# pages/99_Test_Results.py

import os
import time
from pathlib import Path

import pandas as pd
import streamlit as st

from src import ai_core

TEST_OUT_DIR = Path("/data/econsult/tests")
TEST_OUT_DIR.mkdir(parents=True, exist_ok=True)
RESULTS_CSV = TEST_OUT_DIR / "results.csv"
RUN_LOG = TEST_OUT_DIR / "run_logs.txt"


def _write_run_log(text: str):
    with RUN_LOG.open("a", encoding="utf-8") as f:
        f.write(text.rstrip() + "\n")


def _cases():
    """Two stable cases for hybrid mapping validation."""
    lipids = {
        "age": "58",
        "sex": "F",
        "chief_complaint": "Statin intolerance and LDL management",
        "history": (
            "Hx of type 2 diabetes (A1c 7.6), prior NSTEMI 2019, BMI 31. "
            "Reports myalgias on high-intensity statin; stopped atorvastatin 80 mg after 3 weeks. "
            "No rhabdo. Hypothyroidism on levothyroxine. Smokes <5 cigs/day."
        ),
        "medications": "levothyroxine 100 mcg daily; ezetimibe 10 mg daily; no current statin.",
        "allergies": "no known drug allergies",
        "labs": "LDL 162 mg/dL; HDL 41; TG 210; AST/ALT normal; TSH 2.1",
        "question": "Evidence-based plan for LDL lowering in ASCVD with prior statin myalgias.",
    }

    ckd_dose = {
        "age": "72",
        "sex": "M",
        "chief_complaint": "Medication dosing question in CKD",
        "history": (
            "Stage 4 CKD likely diabetic nephropathy; eGFR 22 ml/min. "
            "HTN, CAD s/p PCI (2021). Recent cellulitis requiring antibiotics."
        ),
        "medications": "metoprolol succinate 50 mg daily; lisinopril 20 mg daily; aspirin 81 mg daily; rosuvastatin 10 mg nightly.",
        "allergies": "penicillin (rash)",
        "labs": "Cr 2.8 mg/dL; eGFR 22; K 5.0; A1c 8.1; LFTs WNL",
        "question": "Recommend renal-adjusted antibiotic options and statin dosing safety.",
    }

    return [("lipids", lipids), ("ckd_dose", ckd_dose)]


def main():
    st.title("Step 99 — Hybrid Mapping Validation")

    # Pin validation mode
    os.environ["MAP_MODE"] = "validation"
    st.info("MAP_MODE set to **validation** (registry cap = 20).")

    max_new_tokens = st.number_input("Max new tokens", min_value=300, max_value=1200, value=700, step=50)
    run_btn = st.button("Run validation")

    if not run_btn:
        st.stop()

    results = []
    for case_id, intake in _cases():
        _write_run_log("=== Hybrid Mapping Validation Run ===")
        _write_run_log(f"--- Running case: {case_id} ---")
        st.write(f"Running case: `{case_id}` ...")

        t0 = time.time()
        out = ai_core.generate_soap_draft(
            intake,
            mode="mapping",
            max_new_tokens=int(max_new_tokens),
            temperature=0.2,
            top_p=0.95,
            explain=False,  # default off
        )
        dt = time.time() - t0

        soap = out["soap"]
        mapping = out.get("mapping") or {}
        timings = out.get("timings") or {}

        assess_n = len(soap.get("assessment") or [])
        plan_n = len(soap.get("plan") or [])
        annotated_n = mapping.get("claims_count") or (assess_n + plan_n)
        uniq_evid = mapping.get("unique_evidence_count") or 0
        cap = mapping.get("registry_cap")

        # Console telemetry mirrors prior style
        _write_run_log(
            f"Result: {{'case_id': '{case_id}', 'generate_secs': {timings.get('generate_secs', 0)}, "
            f"'map_secs': {timings.get('map_secs', 0)}, 'total_runtime': {dt:.3f}, "
            f"'assessment_items': {assess_n}, 'plan_items': {plan_n}, 'annotated_items': {annotated_n}, "
            f"'unique_evidence': {uniq_evid}, 'cache_stub': ''}}"
        )

        st.success(
            f"{case_id}: assessed {assess_n} / plan {plan_n} | claims {annotated_n} | unique evidence {uniq_evid} | "
            f"gen {timings.get('generate_secs', 0)}s, map {timings.get('map_secs', 0)}s"
        )
        print(f"[page99] MAP_MODE='validation' cap={cap}  unique_evidence={uniq_evid}")

        results.append(
            {
                "case_id": case_id,
                "generate_secs": timings.get("generate_secs", 0),
                "map_secs": timings.get("map_secs", 0),
                "total_runtime": round(dt, 3),
                "assessment_items": assess_n,
                "plan_items": plan_n,
                "annotated_items": annotated_n,
                "unique_evidence": uniq_evid,
            }
        )

    # Save results
    if RESULTS_CSV.exists():
        prev = pd.read_csv(RESULTS_CSV)
        df = pd.concat([prev, pd.DataFrame(results)], ignore_index=True)
    else:
        df = pd.DataFrame(results)
    df.to_csv(RESULTS_CSV, index=False)
    _write_run_log(f"Results saved to: {RESULTS_CSV}")

    st.dataframe(df)
    st.success(f"Saved to: {RESULTS_CSV} and {RUN_LOG}")


if __name__ == "__main__":
    main()