Token Budget Explorer¶
Run the Token Budget Explorer Fullscreen
About This MicroSim¶
An interactive token budget visualizer that makes context-window usage visible. A horizontal bar shows colored segments for each part of the prompt: System Role, Style Guide, Outline, Prior Chapters, Graph Subgraph, and In-Flight Response. A secondary panel shows the prompt-cache timer (300-second TTL) and whether scheduled runs land inside or outside the cache window.
Controls:
- Instruction Depth (1-10): Scales the Style Guide and Prior Chapters segments.
- Output Length (500-15,000 tokens): Sets the In-Flight Response segment size.
- Context Window (200K / 400K / 1M): Changes the budget bar scale.
- Run Cadence (60s-3600s): Shows whether runs hit or miss the prompt cache.
- Reset: Restores default values.
No real API calls are made — this is purely a budget visualizer.
Learning Objective (Bloom level: Apply): Given a target chapter length and a style-guide size, choose a run cadence that keeps the pipeline inside the prompt-cache window and the context-window bound.