#!/usr/bin/env python3
"""Smoke test: a few direct (non-batch) calls per tier to validate schemas and measure token usage.
Usage: WP20_ENV_FILE=/path/to/.env .venv/bin/python scripts/03_smoke_test.py [n_listings] [n_images]
"""
import json
import sys
from pathlib import Path

import pandas as pd

sys.path.insert(0, str(Path(__file__).resolve().parents[1]))
from src.extraction import client as C  # noqa: E402
from src.extraction.textprep import build_text  # noqa: E402
from src.feature_schema.schema import IMAGE_SCHEMA, TEXT_SCHEMA  # noqa: E402

ROOT = Path(__file__).resolve().parents[1]
n_l = int(sys.argv[1]) if len(sys.argv) > 1 else 2
n_i = int(sys.argv[2]) if len(sys.argv) > 2 else 3

smp = pd.read_parquet(ROOT / "data" / "processed" / "sample.parquet").head(n_l)
usage = []
for r in smp.itertuples():
    print(f"\n=== {r.uid} | {r.source} | {r.roll_municipalite} | built {r.roll_year_built} | {r.roll_floor_area_m2} m² | ${r.price:,.0f}")
    for tier in ["haiku", "sonnet"]:
        for i, url in enumerate(list(r.image_urls_used)[:n_i]):
            out = C.direct(C.image_request(tier, url, IMAGE_SCHEMA), tag="smoke")
            if out["ok"]:
                d = out["data"]
                print(f"  [{tier}] img{i} {d['scene']:24s} q{d['photo_quality']} ext{d['exterior_condition']} int{d['interior_condition']} "
                      f"kit={d['kitchen_grade']} clad={d['cladding']} conf={d['confidence']:.2f} | {d['evidence'][:70]} | tok {out['usage']}")
                usage.append({"tier": tier, "kind": "image", **out["usage"]})
            else:
                print(f"  [{tier}] img{i} FAILED {out.get('error')} {out.get('stop_reason')}")
    text, stats = build_text(r._asdict())
    for tier in ["haiku", "sonnet"]:
        out = C.direct(C.text_request(tier, text, TEXT_SCHEMA), tag="smoke")
        if out["ok"]:
            d = out["data"]
            print(f"  [{tier}] TEXT clad={d['cladding']} roof={d['roof_covering']} kit={d['kitchen_grade']} cond={d['condition_claim']} "
                  f"reno={d['renovation_claim']} yb={d['year_built_claim']} area={d['living_area_claim_sqft']} price_leak={d['price_or_assessment_mentioned']} | tok {out['usage']} | mask {stats}")
            usage.append({"tier": tier, "kind": "text", **out["usage"]})
        else:
            print(f"  [{tier}] TEXT FAILED {out.get('error')}")
u = pd.DataFrame(usage)
print("\nmean tokens per call:\n", u.groupby(["tier", "kind"])[["input_tokens", "output_tokens"]].mean().round(0))