Allow llama.cpp reasoning preservation flags

[cmizzi]

This PR adds --reasoning-preserve and --no-reasoning-preserve to the llama.cpp runtime flag allowlist.

These flags are supported by llama.cpp and are boolean-only options. They do not access files or external resources, so they fit the existing security model used for runtime flag validation. --reasoning-preserve is important for reasoning models because it keeps the reasoning trace in the conversation history, allowing prompt/KV cache reuse when combined with prompt caching.

Without this change, Docker Model Runner rejects the flags before llama.cpp is started.

[gemini-code-assist]

Caution

The consumer version of Gemini Code Assist on GitHub has been sunset. All code review activity has officially ceased.

[sourcery-ai]

Hey - I've reviewed your changes and they look great!

---

Sourcery is free for open source - if you like our reviews please consider sharing them ✨

• X
• Mastodon
• LinkedIn
• Facebook

_{Help me be more useful! Please click 👍 or 👎 on each comment and I'll use the feedback to improve your reviews.}

[Copilot]

Copilot review overview

🟢 Approval recommended

All reviewed changes are covered and no blocking issues remain.

Review effort: Lite
Findings: None

What changed in this PR

Adds llama.cpp reasoning-preservation flags to runtime validation, enabling reasoning trace retention for prompt/KV cache reuse.

Changes:

• Allowlists --reasoning-preserve and --no-reasoning-preserve.
• Adds validation and category test coverage.

File	Description
pkg/​inference/​runtime_flags_test.go	Tests runtime validation for both flags.
pkg/​inference/​runtime_flags_allowlist.go	Adds the two flags to the llama.cpp allowlist.
pkg/​inference/​runtime_flags_allowlist_test.go	Covers reasoning-category membership.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[Copilot]

Copilot review overview

🟢 Approval recommended

No unresolved review issues remain.

Review effort: Lite
Findings: None

[Copilot]

Copilot review overview

🟢 Approval recommended

No unresolved issues were identified, and the changes include relevant test coverage.

Review effort: Lite
Findings: None

[ericcurtin]

Good addition, but it now conflicts with main since #1065 touched the same allowlist map. Please rebase, nothing else to fix. Also, model-runner is being deprecated in favor of llmman, please open future PRs there instead. Marking as draft, please mark ready for review once rebased.