agentic-kvc

Go to file

Gahow Wang d8dc9dc0ce Add --max-single-turn-ratio to control single-turn session fraction

Single-turn sessions with unique prefixes get 0% cache hit, diluting
APC in benchmarks.  --max-single-turn-ratio caps their fraction,
boosting multi-turn density and theoretical APC.

Example: --sample-ratio 0.008 --max-single-turn-ratio 0.3
  Before: 9.2% multi-turn, APC=70.5%
  After:  70.0% multi-turn, APC=85.0%, sharing=53.3%

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

2026-05-23 14:17:25 +08:00

analysis

Add comprehensive research findings document

2026-05-23 07:16:31 +08:00

patches

Add vLLM patches directory for version-controlled patch management

2026-05-22 00:26:14 +08:00

replayer

Fix replay methodology: trace-driven dispatch, no artificial limits

2026-05-23 12:43:41 +08:00

scripts

Add --max-single-turn-ratio to control single-turn session fraction