dokimi subjects

usage: dokimi subjects [-h] {list,show,run} ...

positional arguments:
  {list,show,run}
    list           List subject descriptors
    show           Show normalized subject descriptor
    run            Run subject awareness matrix and write scorecard

options:
  -h, --help       show this help message and exit

dokimi subjects list

usage: dokimi subjects list [-h] [--subjects-root SUBJECTS_ROOT]

options:
  -h, --help            show this help message and exit
  --subjects-root SUBJECTS_ROOT
                        Directory holding subject descriptors (*.toml/*.json). Required; the
                        descriptors live in a separate evaluation-content checkout, e.g.
                        --subjects-root /path/to/content-repo/evaluation/subjects

dokimi subjects show

usage: dokimi subjects show [-h] subject

positional arguments:
  subject     Subject descriptor path (.toml or .json)

options:
  -h, --help  show this help message and exit

dokimi subjects run

usage: dokimi subjects run [-h] [--out-dir OUT_DIR] [--out-json OUT_JSON] [--out-md OUT_MD]
                           [--provider {openai,anthropic,amazon-bedrock,google-vertex,openrouter,openai-codex}]
                           [--context7-id CONTEXT7_ID] [--judge] [--judge-model JUDGE_MODEL]
                           [--judge-provider {openai,anthropic,google,google-vertex,openrouter,amazon-bedrock}]
                           [--top-k TOP_K] [--rerank-threshold RERANK_THRESHOLD]
                           [--max-iterations MAX_ITERATIONS] [--max-tokens MAX_TOKENS]
                           [--grounded-from-topics] [--concurrency CONCURRENCY]
                           [--no-plugin-deltas] [--novelty-artifact PATH]
                           [--work-harnesses WORK_HARNESSES] [--tasks-root TASKS_ROOT]
                           [--work-baseline-harness WORK_BASELINE_HARNESS]
                           [--work-model WORK_MODEL] [--work-output-dir WORK_OUTPUT_DIR]
                           [--work-command-template WORK_COMMAND_TEMPLATE]
                           [--work-repeats WORK_REPEATS] [--work-dry-run] [--skip-work]
                           subject

positional arguments:
  subject               Subject descriptor path (.toml or .json)

options:
  -h, --help            show this help message and exit
  --out-dir OUT_DIR
  --out-json OUT_JSON
  --out-md OUT_MD
  --provider {openai,anthropic,amazon-bedrock,google-vertex,openrouter,openai-codex}
  --context7-id CONTEXT7_ID
  --judge
  --judge-model JUDGE_MODEL
  --judge-provider {openai,anthropic,google,google-vertex,openrouter,amazon-bedrock}
  --top-k TOP_K
  --rerank-threshold RERANK_THRESHOLD
  --max-iterations MAX_ITERATIONS
  --max-tokens MAX_TOKENS
                        Output-token ceiling per answer. Raise for reasoning models, which can
                        spend the provider default entirely on reasoning and return empty text
  --grounded-from-topics
                        Let the judge use a question's expected_topics as ground truth when they
                        are substantive answer statements, enabling the factual_grounding
                        dimension for curated golden sets. Keyword-style topics are ignored
  --concurrency CONCURRENCY
  --no-plugin-deltas
  --novelty-artifact PATH
                        Attach a novelty.v1 artifact from `novelty analyze` (repeatable, one per
                        (artifact, model) pair). Reported per cell and never averaged: novelty is
                        a property of a pair, and ranking artifacts by `added` inside one model
                        correlated with the measured task delta at -0.446. All novelty checks are
                        soft and never block
  --work-harnesses WORK_HARNESSES
                        Comma-separated executable-task harnesses for suite.tasks (for example:
                        codex,docker-claude-skill:/path/to/skills-repo/skills/<name>).
  --tasks-root TASKS_ROOT
                        Directory holding the task directories named in suite.tasks, e.g.
                        <checkout>/evaluation/harbor/tasks. Without it, a suite.tasks entry has to
                        be a full path to a task directory.
  --work-baseline-harness WORK_BASELINE_HARNESS
  --work-model WORK_MODEL
  --work-output-dir WORK_OUTPUT_DIR
  --work-command-template WORK_COMMAND_TEMPLATE
  --work-repeats WORK_REPEATS
  --work-dry-run        Build the task_runs artifact without invoking executable-task agents.
  --skip-work           Skip suite.tasks even when the subject descriptor declares them.