CLI reference¶
Generated from the public argparse tree.
mudidi¶
usage: mudidi [-h] {run,auth,benchmark,config,web} ...
Dictionary OCR and MDF extraction (inference and benchmark modes).
positional arguments:
{run,auth,benchmark,config,web}
run Run production inference.
auth Manage local subscription authentication.
benchmark Benchmark workflows.
config Configuration utilities.
web Run the local production website.
options:
-h, --help show this help message and exit
mudidi run¶
usage: mudidi run [-h] [--config CONFIG] [--auth-mode {api_key,subscription}]
[--auth-provider {openai,google,claude}] [--pages PAGES]
[--dict-pages DICT_PAGES] [--intro INTRO]
[--intro-pages INTRO_PAGES] [--alphabet ALPHABET]
[--ocr-text OCR_TEXT] [--mdf-manual | --no-mdf-manual]
[--stage-1-guides STAGE1_GUIDES_PATH]
[--stage-1-guides-pages STAGE1_GUIDES_PAGES]
[--stage-2-guides STAGE2_GUIDES_PATH]
[--stage-2-guides-pages STAGE2_GUIDES_PAGES]
[--stage-2-guides-scope {pass1,pass2,both}]
[--output-dir OUTPUT_DIR]
[--stage {1,2,all,2-pass-1,2-pass-2}]
[--stage1-typography | --no-stage1-typography]
[--model MODEL] [--stage-1-model STAGE_1_MODEL]
[--stage-2-pass-1-model STAGE_2_PASS_1_MODEL]
[--stage-2-pass-2-model STAGE_2_PASS_2_MODEL] [--overwrite]
[--dry-run] [--stage1-agentic | --no-stage1-agentic]
[--stage2-agentic | --no-stage2-agentic]
[--agentic-max-iterations AGENTIC_MAX_ITERATIONS]
[--agentic-evaluator-model AGENTIC_EVALUATOR_MODEL]
[--agentic-rewriter-model AGENTIC_REWRITER_MODEL]
[--agentic-reasoning {none,minimal,low,medium,high,xhigh,max}]
[--agentic-evaluator-reasoning {none,minimal,low,medium,high,xhigh,max}]
[--agentic-rewriter-reasoning {none,minimal,low,medium,high,xhigh,max}]
[--agentic-min-retry-confidence AGENTIC_MIN_RETRY_CONFIDENCE]
[--agentic-verifier-patches | --no-agentic-verifier-patches]
[--agentic-concrete-retry-gate | --no-agentic-concrete-retry-gate]
options:
-h, --help show this help message and exit
--config CONFIG
--auth-mode {api_key,subscription}
--auth-provider {openai,google,claude}, --provider {openai,google,claude}
Subscription provider; requires --auth-mode
subscription.
--pages PAGES
--dict-pages DICT_PAGES
--intro INTRO
--intro-pages INTRO_PAGES
--alphabet ALPHABET
--ocr-text OCR_TEXT
--mdf-manual, --no-mdf-manual
Inject the bundled SIL MDF manual text into the Stage
2 Pass 1 system prompt.
--stage-1-guides STAGE1_GUIDES_PATH
Stage 1 instruction guide path.
--stage-1-guides-pages STAGE1_GUIDES_PAGES
Selected pages when --stage-1-guides is a PDF.
--stage-2-guides STAGE2_GUIDES_PATH
Stage 2 instruction guide path.
--stage-2-guides-pages STAGE2_GUIDES_PAGES
Selected pages when --stage-2-guides is a PDF.
--stage-2-guides-scope {pass1,pass2,both}
Stage 2 guide routing scope.
--output-dir OUTPUT_DIR
--stage {1,2,all,2-pass-1,2-pass-2}
--stage1-typography, --no-stage1-typography
Mark confident bold and italic text with <b>/<i> tags
in Stage 1 (default: on). Use --no-stage1-typography
for plain text.
--model MODEL
--stage-1-model STAGE_1_MODEL
--stage-2-pass-1-model STAGE_2_PASS_1_MODEL
--stage-2-pass-2-model STAGE_2_PASS_2_MODEL
--overwrite
--dry-run
agentic Evaluator-Editor options:
--stage1-agentic, --no-stage1-agentic
Enable or disable bounded Stage 1 verification and
rewriting.
--stage2-agentic, --no-stage2-agentic
Enable or disable bounded Stage 2 verification and
rewriting.
--agentic-max-iterations AGENTIC_MAX_ITERATIONS
Maximum Evaluator-to-Editor rounds for each enabled
agentic stage.
