model-usage

To analyze local CodexBar cost logs for per-model usage, run the model_usage.py script with --provider codex or --provider claude. It reads local JSON cost logs and generates per-model usage summaries, gracefully handling missing data and optional time filters.

Yes, you can get a current-model view of the most expensive active model from CodexBar data by running the script with --mode current. This quickly identifies the highest-cost active model without relying on external web dashboards.

Yes, the per-model cost breakdown supports both Codex and Claude providers. You specify the target provider using the --provider flag to aggregate usage and costs across multiple days for the selected provider.

The best way to output per-model cost summaries for dashboards or automation is generating structured JSON. The script provides output options for both human-friendly text and structured JSON formats to support reporting and automation workflows.

To aggregate costs by model across multiple days for team budgeting, run the script with --mode all. It processes historical local cost logs to produce full-model breakdowns, with optional time filters to narrow the reporting window.

If local model usage logs have missing data, the script gracefully handles the gaps without failing. It processes available JSON cost logs and applies optional time filters to ensure per-model usage summaries remain accurate.