Transforms complex documents like PDFs and Office docs into LLM-ready markdown/JSON for your Agentic workflows.
Python ai4sciencedocument-analysisdocxextract-datalayout-analysisocrparserpdfpdf-converterpdf-extractor-llmpdf-extractor-pretrainpdf-extractor-ragpdf-parserpptxpythonxlsx
"""Doclib 入库与主动解析共用的默认页范围和后台档位策略。"""
from __future__ import annotations
from ...parser.page_range import expand_page_range, parse_page_range_set
from ...types import Tier, select_default_quality_tier, select_parsing_rule_tier
from ..rows import ParseRow
from .parse_errors import ParseFailure
from ..parse_server_state import ParseServerHealth
def default_parse_range(page_count: int | None) -> str:
"""未知页数请求第一页,已知页数时选择前十页。"""
if page_count and page_count > 0:
return expand_page_range("1-10", page_count)
return "1"
def resolve_parsing_rule_default_tier(*, health: ParseServerHealth, remote: bool = False) -> Tier:
"""后台规则沿用可用质量档位降级到 Flash 的既有策略。"""
supported = health.remote.probe.tiers if remote else health.local.probe.tiers
return select_parsing_rule_tier(supported)
def page_range_covered(request_page_range: str, done_batches: list[ParseRow]) -> bool:
"""按较新完成批次优先的页范围集合判断完整覆盖。"""
needed_page_numbers = parse_page_range_set(request_page_range)
covered_page_numbers: set[int] = set()
for b in sorted(done_batches, key=lambda r: r.get("done_at") or 0, reverse=True):
covered_page_numbers |= parse_page_range_set(b["page_range"])
if needed_page_numbers <= covered_page_numbers:
return True
return False
def page_range_uncovered(request_page_range: str, done_batches: list[ParseRow]) -> set[int]:
"""返回请求中尚未被完成批次覆盖的页号。"""
needed_page_numbers = parse_page_range_set(request_page_range)
covered_page_numbers: set[int] = set()
for b in done_batches:
covered_page_numbers |= parse_page_range_set(b["page_range"])
return needed_page_numbers - covered_page_numbers
def resolve_default_tier(*, health: ParseServerHealth, remote: bool = False) -> Tier:
"""按健康信息选择默认质量档位,并保留认证及无引擎错误。"""
probe = health.remote.probe if remote else health.local.probe
supported = probe.tiers
selected = select_default_quality_tier(supported)
if selected is not None:
return selected
if remote and probe.error_code == "invalid_api_key":
raise ParseFailure(
"invalid_api_key",
probe_error_message("Remote parse-server authentication failed", probe.error_msg),
"parse_server.remote.api_key",
)
if not remote and health.local_mode == "self_hosted" and probe.error_code == "invalid_api_key":
raise ParseFailure(
"invalid_api_key",
probe_error_message("Local self-hosted parse-server authentication failed", probe.error_msg),
"parse_server.local.self_hosted_api_key",
)
actions = ["start a local parse-server"]
if not remote and health.remote.probe.healthy and select_default_quality_tier(health.remote.probe.tiers) is not None:
actions.append("use --remote")
actions.append("explicitly pass --tier flash for text-only preview")
raise ParseFailure(
"quality_tier_unavailable",
f"No basic, standard, or advanced engine available. You can {_format_action_list(actions)}.",
)
def _format_action_list(actions: list[str]) -> str:
"""按原有语法连接错误提示中的操作选项。"""
if len(actions) == 1:
return actions[0]
if len(actions) == 2:
return f"{actions[0]} or {actions[1]}"
return f"{', '.join(actions[:-1])}, or {actions[-1]}"
def probe_error_message(prefix: str, error_msg: str | None) -> str:
"""拼接探测错误详情,保持调用方原有前缀。"""
if error_msg:
return f"{prefix}: {error_msg}"
return prefix
__all__ = [
"default_parse_range",
"resolve_parsing_rule_default_tier",
"resolve_default_tier",
"page_range_covered",
"page_range_uncovered",
"probe_error_message",
]