--agentic-evaluator-model AGENTIC_EVALUATOR_MODEL
Model used for verifier calls; defaults to the current
stage model.
--agentic-rewriter-model AGENTIC_REWRITER_MODEL
Model used for correction calls; defaults to the
current stage model.
--agentic-reasoning {none,minimal,low,medium,high,xhigh,max}
Shared reasoning effort for Evaluator and Editor
calls.
--agentic-evaluator-reasoning {none,minimal,low,medium,high,xhigh,max}
Verifier reasoning effort; overrides --agentic-
reasoning.
--agentic-rewriter-reasoning {none,minimal,low,medium,high,xhigh,max}
Editor reasoning effort; overrides --agentic-
reasoning.
--agentic-min-retry-confidence AGENTIC_MIN_RETRY_CONFIDENCE
Minimum verifier confidence required before a rewrite.
--agentic-verifier-patches, --no-agentic-verifier-patches
Deprecated and ignored: the correction model applies
every edit itself.
--agentic-concrete-retry-gate, --no-agentic-concrete-retry-gate
Deprecated and ignored: targeted edits always name the
text to change.
mudidi auth¶
usage: mudidi auth [-h] {status,login,logout} ...
Manage credentials in MUDIDI's local encrypted subscription store. No API keys
or tokens are accepted by these commands.
positional arguments:
{status,login,logout}
status Show safe status for one subscription provider.
login Open one provider's subscription login in the browser.
logout Delete one provider's local subscription credential.
options:
-h, --help show this help message and exit
mudidi auth status¶
usage: mudidi auth status [-h] [--provider {openai,google,claude}]
[{openai,google,claude}]
positional arguments:
{openai,google,claude}
Subscription provider (openai, google, or claude).
options:
-h, --help show this help message and exit
--provider {openai,google,claude}
Subscription provider (also accepted positionally).
mudidi auth login¶
usage: mudidi auth login [-h] [--provider {openai,google,claude}]
[{openai,google,claude}]
positional arguments:
{openai,google,claude}
Subscription provider (openai, google, or claude).
options:
-h, --help show this help message and exit
--provider {openai,google,claude}
Subscription provider (also accepted positionally).
mudidi auth logout¶
usage: mudidi auth logout [-h] [--provider {openai,google,claude}]
[{openai,google,claude}]
positional arguments:
{openai,google,claude}
Subscription provider (openai, google, or claude).
options:
-h, --help show this help message and exit
--provider {openai,google,claude}
Subscription provider (also accepted positionally).
mudidi benchmark¶
usage: mudidi benchmark [-h] {run,sweep,evaluate} ...
positional arguments:
{run,sweep,evaluate}
run Run benchmark extraction.
sweep Run a typed benchmark experiment sweep.
evaluate Evaluate predictions.
options:
-h, --help show this help message and exit
mudidi benchmark run¶
usage: mudidi benchmark run [-h] [--config CONFIG]
[--auth-mode {api_key,subscription}]
[--auth-provider {openai,google,claude}]
[--pages PAGES] [--dict-pages DICT_PAGES]
[--intro INTRO] [--intro-pages INTRO_PAGES]
[--alphabet ALPHABET] [--ocr-text OCR_TEXT]
[--dictionary-languages DICTIONARY_LANGUAGES]
[--mdf-manual | --no-mdf-manual]
[--stage-1-guides STAGE1_GUIDES_PATH]
[--stage-1-guides-pages STAGE1_GUIDES_PAGES]
[--stage-2-guides STAGE2_GUIDES_PATH]
[--stage-2-guides-pages STAGE2_GUIDES_PAGES]
[--stage-2-guides-scope {pass1,pass2,both}]
[--output-dir OUTPUT_DIR]
[--stage {1,2,all,2-pass-1,2-pass-2}]
[--stage1-typography | --no-stage1-typography]
[--model MODEL] [--stage-1-model STAGE_1_MODEL]
[--stage-2-pass-1-model STAGE_2_PASS_1_MODEL]
[--stage-2-pass-2-model STAGE_2_PASS_2_MODEL]
[--overwrite] [--dry-run]
[--stage1-agentic | --no-stage1-agentic]
[--stage2-agentic | --no-stage2-agentic]
[--agentic-max-iterations AGENTIC_MAX_ITERATIONS]
[--agentic-evaluator-model AGENTIC_EVALUATOR_MODEL]
[--agentic-rewriter-model AGENTIC_REWRITER_MODEL]
[--agentic-reasoning {none,minimal,low,medium,high,xhigh,max}]
[--agentic-evaluator-reasoning {none,minimal,low,medium,high,xhigh,max}]
[--agentic-rewriter-reasoning {none,minimal,low,medium,high,xhigh,max}]
[--agentic-min-retry-confidence AGENTIC_MIN_RETRY_CONFIDENCE]
[--agentic-verifier-patches | --no-agentic-verifier-patches]
[--agentic-concrete-retry-gate | --no-agentic-concrete-retry-gate]
[--dataset-dir DATASET_DIR]
[--samples-dir SAMPLES_DIR]
[--languages LANGUAGES [LANGUAGES ...]]
