Module token 

Token estimation and usage tracking for agent sessions.

This module provides approximate token counting for messages and cumulative usage tracking across turns. Token estimation uses character-based heuristics:

• ASCII text: ~4 characters per token
• Non-ASCII text (CJK, etc.): ~2 characters per token

The request admission layer adds an explicit safety margin for text. Image parts use a conservative declared-size policy because dimensions are not always available at this boundary.

Structs

ModelPricing

Pricing information for a language model, expressed as cost per 1,000 tokens.

TokenEstimator

Estimates token counts for messages and tracks cumulative usage across turns.