Preserve exact GPT-5 Codex pricing keys

[iam-brain]

Summary

• Stop collapsing Codex/OpenAI GPT-5 model IDs onto base GPT-5 entries during pricing lookup. We were previously stripping -codex and effectively pricing persisted usage as a different model than the one the scanner recorded, which made local cost reports harder to trust and hid model-specific behavior in breakdowns.
• Expand the Codex pricing catalog to include the missing non-chat GPT-5 family entries that now appear in real persisted usage data. This keeps pricing aligned with the exact model IDs we ingest instead of relying on partial aliases or nonexistent base-model fallbacks.
• Treat gpt-5.3-codex-spark as a zero-cost Research Preview case. The goal is to keep preview usage visible in scanner output, CLI payloads, and chart details without incorrectly charging it or dropping it from breakdowns just because the cost is zero.
• Preserve exact model IDs end to end through normalization, scanner aggregation, JSON payloads, and cost-history detail rendering. We took the explicit-table approach here so the supported pricing surface is obvious in code, dated suffixes are only stripped when the exact undated model exists, and we do not silently map future GPT-5 variants onto the wrong model.

Validation

• swift test --filter 'CostUsagePricingTests|CostUsageScannerTests|CostUsageDecodingTests|CLICostTests|UsageFormatterTests'
• pnpm check
• ./Scripts/compile_and_run.sh

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 3e74f2e168

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[ratulsarna]

@codex review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 3e74f2e168

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[ratulsarna]

Thanks @iam-brain !

[iam-brain]

Thought I asked this earlier but forgot @ratulsarna, bumping Codex's cache to v2 to clear the old cache so sessions can revalidate using the new models was the smallest change doable to introduce the new models and wanted to ask your opnion on that. Keep that small bandaid or should something be introduced to flush the cache with each model update?