# Generated from run-config.json. Contains no credentials or cluster endpoints.
schema_version: 1
source: "run-config.json"
tested_configuration:
  package: "request-concurrency-priority-tuning"
  runner:
    source: "pipeline/archive/benchmark-2026-07.py"
    published_sha256: "0d5296e42922f6691db267598afe3ec5ac4f653b96a9935f5fefbf8e4c2dbce3"
    executed_sha256: "cfc227b88faa9041ba88c2e352c27fae328a97d1fce6ec0b9c6859c32d136998"
    sanitized_deployment_defaults_only: true
    traffic_driver: "native closed loop"
    historical_runner: true
  endpoint_picker:
    version: "llm-d Endpoint Picker v0.9.0"
    detector: "request-concurrency"
    max_concurrency_values:
      - 32
      - 48
      - 64
      - 96
      - 128
  traffic:
    input_tokens: 512
    output_tokens: 128
    prompt_pool_size: 384
    scenario_duration_seconds: 120
    repeats_per_value: 2
    traffic_seed: 42
    cache_mode: "off"
  scope: "Historical priority-tuning calibration. New runs use pipeline/benchmark.py."