chore(config): backfill serving component metadata

[cquil11]

Summary

• backfill source-verified router metadata across the active AMD and NVIDIA master configs
• identify multinode P2P KV transfer engines by name without exposing their package versions
• classify CPU/DRAM cache movement under kv-offload-backend instead of P2P transfer
• distinguish vllm-native from vllm-simple, omit versions for framework-native vLLM/HiCache backends, and retain exact independent versions for LMCache and Mooncake
• backfill the Qwen3.5 GB300 Dynamo/SGLang config added on main

Validation

• python -m pytest utils/matrix_logic/ utils/test_process_result.py utils/agentic/aggregation/test_process_agentic_result.py -q
• full NVIDIA and AMD master-config generation

[github-actions]

Thanks for the contribution! Please reach out to respective companies' CODEOWNER to fill in the latest PR_REVIEW_CHECKLIST.md before pinging core maintainer on Slack for review. In order for the signoff PR check bot to trigger, you must follow the PR_REVIEW_CHECKLIST.md template correctly, including the phrase As a PR reviewer and CODEOWNER, I have reviewed this and have.

For PR verification, add the full-sweep-fail-fast label (strongly recommended) to this PR — the benchmark sweep only runs on labeled PRs. Use full-sweep-enabled only if you need matrix jobs to keep running past a failure.

PR authors are responsible for ensuring that after merging, all GitHub Action jobs fully pass. A lot of the time, failures are just flakes and simply re-running the failed jobs will fix it. See GitHub's docs on re-running failed jobs

---

感谢你的贡献！请联系相应公司的 CODEOWNER 填写最新的 PR_REVIEW_CHECKLIST.md，然后再在 Slack 上联系核心维护者进行审阅。为了触发 signoff PR 检查机器人，你必须正确遵循 PR_REVIEW_CHECKLIST.md 模板，包括保留英文语句 As a PR reviewer and CODEOWNER, I have reviewed this and have。

如需进行 PR 验证，请为此 PR 添加 full-sweep-fail-fast 标签（强烈推荐）— 基准测试 sweep 仅在带有标签的 PR 上运行。仅当需要矩阵任务在失败后继续运行时才使用 full-sweep-enabled。

PR 作者有责任确保合并后所有 GitHub Action 任务完全通过。 很多时候失败只是偶发抖动（flake），重新运行失败的任务即可解决。参见 GitHub 关于重新运行失败任务的文档

[github-actions]

Thanks for the contribution! Please reach out to respective companies' CODEOWNER to fill in the latest PR_REVIEW_CHECKLIST.md before pinging core maintainer on Slack for review. In order for the signoff PR check bot to trigger, you must follow the PR_REVIEW_CHECKLIST.md template correctly, including the phrase As a PR reviewer and CODEOWNER, I have reviewed this and have.

For PR verification, add the full-sweep-fail-fast label (strongly recommended) to this PR — the benchmark sweep only runs on labeled PRs. Use full-sweep-enabled only if you need matrix jobs to keep running past a failure.

PR authors are responsible for ensuring that after merging, all GitHub Action jobs fully pass. A lot of the time, failures are just flakes and simply re-running the failed jobs will fix it. See GitHub's docs on re-running failed jobs

---

感谢你的贡献！请联系相应公司的 CODEOWNER 填写最新的 PR_REVIEW_CHECKLIST.md，然后再在 Slack 上联系核心维护者进行审阅。为了触发 signoff PR 检查机器人，你必须正确遵循 PR_REVIEW_CHECKLIST.md 模板，包括保留英文语句 As a PR reviewer and CODEOWNER, I have reviewed this and have。

如需进行 PR 验证，请为此 PR 添加 full-sweep-fail-fast 标签（强烈推荐）— 基准测试 sweep 仅在带有标签的 PR 上运行。仅当需要矩阵任务在失败后继续运行时才使用 full-sweep-enabled。

PR 作者有责任确保合并后所有 GitHub Action 任务完全通过。 很多时候失败只是偶发抖动（flake），重新运行失败的任务即可解决。参见 GitHub 关于重新运行失败任务的文档

[claude]

LGTM — purely additive metadata backfill on YAML configs, no code changes.

Extended reasoning...

Overview

This PR backfills two optional metadata fields (router and kv-transfer, each with name + version) across four YAML config files: configs/amd-master.yaml, configs/nvidia-master.yaml, and the two deprecated minimaxm2.5-m2.7-master.yaml archives. Values are set either at the entry level (for uniform disaggregated setups) or on individual search-space rows where components differ per-config. It follows commit 438b018 ("feat(config): add serving component metadata") which introduced the schema.

Security risks

None. This is data-only YAML metadata. router / kv-transfer are declared Optional[ComponentMetadata] in utils/matrix_logic/validation.py, so no schema/code path changes. The version: strings are descriptive labels (many are just "image:<existing-image-ref>"), not resolved as URLs or executed anywhere.

Level of scrutiny

Very low. Every image:<ref> value mirrors the same-entry pre-existing image: field (including the entry-by-entry nvcr.io/nvidia/... vs nvcr.io#nvidia/... split, which is the pre-existing enroot-URI convention in these files). Non-image versions (0.1.14, 0.3.11.post1, nightly-20260617-e667ebb, git SHAs) are plain descriptive labels for components not built from a single image.

Other factors

No prior human reviews or bot findings; timeline only contains the two duplicate recipe-reminder bot comments. Config-only chore that's a natural follow-up to a landed feat commit — safe to shadow-approve.