XmLLM / src /app /enrichers /__init__.py
Claude
Code quality: fix all ruff warnings, add CI/CD, improve test coverage
bbbfba8 unverified
Raw
History Blame Contribute Delete
2.01 kB
"""Enricher subsystem — post-normalization enrichment pipeline.
Enrichers transform a CanonicalDocument into a new CanonicalDocument
with additional derived data. They produce new instances (models are frozen).
Every enricher MUST:
- Set provenance.evidence_type to 'derived' or 'inferred'
- Update geometry.status if geometry was modified
- Add warnings when appropriate
- Respect the active DocumentPolicy
"""
from __future__ import annotations
from abc import ABC, abstractmethod
from typing import TYPE_CHECKING
from src.app.policies.document_policy import DocumentPolicy
if TYPE_CHECKING:
from src.app.domain.models import CanonicalDocument
class BaseEnricher(ABC):
"""Abstract base for enrichers."""
@property
@abstractmethod
def name(self) -> str:
"""Enricher name used in audit trail."""
...
@abstractmethod
def enrich(
self, doc: CanonicalDocument, policy: DocumentPolicy
) -> CanonicalDocument:
"""Apply enrichment, returning a new document."""
...
class EnricherPipeline:
"""Runs a configurable chain of enrichers."""
def __init__(self, enrichers: list[BaseEnricher] | None = None) -> None:
self._enrichers = enrichers or []
def add(self, enricher: BaseEnricher) -> None:
self._enrichers.append(enricher)
def run(
self, doc: CanonicalDocument, policy: DocumentPolicy | None = None
) -> CanonicalDocument:
"""Run all enrichers in order, threading the document through."""
if policy is None:
policy = DocumentPolicy()
applied: list[str] = list(doc.audit.enrichers_applied)
current = doc
for enricher in self._enrichers:
current = enricher.enrich(current, policy)
applied.append(enricher.name)
# Update audit with enrichers applied
new_audit = doc.audit.model_copy(update={"enrichers_applied": applied})
return current.model_copy(update={"audit": new_audit})