v0.91.0

PromptingTools v0.91.0

Diff since v0.90.0

Breaking changes

None!

Added

• Added TokenUsage struct for standardized token usage tracking across all LLM providers (input/output tokens, cache tokens, reasoning tokens, cost, elapsed time). New usage field added to AIMessage, DataMessage, and AIToolRequest with backwards compatibility for legacy tokens, cost, elapsed fields.
• Added cache-aware cost calculation with provider-specific discounts via call_cost_with_cache and CACHE_DISCOUNTS.
• Added support for OpenAI's GPT-5.3 Codex model (gpt-5.3-codex, alias gpt53 / gpt53c). GPT-5.3 Codex is OpenAI's most capable agentic coding model combining frontier coding and reasoning capabilities. 400K context, 128K output. Requires the Responses API.
• Added support for Anthropic's Claude Opus 4.6 model (claude-opus-4-6, alias claudeo). Claude Opus 4.6 is Anthropic's latest Opus-class model with 1M context and 128K output.

Commits

Merged pull requests:

• New AIMessage.usage field (#327) (@svilupp)
• Add support for GPT-5.3 Codex and Claude Opus 4.6 models (#328) (@svilupp)