"""Normalize and compile DeerFlow-owned Wiki generation guidance.

The values accepted by the public API are raw user configuration.  They are
kept separately from the effective strings sent to WeKnora so editing a focus
or directory constraint never appends a second copy of an older compiled
block.
"""

from __future__ import annotations

from dataclasses import dataclass

WEKNORA_CUSTOM_INSTRUCTION_MAX = 4000
WIKI_FOCUS_INSTRUCTION_MAX = 2000
WIKI_CONFIG_SCHEMA_VERSION = 3


@dataclass(frozen=True, slots=True)
class CompiledWikiInstructions:
    content_instructions: str
    extraction_instructions: str
    directory_prompt_coverage: str
    directory_prompt_covered_nodes: int
    directory_prompt_total_nodes: int


def normalize_focus_instructions(value: str | None) -> str:
    """Return the canonical raw focus text used for persistence and hashing."""

    if value is None:
        return ""
    normalized = str(value).replace("\r\n", "\n").replace("\r", "\n").strip()
    if len(normalized) > WIKI_FOCUS_INSTRUCTION_MAX:
        raise ValueError(
            f"重点要求不能超过 {WIKI_FOCUS_INSTRUCTION_MAX} 个字符"
        )
    return normalized


def _join_blocks(*blocks: str) -> str:
    return "\n\n".join(block.strip() for block in blocks if block.strip())


def _focus_content_block(focus: str) -> str:
    if not focus:
        return ""
    return _join_blocks(
        f"【用户重点要求】\n{focus}",
        (
            "这是生成范围的硬约束：只生成与上述重点直接相关且有来源支撑的页面；"
            "不要为背景材料中与重点无关的人物、机构、地点或主题单独生成页面。"
            "不得为了满足重点要求而编造信息。"
        ),
    )


def _focus_extraction_block(focus: str) -> str:
    if not focus:
        return ""
    return _join_blocks(
        f"【用户重点要求】\n{focus}",
        (
            "这是抽取范围的硬约束：只抽取与上述重点直接相关、可追溯到来源的"
            "实体、事件、时间线、指标和关系；与重点无关的内容必须跳过。"
        ),
    )


def _directory_lines(directory_paths: list[list[str]]) -> list[str]:
    return [f"{'  ' * (len(path) - 1)}- {path[-1]}" for path in directory_paths]


def compile_wiki_instructions(
    *,
    content_instructions: str | None,
    extraction_instructions: str | None,
    focus_instructions: str | None,
    directory_paths: list[list[str]],
    max_length: int = WEKNORA_CUSTOM_INSTRUCTION_MAX,
) -> CompiledWikiInstructions:
    """Compile raw inputs into the two instruction strings supported upstream.

    Raw free text and focus guidance are never truncated. Directory nodes are
    appended in deterministic parent-before-child order until the remaining
    content budget is exhausted. They are examples for the generator rather
    than a post-generation allowlist: WeKnora may adapt the hierarchy when the
    source material needs a clearer organization.
    """

    raw_content = str(content_instructions or "").strip()
    raw_extraction = str(extraction_instructions or "").strip()
    focus = normalize_focus_instructions(focus_instructions)

    content_base = _join_blocks(raw_content, _focus_content_block(focus))
    extraction = _join_blocks(raw_extraction, _focus_extraction_block(focus))
    if len(content_base) > max_length:
        raise ValueError(
            f"内容生成要求与重点要求合并后不能超过 {max_length} 个字符"
        )
    if len(extraction) > max_length:
        raise ValueError(
            f"信息抽取要求与重点要求合并后不能超过 {max_length} 个字符"
        )

    total_nodes = len(directory_paths)
    covered_nodes = 0
    content = content_base
    if total_nodes:
        header = "【参考目录层级】"
        footer = (
            "上述目录仅用于帮助理解期望的组织方式，不是必须逐项套用的目录模板，也不是目录"
            "白名单。请结合原始资料自行合并、拆分、改名或补充目录；内容结构与参考目录冲突时，"
            "以资料实际内容和清晰表达为准。目录只用于组织，不得作为事实来源。"
        )
        accepted: list[str] = []
        for line in _directory_lines(directory_paths):
            candidate_block = "\n".join([header, *accepted, line, "", footer])
            candidate = _join_blocks(content_base, candidate_block)
            if len(candidate) > max_length:
                break
            accepted.append(line)
        if accepted:
            content = _join_blocks(
                content_base,
                "\n".join([header, *accepted, "", footer]),
            )
            covered_nodes = len(accepted)

    coverage = (
        "none"
        if total_nodes == 0 or covered_nodes == 0
        else "full"
        if covered_nodes == total_nodes
        else "partial"
    )
    return CompiledWikiInstructions(
        content_instructions=content,
        extraction_instructions=extraction,
        directory_prompt_coverage=coverage,
        directory_prompt_covered_nodes=covered_nodes,
        directory_prompt_total_nodes=total_nodes,
    )


__all__ = [
    "CompiledWikiInstructions",
    "WEKNORA_CUSTOM_INSTRUCTION_MAX",
    "WIKI_CONFIG_SCHEMA_VERSION",
    "WIKI_FOCUS_INSTRUCTION_MAX",
    "compile_wiki_instructions",
    "normalize_focus_instructions",
]