scripts/reviews/02_build_data.py

#!/usr/bin/env python3
import json
import logging
import sys
from pathlib import Path

# ---------------------------------------------------------------------------
# Config
# ---------------------------------------------------------------------------

ROOT = Path(__file__).parent.parent.parent
SOURCE_DIR = ROOT / "source" / "reviews" / "detail"
OUTPUT_FILE = ROOT / "data" / "reviews.jsonl"
PREFIX = "review"

logging.basicConfig(level=logging.INFO, format="%(message)s")
log = logging.getLogger(__name__)

# ---------------------------------------------------------------------------
# Mapper
# ---------------------------------------------------------------------------

def transform_review(data: dict) -> dict:
    """Flatten nested review JSON into a ChromaDB-ready format."""
    
    # 1. Generate Globally Unique ID
    chroma_id = f"{PREFIX}:{int(data['id']):06d}"
    
    # 2. Build searchable document text
    sections = data.get("sections", {})
    text_blocks = []
    if data.get("kit"): text_blocks.append(f"Review of the {data['kit']}.")
    
    # Add all narrative sections
    for title, content in sections.items():
        text_blocks.append(f"{title}: {content}")
    
    document = " ".join(text_blocks)
    
    # 3. Flatten metadata (simple types only)
    metadata = {
        "id": data["id"],
        "type": data.get("type"),
        "date": data.get("date"),
        "kit_name": data.get("kit"),
        "manufacturer_name": data.get("manufacturer", {}).get("name") if isinstance(data.get("manufacturer"), dict) else data.get("manufacturer"),
        "rating_construction": data.get("ratings", {}).get("construction"),
        "rating_flight": data.get("ratings", {}).get("flight"),
        "rating_overall": data.get("ratings", {}).get("overall"),
        "url": data.get("url")
    }
    
    # Remove nulls to keep metadata clean
    metadata = {k: v for k, v in metadata.items() if v is not None}
    
    return {
        "id": chroma_id,
        "document": document,
        "metadata": metadata
    }

# ---------------------------------------------------------------------------
# Main
# ---------------------------------------------------------------------------

def main():
    if not SOURCE_DIR.exists():
        log.error(f"Source directory {SOURCE_DIR} not found.")
        return

    OUTPUT_FILE.parent.mkdir(parents=True, exist_ok=True)
    
    count = 0
    with OUTPUT_FILE.open("w", encoding="utf-8") as out:
        # Walk the sharded detail directory
        for shard_dir in sorted(SOURCE_DIR.iterdir()):
            if not shard_dir.is_dir(): continue
            
            for file_path in sorted(shard_dir.glob("*.json")):
                try:
                    with file_path.open("r", encoding="utf-8") as f:
                        raw_data = json.load(f)
                    
                    processed = transform_review(raw_data)
                    out.write(json.dumps(processed, ensure_ascii=False) + "\n")
                    count += 1
                except Exception as e:
                    log.error(f"Error processing {file_path}: {e}")

    log.info(f"Successfully built {count} records in {OUTPUT_FILE}")

if __name__ == "__main__":
    main()