[sernote]

Orca-Code-Review — push 1

Severity	Count
P0	0
P1	0
P2	0
P3	0

✅ no blocking findings

OpenAI-compatible requests silently lost prompt_cache_key and prompt_cache_options during request validation, before reaching LiteLLM. Explicit cache mode and request TTL therefore never reached the selected provider.

Declare the optional key and typed options in the request schema. Preserve supplied members without inserting defaults; reject invalid mode/TTL values. This covers blocking and streaming Chat Completions. prompt_cache_retention is outside this change.

Validation: 831 tests passed in the offline Python 3.12 suite. Regressions exercise the HTTP route, schema, Orca adapter, real LiteLLM Router and OpenAI SDK through HTTP MockTransport. Independent review ran 43 targeted tests, including additional invalid-input and serialized-payload checks. Ruff and git diff --check passed. No live provider cache hit or savings claim.

Discovered during a repository audit using audit-prompt-caching, then independently reproduced and reviewed.

[orcacode-review]

🐳 OrcaCode Review

✅ No findings — nothing to flag in this PR. Great work!

_{OrcaCode Review — Route Smarter. Ship Safer. Spend Less.
Engine-reported: 179 calls · 4.6M tokens · 92% cached}

❤️ Share · Install OrcaCode Review

Free on GitHub — the review runs on your own OrcaRouter key. If it helped, a shout-out goes a long way.

Share: X · Reddit · LinkedIn
Follow: X · Discord · LinkedIn · OrcaRouter