docs: PolyKV stress benchmark section (#742)

README.md

@@ -119,6 +119,33 @@ unzip -p opencoti-llamafile-*.llamafile ggml-cuda.so | sha256sum
 Apache-2.0 (llamafile) + MIT (llama.cpp and bundled projects).
 opencoti patches and packaging are MIT.
 
+
+## PolyKV multi-session stress benchmark (package_courier)
+
+An agentic multi-session benchmark: LangGraph courier agents (tool-calling,
+10-step delivery chains, 100 packages) run against ONE engine at
+`--parallel 32`, ctx 163840, q8_0 KV, with a tps-floor scheduler that admits
+new sessions only while every live session keeps ≥ floor tok/s (PolyKV
+`/polykv/tps` SSE telemetry drives admission). Harness source:
+`opencoti/langgraph/bench/package_courier/` in the repo; reports under
+[`bench/reports/`](bench/reports/) (per-run HTML + raw stats).
+
+| model (MTP) | floor | score | wall | delivery mean | fan-out |
+|---|---|---|---|---|---|
+| Qwen3.6-27B-Omnimerge-v4 Q6_K (NextN self-spec) | 15 | 99/100 | 2171 s | 146 s | ~7 |
+| ″ | 5 | 99/100 | 2509 s | 467 s | →22 |
+| ″ | 1 | 98/100 | 2803 s | 818 s | 32 (max) |
+| Gemma-4 26B-A4B-it Q4_K_M (assistant-MTP) | 15 | 100/100 | 864 s | 51 s | med 6 |
+| ″ | 5 | 99/100 | 1090 s | 145 s | med 14 |
+| ″ | 1 | 95/100 | 1340 s | 346 s | 32 (max) |
+
+Aggregate engine throughput stays pinned (Qwen ~75-80 tok/s regardless of
+concurrency), so equilibrium workers ≈ aggregate/floor. The A4B + assistant-MTP
+pair is ~2.5× faster end-to-end at every rung. All losses across all runs are
+agent-side (turn-budget wander on hard packages) — zero tool errors, zero LLM
+errors, engines clean throughout. The `ab-pooled` vs `ab-naive` pair isolates
+the SharedKVPool win at identical workload.
+
 ---
 
 <img src="vodafone-logo.png" alt="Vodafone" width="96"/>