[--experiment-name EXPERIMENT_NAME]
options:
-h, --help show this help message and exit
--config CONFIG
--auth-mode {api_key,subscription}
--auth-provider {openai,google,claude}, --provider {openai,google,claude}
Subscription provider; requires --auth-mode
subscription.
--pages PAGES
--dict-pages DICT_PAGES
--intro INTRO
--intro-pages INTRO_PAGES
--alphabet ALPHABET
--ocr-text OCR_TEXT
--dictionary-languages DICTIONARY_LANGUAGES
Legacy benchmark language metadata file.
--mdf-manual, --no-mdf-manual
Inject the bundled SIL MDF manual text into the Stage
2 Pass 1 system prompt.
--stage-1-guides STAGE1_GUIDES_PATH
Stage 1 instruction guide path.
--stage-1-guides-pages STAGE1_GUIDES_PAGES
Selected pages when --stage-1-guides is a PDF.
--stage-2-guides STAGE2_GUIDES_PATH
Stage 2 instruction guide path.
--stage-2-guides-pages STAGE2_GUIDES_PAGES
Selected pages when --stage-2-guides is a PDF.
--stage-2-guides-scope {pass1,pass2,both}
Stage 2 guide routing scope.
--output-dir OUTPUT_DIR
--stage {1,2,all,2-pass-1,2-pass-2}
--stage1-typography, --no-stage1-typography
Mark confident bold and italic text with <b>/<i> tags
in Stage 1 (default: on). Use --no-stage1-typography
for plain text.
--model MODEL
--stage-1-model STAGE_1_MODEL
--stage-2-pass-1-model STAGE_2_PASS_1_MODEL
--stage-2-pass-2-model STAGE_2_PASS_2_MODEL
--overwrite
--dry-run
--dataset-dir DATASET_DIR
--samples-dir SAMPLES_DIR
--languages LANGUAGES [LANGUAGES ...]
--experiment-name EXPERIMENT_NAME
agentic Evaluator-Editor options:
--stage1-agentic, --no-stage1-agentic
Enable or disable bounded Stage 1 verification and
rewriting.
--stage2-agentic, --no-stage2-agentic
Enable or disable bounded Stage 2 verification and
rewriting.
--agentic-max-iterations AGENTIC_MAX_ITERATIONS
Maximum Evaluator-to-Editor rounds for each enabled
agentic stage.
--agentic-evaluator-model AGENTIC_EVALUATOR_MODEL
Model used for verifier calls; defaults to the current
stage model.
--agentic-rewriter-model AGENTIC_REWRITER_MODEL
Model used for correction calls; defaults to the
current stage model.
--agentic-reasoning {none,minimal,low,medium,high,xhigh,max}
Shared reasoning effort for Evaluator and Editor
calls.
--agentic-evaluator-reasoning {none,minimal,low,medium,high,xhigh,max}
Verifier reasoning effort; overrides --agentic-
reasoning.
--agentic-rewriter-reasoning {none,minimal,low,medium,high,xhigh,max}
Editor reasoning effort; overrides --agentic-
reasoning.
--agentic-min-retry-confidence AGENTIC_MIN_RETRY_CONFIDENCE
Minimum verifier confidence required before a rewrite.
--agentic-verifier-patches, --no-agentic-verifier-patches
Deprecated and ignored: the correction model applies
every edit itself.
--agentic-concrete-retry-gate, --no-agentic-concrete-retry-gate
Deprecated and ignored: targeted edits always name the
text to change.
mudidi benchmark sweep¶
usage: mudidi benchmark sweep [-h] --config CONFIG [--experiment EXPERIMENT]
[--select SELECT] [--max-runs MAX_RUNS]
[--dry-run]
options:
-h, --help show this help message and exit
--config CONFIG
--experiment EXPERIMENT
--select SELECT
--max-runs MAX_RUNS
--dry-run
mudidi benchmark evaluate¶
usage: mudidi benchmark evaluate [-h] {stage1,stage2} ...
positional arguments:
{stage1,stage2}
options:
-h, --help show this help message and exit
mudidi benchmark evaluate stage1¶
usage: mudidi benchmark evaluate stage1 [-h] [--config CONFIG]
[--predicted PREDICTED] [--gold GOLD]
[--dataset-dir DATASET_DIR]
[--pred-root PRED_ROOT]
[--samples-dir SAMPLES_DIR]
[--output-dir OUTPUT_DIR]
[--languages LANGUAGES [LANGUAGES ...]]
