StormShadow308's picture
Deploy RICS v2 backend (CPU embedder + reranker, baked jina models)
a671976
Raw
History Blame Contribute Delete
2.06 kB
"""Result models for a generated report."""
from __future__ import annotations
from pydantic import BaseModel, Field
class ReferenceSource(BaseModel):
"""Provenance for content mapped from an uploaded past report (REFERENCE tier)."""
report_filename: str
section_id: str = ""
section_title: str = ""
paragraph_index: int = 0
tier: str = Field(
default="reference", description="Always reference for user-facing attribution."
)
class GeneratedSection(BaseModel):
"""One section of mapped output."""
section_id: str
title: str
text: str = Field(default="")
rating_value: str | None = Field(default=None)
status: str = Field(
default="OK",
description="OK | empty | NO_RAG_MATCH | GROUNDING_REVIEW | UNASSIGNED",
)
notes: str = Field(default="", description="Diagnostic note for this section.")
rag_sources: list[str] = Field(
default_factory=list,
description="Human-readable REFERENCE provenance strings.",
)
reference_sources: list[ReferenceSource] = Field(
default_factory=list,
description="Structured provenance from uploaded past reports only.",
)
grounding_passed: bool = Field(default=True)
unmatched_observations: list[str] = Field(default_factory=list)
shorthand_expanded: str | None = Field(
default=None,
description="Optional expanded shorthand for this section's notes.",
)
class ReportResult(BaseModel):
"""Full generation result before DOCX assembly."""
tenant_id: str
schema_version: int
property_type: str = Field(default="")
tenure: str = Field(default="")
sections: list[GeneratedSection] = Field(default_factory=list)
unassigned_text: str = Field(default="")
active_section_count: int = Field(
default=0,
description="Leaf subsections with notes and/or AI-selected photos.",
)
processed_section_count: int = Field(
default=0,
description="Leaf subsections actually processed in this run.",
)