import argparse, os, json
from pathlib import Path
from dotenv import dotenv_values
import pandas as pd

from app.ingest import ingest
from app.search import search


def get_env():
    """
    Load environment paths. On Hugging Face Spaces, fall back to local 'data/' dirs.
    Never use absolute paths like /Users/... which don't exist in the container.
    """
    env = dotenv_values(".env") or {}

    # Root of the repo (parent of this file's folder)
    ROOT = Path(__file__).resolve().parents[1]

    def default_path(sub):
        return str(ROOT / sub)

    env.setdefault("DATA_DIR", default_path("data"))
    env.setdefault("DOCSTORE_DIR", default_path("data/docstore"))
    env.setdefault("INDEX_DIR", default_path("data/index"))
    env.setdefault("EXPORT_DIR", default_path("data/exports"))

    for k in ["DATA_DIR", "DOCSTORE_DIR", "INDEX_DIR", "EXPORT_DIR"]:
        os.makedirs(env[k], exist_ok=True)

    return env



def ensure_index_exists(env: dict):
    """
    Ensure a FAISS index exists in env['INDEX_DIR'].
    If missing, run a minimal ingest using config/sources.yaml.
    This lets the Hugging Face Space self-heal on first boot.
    """
    index_dir = Path(env["INDEX_DIR"])
    faiss_idx = index_dir / "faiss.index"
    meta_json = index_dir / "meta.json"

    if faiss_idx.exists() and meta_json.exists():
        return  # already built

    print("Index not found. Building now via ingest() …")
    # NOTE: This uses your existing ingestion pipeline.
    # If your ingest relies on API keys, set them in the Space's
    # Settings → Variables and secrets, then Restart the Space.
    path, n = ingest("config/sources.yaml", env)
    print(f"Ingest complete. {n} records. Docstore: {path}")


def cmd_ingest(args):
    env = get_env()
    path, n = ingest("config/sources.yaml", env)
    print(f"Ingest complete. {n} records. Docstore: {path}")


def cmd_search(args):
    env = get_env()
    ensure_index_exists(env)  # <— NEW: auto-build if missing
    filters = {}
    if args.geo:
        filters["geo"] = args.geo.split(",")
    if args.categories:
        filters["categories"] = args.categories.split(",")
    res = search(args.q, env, top_k=args.k, filters=filters)
    for r in res:
        print(f"- {r['title']} [{r['source']}] ({r['geo']}) score={r['score']:.3f}")
        print(f"  {r['url']}")


def cmd_export(args):
    env = get_env()
    ensure_index_exists(env)  # <— NEW: auto-build if missing
    filters = {}
    if args.geo:
        filters["geo"] = args.geo.split(",")
    if args.categories:
        filters["categories"] = args.categories.split(",")
    res = search(args.q, env, top_k=args.k, filters=filters)
    out = Path(env["EXPORT_DIR"]) / (args.out or "results.csv")
    pd.DataFrame(res).to_csv(out, index=False)
    print(f"Exported {len(res)} rows to {out}")


if __name__ == "__main__":
    p = argparse.ArgumentParser()
    sub = p.add_subparsers(dest="cmd")

    p_ing = sub.add_parser("ingest", help="Ingest sources and build index")
    p_ing.set_defaults(func=cmd_ingest)

    p_search = sub.add_parser("search", help="Search index")
    p_search.add_argument("--q", required=True)
    p_search.add_argument("--k", type=int, default=15)
    p_search.add_argument("--geo", default="")
    p_search.add_argument("--categories", default="")
    p_search.set_defaults(func=cmd_search)

    p_export = sub.add_parser("export", help="Export search results to CSV")
    p_export.add_argument("--q", required=True)
    p_export.add_argument("--k", type=int, default=50)
    p_export.add_argument("--geo", default="")
    p_export.add_argument("--categories", default="")
    p_export.add_argument("--out", default="results.csv")
    p_export.set_defaults(func=cmd_export)

    args = p.parse_args()
    if not args.cmd:
        p.print_help()
    else:
        args.func(args)