Record MiniCPM-o throughput search, quality gates and peer baselines

docs/contributing.md

@@ -159,9 +159,10 @@ campaign, template, source-import, review-summary and external-archive layouts;
 new directories and recognizable manifests require an explicit classification.
 The nightly lane runs this CPU-only check. It verifies available bytes using
 existing bundle SHA-256 maps, historical file mappings or archive SHA-256/size
-records. Historical mappings point only to preserved snapshots, never current
-implementations. Do not invent hashes or update frozen manifests to silence a
-failure.
+records. A descriptive manifest classifies a JSON record that is not a digest
+map; the checker does not invent or verify content hashes for it. Historical
+mappings point only to preserved snapshots, never current implementations. Do
+not invent hashes or update frozen manifests to silence a failure.
 
 The JSON result separates structural/hash errors (nonzero exit) from unavailable
 external dependencies. Remote archives are not downloaded; if supplied at their

docs/reference-matrix.md

@@ -552,3 +552,11 @@ create a repo runtime descriptor, backend port or measured resource effect.
 | Source / revision | Inspected scope | Extraction and qualification boundary |
 | --- | --- | --- |
 | vllm-project/vllm-ascend at `3b31886237ed65c435e0965b10e8002ff5769766` | Release table, plugin hooks, DeepSeek-V4/DSpark evolution, Kimi K3/MiniMax-M3/GLM5+ optimization, DSA-CP/FlashComm1/Fused-MC2 and fusion dispatch, reversals and numerical oracles. [History screening and semantic scope](../src/inferrsi/data/skill_library/repositories/vllm-ascend/references/history-scope.md). [Source locators](../src/inferrsi/data/skill_library/repositories/vllm-ascend/references/sources.md) | [Release-specific diagnosis](../src/inferrsi/data/skill_library/repositories/vllm-ascend/SKILL.md). Candidate; device execution and downstream efficacy unverified. Older ACLGraph and checkpoint-adaptation owners retain their own source revisions. |
+
+## Issue #49: conditioning transport diagnostic
+
+[Owner-local case](../src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/minicpm-conditioning-transport.md) preserves the pinned ordinary-turn bridge,
+worker list traversal, invalidated high-rate profiler and rehashed CPU component
+measurements. This is an internal diagnostic lineage, not independent evidence
+from the peer engine. Component value equality and synthetic timing do not
+establish service throughput or speech-quality acceptance.

docs/searchspace-matrix.md

@@ -622,3 +622,10 @@ NOASSERTION and InferRSI's repository license.
 | Accelerator / Host | Diagnose engine-process plugin discovery and MiniMax-M3 mixed KV storage from hardware capability through per-layer resolution | [vllm-ascend](../src/inferrsi/data/skill_library/repositories/vllm-ascend/SKILL.md), [cache path](../src/inferrsi/data/skill_library/repositories/vllm-ascend/references/minimax-cache.md) | Entry-point metadata and cache specs can be inspected separately from device dispatch and numerics. |
 | Communication / Accelerator | Verify DSpark draft budget, effective P/D ranks and target capture tiers | [vllm-ascend](../src/inferrsi/data/skill_library/repositories/vllm-ascend/SKILL.md), [DSpark PD](../src/inferrsi/data/skill_library/repositories/vllm-ascend/references/dspark-pd.md) | TP4 acceptance-length fixture and two-node 950DT deployment have separate evidence requirements. |
 | Accelerator / RAM / Host | Diagnose fused preprocessing, grouped GEMM/activation/quant, residual specialization, KV index reuse and reverted optimizations | [vllm-ascend](../src/inferrsi/data/skill_library/repositories/vllm-ascend/SKILL.md), [fusion](../src/inferrsi/data/skill_library/repositories/vllm-ascend/references/fused-operators.md), [model performance](../src/inferrsi/data/skill_library/repositories/vllm-ascend/references/model-performance.md) | Whole-history discovery and selected semantic review remain separate from native performance validation. |
+
+## Issue #49: Host Observe / Diagnose / Verify
+
+[Conditioning transport case](../src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/minicpm-conditioning-transport.md) covers scalar-list materialization,
+untyped tensor metadata, owned snapshots and profiler perturbation. It adds
+a bounded CPU component observation; no new green service-performance cell,
+GPU-kernel gain or quality qualification is credited.

docs/skill-coverage.md

@@ -4,7 +4,7 @@ Generated from the registry, the skill sidecars and the two maintained
 matrices. **Do not edit by hand**; regenerate after any registry, skill,
 matrix or frozen-plan change.
 
-- Input digest: `sha256:ec4e4b8fbed28d0883fc54ba7bfaf6a76ee0a5f1f28774448f3b49ebed8addb9`
+- Input digest: `sha256:b96c82e08ff3892763f6a34f042f51a4613914300c7fdd4c3f774e4d9b36b349`
 - Regenerate: `python tools/build_skill_coverage.py`
 - Verify without writing: `python tools/build_skill_coverage.py --check`