backend/services/reference_data_service.py

from __future__ import annotations

import re
import shutil
from dataclasses import dataclass
from pathlib import Path

from sqlalchemy import select

from core.db import SessionLocal
from core.models import Candidate
from services.ingestion_service import IngestionService

DEFAULT_REFERENCE_RESUMES_DIR = (
    Path(__file__).resolve().parents[1]
    / "data"
    / "reference_resumes"
)

FALLBACK_REFERENCE_RESUME_DIRS = [
    Path("E:/SaturnIQ/engazeiq-adk")
    / "multiagent_platform_backend"
    / "services"
    / "agent_services"
    / "mcp_tools"
    / "organized_docs"
    / "resumes",
    Path("E:/SaturnIQ/engazeiq-adk")
    / "multiagent_platform_backend"
    / "services"
    / "agent_services"
    / "mcp_tools"
    / "organized_docs"
    / "resumes_",
]


@dataclass
class ReferenceIngestionSummary:
    source_dir: str
    requested_limit: int
    files_seen: int
    ingested: int
    skipped: int
    failed: int
    failures: list[dict[str, str]]


def _resolve_source_dir(source_dir: str | None) -> Path:
    if source_dir:
        explicit = Path(source_dir)
        if explicit.exists() and explicit.is_dir():
            return explicit
        raise FileNotFoundError(f"Reference resume directory not found: {explicit}")

    candidate_dirs = [DEFAULT_REFERENCE_RESUMES_DIR, *FALLBACK_REFERENCE_RESUME_DIRS]
    for directory in candidate_dirs:
        if directory.exists() and directory.is_dir() and any(directory.rglob("*.pdf")):
            return directory

    checked = "\n".join(str(path) for path in candidate_dirs)
    raise FileNotFoundError(f"No usable reference resume directory found. Checked:\n{checked}")


def _normalize_external_id(file_path: Path) -> str:
    normalized = re.sub(r"[^a-zA-Z0-9]+", "_", file_path.stem).strip("_")
    if not normalized:
        normalized = "resume"
    return f"ref_{normalized.lower()}"


def sync_reference_resumes_to_local(
    *,
    max_files: int = 120,
    source_dir: str | None = None,
) -> dict[str, object]:
    resolved_source = _resolve_source_dir(source_dir)
    DEFAULT_REFERENCE_RESUMES_DIR.mkdir(parents=True, exist_ok=True)

    copied = 0
    skipped = 0
    seen_names: set[str] = set()

    pdf_files = sorted(resolved_source.rglob("*.pdf"))[: max(1, max_files)]
    for idx, file_path in enumerate(pdf_files, start=1):
        base_name = file_path.name
        target_name = base_name

        same_name_target = DEFAULT_REFERENCE_RESUMES_DIR / base_name
        if same_name_target.exists() and same_name_target.stat().st_size == file_path.stat().st_size:
            skipped += 1
            seen_names.add(base_name.lower())
            continue

        if target_name.lower() in seen_names or (DEFAULT_REFERENCE_RESUMES_DIR / target_name).exists():
            target_name = f"{idx:04d}_{base_name}"

        seen_names.add(target_name.lower())
        target_path = DEFAULT_REFERENCE_RESUMES_DIR / target_name

        if target_path.exists():
            skipped += 1
            continue

        shutil.copy2(file_path, target_path)
        copied += 1

    return {
        "source_dir": str(resolved_source),
        "target_dir": str(DEFAULT_REFERENCE_RESUMES_DIR),
        "files_seen": len(pdf_files),
        "copied": copied,
        "skipped": skipped,
    }


def ingest_reference_resumes(
    *,
    limit: int = 25,
    source_dir: str | None = None,
    force_reingest: bool = False,
) -> ReferenceIngestionSummary:
    base_dir = _resolve_source_dir(source_dir)

    service = IngestionService()
    pdf_files = sorted(base_dir.rglob("*.pdf"))[: max(1, limit)]

    failures: list[dict[str, str]] = []
    ingested = 0
    skipped = 0

    with SessionLocal() as db:
        for pdf in pdf_files:
            external_id = _normalize_external_id(pdf)

            if not force_reingest:
                exists = db.execute(
                    select(Candidate.id).where(Candidate.external_id == external_id)
                ).scalar_one_or_none()
                if exists is not None:
                    skipped += 1
                    continue

            try:
                file_bytes = pdf.read_bytes()
                service.ingest_pdf(
                    db,
                    file_name=pdf.name,
                    file_bytes=file_bytes,
                    external_id=external_id,
                )
                ingested += 1
            except Exception as exc:
                failures.append({"file": str(pdf), "error": str(exc)})

    return ReferenceIngestionSummary(
        source_dir=str(base_dir),
        requested_limit=max(1, limit),
        files_seen=len(pdf_files),
        ingested=ingested,
        skipped=skipped,
        failed=len(failures),
        failures=failures,
    )