Support GPT-6.1 Sol pricing and release v1.5.5

[capisoft-lib]

GPT-6.1 Sol calls were collected but remained unrated. Add the distinct gpt-6.1-sol model to the shared dated catalog from its September 29 launch: API $2 / $0.10 / $10 and Codex 50 / 2.5 / 250 credits per million input / cached input / output tokens.

Handle cache writes, the strict >272K API context threshold and Fast independently. Sol 6.1 purchased-credit Fast totals use 2x; subscription quota calibration/forecasts use 2.5x. Earlier model periods and saved custom settings retain their behavior. Pin the previously observed September 23 Fast gates so advancing catalog verification does not remove historical coverage.

Prepare v1.5.5 and updated Docker documentation. Generated local/Sites assets are built from the shared source and stay untracked.

Validation: 323 root tests; 15 hosted tests and hosted build; hosted lint/typecheck; Docker build and read-only relational persistence/local collector smoke test. Ten new regressions cover boundaries, precise IDs, custom settings, cache accounting, both Fast tier names, long context, report reproducibility, collector/Mesh preservation, SQL filters/aggregation and quota weighting.

Evidence: API pricing, Sol 6.1 model, September 29 launch, Codex credits, Fast billing.

Day-only launch dates use the existing 00:00 UTC convention, with launch-day calls marked estimated. Batch/Flex/Ultrafast and undocumented aliases remain unrated; tool and regional fees remain outside token estimates. The newly documented credit/subscription distinction is scoped to Sol 6.1 rather than retroactively rewriting previously recorded model periods.

Development-tool dependencies also receive bounded security patches: undici 7.29.1, brace-expansion 1.1.21/5.0.12 and fast-uri 3.1.8. Application/framework versions remain pinned. Both dependency audits report zero vulnerabilities; Mesh tests/build and hosted tests/build/lint/typecheck pass with the patched dependencies.

[capisoft-lib]

Self-review completed on commit 7dcbc97.

No blocking findings. Checked the exact model identity (no broad prefix fallback), UTC launch boundaries, preservation of Sol 6 and custom settings, cache writes charged once, strict >272K API context bands, historical Fast gates, and the distinction between purchased credits and subscription quota weighting in both browser and server forecasts. Verified that the shared source reaches generated local/Sites bundles and Docker runtime imports; the collector/Mesh contract needs no format change.

Validation: 323 root tests, 15 hosted tests/build, hosted lint/typecheck, Docker build, and read-only Docker persistence/replacement/local-collector smoke test passed. The 10 new regressions cover the new model through collection, Mesh, SQL aggregation, reports and quota calibration.

The rate-card evidence and limits are documented in the PR. Previously recorded model periods remain intact. Merge is contingent on both GitHub CI jobs passing on this exact commit and the branch being up to date with main.

[capisoft-lib]

Self-review completed on commit b8fc451.

No blocking findings. Checked the exact model identity (no broad prefix fallback), UTC launch boundaries, preservation of Sol 6 and custom settings, cache writes charged once, strict >272K API context bands, historical Fast gates, and the distinction between purchased credits and subscription quota weighting in both browser and server forecasts. Verified that the shared source reaches generated local/Sites bundles and Docker runtime imports; the collector/Mesh contract needs no format change.

Validation: 323 root tests, 15 hosted tests/build, hosted lint/typecheck, Docker build, and read-only Docker persistence/replacement/local-collector smoke test passed. The 10 new regressions cover the new model through collection, Mesh, SQL aggregation, reports and quota calibration.

The rate-card evidence and limits are documented in the PR. Previously recorded model periods remain intact. Merge is contingent on both GitHub CI jobs passing on this exact commit and the branch being up to date with main.

Reviewed the bounded development-only dependency patches as well: undici 7.29.1 and compatible brace-expansion/fast-uri patches. No application/framework upgrade. Both audits report zero vulnerabilities, and Mesh tests/build plus hosted tests/build/lint/typecheck pass after the changes.