filter_comments.py

#!/usr/bin/env python3
"""Run a JSON file of comments through all filters and print the result.

Usage:
    python filter_comments.py <comments.json>

The input file is a JSON list of comments in the shape produced by the API
clients' ``get_comments`` (see fetch_comments.py). The filtered list is printed
to stdout as JSON; debug logs are written to a log file (see
chorus.logging_config).
"""
import json
import os
import sys

from dotenv import load_dotenv
from chorus.filters import apply_filters
from chorus.llm.llama_client import OpenAICompatibleLLMClient
from chorus.logging_config import configure_logging


def build_llm_client() -> OpenAICompatibleLLMClient:
    """Build the LLM client used by the LLM-based filters.

    The endpoint/model can be overridden via the environment (see
    .env.example); otherwise the defaults are used.
    """
    return OpenAICompatibleLLMClient.from_env()


def main(argv):
    if len(argv) != 2:
        print(f"Usage: {argv[0]} <comments.json>", file=sys.stderr)
        return 1

    log = configure_logging()
    load_dotenv()

    with open(argv[1], encoding="utf-8") as f:
        data = json.load(f)
    
    is_dict = isinstance(data, dict)
    if is_dict:
        comments = data.get("comments") or []
        metadata = {k: v for k, v in data.items() if k != "comments"}
    else:
        comments = data

    log.info("Loaded %d comment(s) from %s", len(comments), argv[1])

    llm_client = build_llm_client()
    
    if is_dict:
        print("{")
        for k, v in metadata.items():
            print(f'  "{k}": {json.dumps(v, ensure_ascii=False)},')
        print('  "comments": [')
    else:
        print("[")

    count = 0
    for i, comment in enumerate(apply_filters(comments, llm_client=llm_client)):
        if i > 0:
            print(",")
        json_str = json.dumps(comment, indent=2, ensure_ascii=False)
        # Indent each line by 2 or 4 spaces to maintain the look of a JSON list
        indent = "    " if is_dict else "  "
        indented = "\n".join(indent + line for line in json_str.splitlines())
        print(indented, end="", flush=True)
        count = i + 1

    if is_dict:
        print("\n  ]")
        print("}")
    else:
        print("\n]")
    
    log.info("Filtering produced %d comment(s)", count)
    return 0


if __name__ == "__main__":
    sys.exit(main(sys.argv))