#!/usr/bin/env python3 """Full pipeline dry-run on sample posts with detailed per-post reporting. Processes data/samples/posts.json against real catalog.db, shows: - extraction result, matching details, dgred payload — WITHOUT sending to CRM. """ from __future__ import annotations import json import logging import sys from collections import Counter from datetime import datetime, timedelta from pathlib import Path # Ensure src is importable sys.path.insert(0, str(Path(__file__).resolve().parent.parent / "src")) from agent_samochodowy.config import Settings from agent_samochodowy.dgred.client import DgredClient from agent_samochodowy.dgred.writer import LeadWriter from agent_samochodowy.extraction import extract from agent_samochodowy.extraction.prefilter import is_part_request as prefilter_check from agent_samochodowy.matcher.engine import match from agent_samochodowy.matcher.index import CatalogIndex from agent_samochodowy.models import Post from agent_samochodowy.status import map_status logging.basicConfig( level=logging.WARNING, format="%(asctime)s %(name)s %(levelname)s %(message)s", ) # Silence httpx noise logging.getLogger("httpx").setLevel(logging.WARNING) logging.getLogger("openai").setLevel(logging.WARNING) STATUS_ID_MAP = {"Dopasowano": 67, "Do weryfikacji": 68, "Brak dopasowania": 69} def main() -> None: settings = Settings() settings.dry_run = True # force dry-run print(f"LLM Provider: {settings.llm_provider}") print(f"LLM Model: {settings.llm_model}") print(f"DB Path: {settings.db_path}") print(f"Post max age: {settings.post_max_age_hours}h") print() # Load sample posts posts_path = Path(__file__).resolve().parent.parent / "data" / "samples" / "posts.json" raw_posts = json.loads(posts_path.read_text()) posts = [Post(**p) for p in raw_posts] total_fetched = len(posts) # Freshness filter cutoff = datetime.now() - timedelta(hours=settings.post_max_age_hours) if settings.post_max_age_hours > 0 else None if cutoff: fresh = [p for p in posts if p.timestamp.replace(tzinfo=None) >= cutoff] skipped_old = len(posts) - len(fresh) posts = fresh else: skipped_old = 0 print(f"Posts fetched: {total_fetched}") print(f"Too old (>{settings.post_max_age_hours}h): {skipped_old}") print(f"To analyze: {len(posts)}") print("=" * 80) index = CatalogIndex(settings.db_path) dgred = DgredClient(settings) writer = LeadWriter(dgred, settings) # Stats statuses: Counter[str] = Counter() methods: Counter[str] = Counter() part_requests = 0 for post in posts: print() print(f"{'─' * 80}") print(f"POST {post.post_id} | Grupa: {post.group}") print(f"Autor: {post.author_name}") print(f"Tekst: {post.text}") print() # Pre-filter passes_prefilter = prefilter_check(post.text) if not passes_prefilter: print(f" PREFILTER: ODRZUCONY (brak intencji kupna)") print(f" STATUS: --- (pominięty)") continue # Extract query = extract(post, settings) if query is None: print(f" PREFILTER: przeszedł, ale ekstraktor: ODRZUCONY") print(f" STATUS: --- (pominięty)") continue part_requests += 1 print(f" KLASYFIKACJA: zapytanie o część = {query.is_part_request}") print(f" POWÓD: {query.reason}") print(f" ATRYBUTY:") print(f" Marka: {query.brand or '—'}") print(f" Model: {query.model or '—'}") print(f" Rocznik: {query.year or '—'}") print(f" Kategoria: {query.category or '—'}") print(f" Kod silnika: {query.engine_code or '—'}") print(f" Numery OE: {', '.join(query.oe_numbers) if query.oe_numbers else '—'}") # Match result = match(query, index) status_name = map_status(result, settings) status_id = STATUS_ID_MAP.get(status_name, 0) statuses[status_name] += 1 methods[result.method] += 1 print() print(f" DOPASOWANIE:") print(f" Metoda: {result.method}") print(f" Pewność: {result.confidence:.0%}") print(f" Trafień: {len(result.offers)}") if result.offers: print(f" Oferty (top 3):") for offer in result.offers[:3]: price_str = f"{offer.price} {offer.currency}" if offer.price else "brak ceny" print(f" - {offer.title[:70]}") print(f" {offer.url}") print(f" Cena: {price_str} | OE: {', '.join(offer.oe_numbers[:3])}") print() print(f" STATUS: {status_id} = {status_name}") # Build dgred payload (dry-run) note = writer._build_note(post, query, result, status_name) custom_fields: dict = {} if query.brand: custom_fields["marka"] = query.brand if query.model: custom_fields["model"] = query.model if query.year: custom_fields["rocznik"] = query.year if query.category: custom_fields["kategoria"] = query.category if query.engine_code: custom_fields["kod_silnika"] = query.engine_code if query.oe_numbers: custom_fields["numery_oe"] = ", ".join(query.oe_numbers) custom_fields["fb_profil"] = post.author_profile_url custom_fields["url_posta"] = post.post_url custom_fields["grupa"] = post.group payload = { "client": { "name": post.author_name, "customFields": custom_fields, "tags": [{"tag": "FB Agent", "color": 1}], "externalId": post.author_profile_url, }, "clientStatus": status_id, "note": note, } print() print(f" PAYLOAD DO DGRED (dry-run):") print(json.dumps(payload, indent=4, ensure_ascii=False)) index.close() dgred.close() # Summary print() print("=" * 80) print("PODSUMOWANIE") print("=" * 80) print(f"Pobrano postów: {total_fetched}") print(f"Za stare (>{settings.post_max_age_hours}h): {skipped_old}") print(f"Do analizy: {len(posts)}") print(f"Zapytania o część: {part_requests}") print(f"Odrzucone (prefilter/LLM): {len(posts) - part_requests}") print() print("Rozkład statusów:") for s in ["Dopasowano", "Do weryfikacji", "Brak dopasowania"]: print(f" {STATUS_ID_MAP.get(s, '?')} {s:25s} {statuses.get(s, 0)}") print() print("Trafienia wg metody:") for m in ["oe", "engine_code", "atrybuty", "fuzzy", "brak"]: if methods.get(m, 0): print(f" {m:20s} {methods[m]}") print("=" * 80) if __name__ == "__main__": main()