[--experiment-name EXPERIMENT_NAME]
[--all-experiments]
[--experiment-name-contains EXPERIMENT_NAME_CONTAINS]
[--include-vlm-ocr]
[--stage1-output-subdir STAGE1_OUTPUT_SUBDIR]
[--metrics {full,minimal}]
[--alignment-threshold ALIGNMENT_THRESHOLD]
[--character-alignment {collapsed,quick_match}]
[--per-language-script] [--overwrite]
[--workers WORKERS]
options:
-h, --help show this help message and exit
--config CONFIG
--predicted PREDICTED, -p PREDICTED
--gold GOLD, -g GOLD
--dataset-dir DATASET_DIR
--pred-root PRED_ROOT
--samples-dir SAMPLES_DIR
--output-dir OUTPUT_DIR, -o OUTPUT_DIR
--languages LANGUAGES [LANGUAGES ...]
--experiment-name EXPERIMENT_NAME
--all-experiments
--experiment-name-contains EXPERIMENT_NAME_CONTAINS
--include-vlm-ocr
--stage1-output-subdir STAGE1_OUTPUT_SUBDIR
--metrics {full,minimal}
--alignment-threshold ALIGNMENT_THRESHOLD
--character-alignment {collapsed,quick_match}
--per-language-script
--overwrite
--workers WORKERS
mudidi benchmark evaluate stage2¶
usage: mudidi benchmark evaluate stage2 [-h] [--config CONFIG]
[--predicted PREDICTED] [--gold GOLD]
[--dataset-dir DATASET_DIR]
[--pred-root PRED_ROOT]
[--samples-dir SAMPLES_DIR]
[--output-dir OUTPUT_DIR]
[--languages LANGUAGES [LANGUAGES ...]]
[--experiment-name EXPERIMENT_NAME]
[--all-experiments]
[--baseline-summary BASELINE_SUMMARY]
[--baseline-experiment BASELINE_EXPERIMENT]
[--comparison-output COMPARISON_OUTPUT]
[--record-threshold RECORD_THRESHOLD]
[--line-threshold LINE_THRESHOLD]
[--marker-sub-list MARKER_SUB_LIST]
[--dictionary-languages DICTIONARY_LANGUAGES]
options:
-h, --help show this help message and exit
--config CONFIG
--predicted PREDICTED, -p PREDICTED
--gold GOLD, -g GOLD
--dataset-dir DATASET_DIR
--pred-root PRED_ROOT
--samples-dir SAMPLES_DIR
--output-dir OUTPUT_DIR, -o OUTPUT_DIR
--languages LANGUAGES [LANGUAGES ...]
--experiment-name EXPERIMENT_NAME
--all-experiments
--baseline-summary BASELINE_SUMMARY
--baseline-experiment BASELINE_EXPERIMENT
--comparison-output COMPARISON_OUTPUT
--record-threshold RECORD_THRESHOLD
--line-threshold LINE_THRESHOLD
--marker-sub-list MARKER_SUB_LIST
--dictionary-languages DICTIONARY_LANGUAGES
mudidi config¶
usage: mudidi config [-h] {validate} ...
positional arguments:
{validate}
validate Validate a YAML config.
options:
-h, --help show this help message and exit
mudidi config validate¶
usage: mudidi config validate [-h] config
positional arguments:
config
options:
-h, --help show this help message and exit
mudidi web¶
usage: mudidi web [-h] [--host {127.0.0.1,localhost}] [--port PORT]
[--data-dir DATA_DIR]
[--max-request-bytes MAX_REQUEST_BYTES]
[--max-upload-bytes MAX_UPLOAD_BYTES] [--container]
[--no-browser]
options:
-h, --help show this help message and exit
--host {127.0.0.1,localhost}
Loopback interface to bind (default: 127.0.0.1).
--port PORT
--data-dir DATA_DIR
--max-request-bytes MAX_REQUEST_BYTES
Optional cap on the raw HTTP request body size,
including multipart framing. Unlimited by default.
--max-upload-bytes MAX_UPLOAD_BYTES
Optional cap on the cumulative managed upload size per
run. Unlimited by default.
--container Bind to the container network interface. Use only
inside a container whose published port is restricted
to host loopback.
--no-browser Do not open the website in the default browser.