data — no need to generate real traffic.

Usage

Seed demo data

python3 demo.py --seed

Populates the database with: - 500 requests over 24 hours - A realistic cache hit rate (modeling repeated queries) - Multiple models (Haiku, Sonnet, GPT-4) - Cost breakdown, latency, cache efficiency - Labeled as "Demo Data" for clarity

Custom seed size

python3 demo.py --seed 1000 12  # 1000 requests over 12 hours

Check demo data

python3 demo.py --check

Shows: - Total requests - Cache hit rate and percentage - Time range - Cost breakdown (misses vs. hits) - Estimated savings

Clear demo data

python3 demo.py --clear

Removes all demo data — safe to run before a live demo.

Dashboard Integration

Once seeded, the TokenPak dashboard automatically displays: - Request timeline (scatter plot over 24h) - Cache hit rate (a high hit rate on repeated queries) - Cost breakdown by model - Latency comparison (cache hits respond much faster) - Savings estimate (realistic for production scenarios)

Data Isolation

  • Demo data is labeled internally (agent_id: "demo-data")
  • Real traffic never mixes with demo data
  • Cleanup is safe — only removes records marked as demo
  • Idempotent — seed multiple times without duplicates

Data Isolation Guarantees

Demo data is fully isolated from real traffic and safe to seed or clear at any time — see the Data Isolation notes above.