Module usage 

Uniform usage reporting across all providers.

This module provides the UsageReport type that normalizes token usage from all LLM providers into a consistent format. After each turn, the session loop extracts input_tokens and output_tokens from the UsageMetadata in the LLM response and produces a uniform UsageReport.

Provider Parity (Requirement 5.3)

Streaming token output and usage metadata (input/output tokens) SHALL be reported uniformly so the platform can meter cost per provider.

Different providers report usage with different field names:

• Gemini: prompt_token_count / candidates_token_count
• OpenAI: prompt_tokens / completion_tokens
• Anthropic: input_tokens / output_tokens

All of these are normalized into adk-core’s UsageMetadata by each provider’s client. This module further simplifies into UsageReport for the managed runtime’s uniform reporting.

Integration

The session loop calls UsageReport::from_usage_metadata after each turn to extract usage information. The platform can then use this for billing, monitoring, and cost tracking.

Structs

SessionUsageTracker

Accumulated usage tracking for an entire session.

UsageReport

Uniform usage report emitted after each turn.