"""search_course_content — RAG over the official course material."""

from __future__ import annotations

import re
from typing import Any

from pydantic import BaseModel, Field

from app.llm.openrouter import get_llm
from app.llm.schemas import ToolResult
from app.rag import retriever
from app.tools.registry import ToolContext, registry


def _excerpt(text: str, limit: int = 280) -> str:
    """Plain-text excerpt for the UI card: drop LaTeX delimiters/commands and headings."""
    t = re.sub(r"\$\$?(.*?)\$\$?", lambda m: re.sub(r"\\[a-zA-Z]+|[{}]", "", m.group(1)), text, flags=re.S)
    t = re.sub(r"^#+\s*", "", t, flags=re.M).replace("**", "")
    t = re.sub(r"\s+", " ", t).strip()
    return t if len(t) <= limit else t[: limit - 1] + "…"


class SearchCourseArgs(BaseModel):
    query: str = Field(..., min_length=2, max_length=400)
    course: str | None = Field(None, description="IMM1003, IMM1033 ou null pour les deux")
    top_k: int = Field(6, ge=1, le=10)


async def run(args: dict[str, Any], ctx: ToolContext) -> ToolResult:
    await ctx.report("running", "Recherche dans le matériel du cours…")
    courses: set[str] | None = None
    if args.get("course"):
        courses = {str(args["course"]).upper()}
    q_emb = None
    if ctx.settings.MODEL_EMBEDDINGS and retriever.index.embeddings is not None:
        try:
            vecs = await get_llm().embed([args["query"]])
            q_emb = vecs[0] if vecs else None
        except Exception:  # noqa: BLE001
            q_emb = None
    hits = retriever.index.search(
        args["query"], top_k=args["top_k"], courses=courses, boost_course=ctx.course_code,
        include_professor=ctx.role in {"professor", "admin"}, query_embedding=q_emb)
    sources = [{
        "index": i + 1, "id": h.chunk_id, "course": h.course, "module": h.module,
        "section": h.section, "page": h.page, "url": h.url,
        "excerpt": _excerpt(h.content), "score": round(h.score, 3),
    } for i, h in enumerate(hits)]
    return ToolResult(
        content=retriever.format_for_model(hits),
        payload={"query": args["query"], "sources": sources},
        meta={"summary": f"{len(hits)} passage(s) du cours"},
    )


registry.register("search_course_content", run, SearchCourseArgs)