Question 1

How does the Context Window Budget Planner work?

Accepted Answer

Select a model context window size, then adjust sliders for each budget section. The pie chart updates in real time, and warnings appear if any section exceeds 50% of the total budget. Use presets to auto-allocate tokens for common use cases.

Question 2

What are the preset allocations for chatbot, RAG, code generation, and data extraction?

Accepted Answer

Chatbot: 15% system, 5% tools, 10% RAG, 50% history, 10% examples, 10% output. RAG: 10% system, 5% tools, 50% RAG, 20% history, 5% examples, 10% output. Code generation: 20% system, 10% tools, 10% RAG, 25% history, 20% examples, 15% output. Data extraction: 25% system, 15% tools, 30% RAG, 10% history, 10% examples, 10% output.

Question 3

How is cost per request calculated?

Accepted Answer

Cost per request assumes $0.01 per 1K tokens for the full context window. Trimming reduces token usage proportionally, and cost savings are shown as a percentage and dollar amount.

Context Window Budget Planner

Budget allocation

Allocation pie