Executor

Rule execution engine. Rules run deterministically at inference time with no LLM calls.

When several rules produce overlapping or conflicting matches, the executor orders them by priority, then validated precision (falling back to confidence when a rule has no validated estimate), so the higher-trust rule wins. Each rule also runs under a wall-clock budget (RULE_TIMEOUT_S): a catastrophically-backtracking regex is bounded via the regex module's matching timeout, and a runaway code rule is interrupted with a signal alarm, so a single bad LLM-written rule cannot freeze evaluation.

RuleExecutor

RuleExecutor(use_spacy_ner=False)

Executes rules against input data

Initialize the rule executor.

Parameters:

Name	Type	Description	Default
use_spacy_ner	bool	If True, run spaCy's NER pipeline during rule execution so that ENT_TYPE/ENT_ID patterns are available.	False

apply_rules(rules, input_data, task_type=None, text_field=None)

Apply rules to input and return aggregated output.

Rules are sorted by priority and applied sequentially. Results are deduplicated by span position. If a rule has no output_key, it is inferred from task_type using DEFAULT_OUTPUT_KEYS.

Parameters:

Name	Type	Description	Default
rules	list[Rule]	List of rules to apply.	required
input_data	dict	Input data dict.	required
task_type	TaskType | None	Task type for inferring the default output_key.	None
text_field	str | None	Input key to use for regex/spaCy matching.	None

Returns:

Type	Description
dict	Output dict with results keyed by output_key (e.g. 'entities',
dict	'spans', 'label'). Empty dict if no rules matched.

execute_rule(rule, input_data, text_field=None)

Execute a single rule against input data.

Parameters:

Name	Type	Description	Default
rule	Rule	The rule to execute.	required
input_data	dict	Input data dict.	required
text_field	str | None	Input key to use for regex/spaCy matching.	None

Returns:

Type	Description
Any	List of Span or dict results for regex/spaCy rules, or
Any	arbitrary return value from code rules. Empty list on no match.

Functions

substitute_template

substitute_template(template, match_text, start, end, groups=(), ent_type=None, ent_label=None, token_spans=None)

Substitute template variables with actual values from a match.

Parameters:

Name	Type	Description	Default
template	dict[str, Any]	Output template dict with variable placeholders.	required
match_text	str	The full matched text ($0).	required
start	int	Start character offset ($start).	required
end	int	End character offset ($end).	required
groups	tuple	Regex capture groups ($1, $2, ...).	()
ent_type	str | None	spaCy entity type string ($ent_type).	None
ent_label	str | None	spaCy entity label string ($ent_label).	None
token_spans	list[dict[str, Any]] | None	Per-token span dicts for spaCy dependency matches ($1.text, $1.start, $1.end, etc.).	None

Returns:

Type	Description
dict[str, Any]	Dict with all template variables replaced by their actual values.

Variables: - $0: Full match text - $1, $2, ...: Capture groups - $start: Start character offset - $end: End character offset - $ent_type: Entity type (spaCy only) - $ent_label: Entity label (spaCy only)