"""Offline reprocessing of residual questions against the final canonical identity map."""
import argparse
import json
from collections import defaultdict
from pathlib import Path

from central.app.ai.kb_consolidation import canonical_product_id
from central.app.ai.workitems import build_work_items


def _erp_code(question):
    import re
    match = re.search(r"PRODUCT\s+(\d+)", question.get("QUESTION", ""))
    return match.group(1) if match else str(question.get("DATABASE_CODE") or "")


def main(questions_path, payload_path, output_path):
    questions_doc = json.loads(Path(questions_path).read_text(encoding="utf-8"))
    payload = json.loads(Path(payload_path).read_text(encoding="utf-8"))
    canonical = {}
    fingerprints = {}
    by_observation = defaultdict(list)
    for product in payload.get("products", []):
        cid = canonical_product_id(product)
        for observation in product.get("observations", []):
            key = f"{observation.get('database')}:{observation.get('erp_code')}"
            canonical[key] = cid
            fingerprints[key] = product.get("fingerprint") or cid
    questions = [dict(question, DATABASE_CODE=_erp_code(question)) for question in questions_doc.get("questions", [])]
    items = build_work_items(questions, fingerprints, canonical)
    by_task = {}
    for item in items:
        bucket = by_task.setdefault(item.task_type, {"items": 0, "cells": 0})
        bucket["items"] += 1
        bucket["cells"] += item.cells_covered
    report = {
        "artifact": "AI_RESIDUAL_PLAN_V3",
        "work_items_before": 750,
        "work_items_after": len(items),
        "llm_items_before": 511,
        # Candidate preparation is evidence-only; without an official validated
        # answer it must not reclassify a selection item as resolved.
        "llm_items_after": 511,
        "llm_reclassification": "UNCHANGED_PENDING_FINAL_PROVIDER_SELECTION",
        "by_task_type": by_task,
        "canonical_observation_keys": len(canonical),
        "ai_write_allowed": False,
        "semantic_decisions": 0,
    }
    Path(output_path).write_text(json.dumps(report, indent=2, ensure_ascii=False) + "\n", encoding="utf-8")
    return report


if __name__ == "__main__":
    parser = argparse.ArgumentParser()
    parser.add_argument("--questions", default="docs/LOCAL_FISCAL_CONTEXT_QUESTIONS_V2.json")
    parser.add_argument("--payload", default="docs/runs/ai_kb_seed_payload.json")
    parser.add_argument("--output", default="docs/AI_RESIDUAL_PLAN_V3.json")
    args = parser.parse_args()
    print(json.dumps(main(args.questions, args.payload, args.output), indent=2, ensure_ascii=False))