config.py

"""
Centralized configuration helpers for the Mixedbread CVE RAG workflow.
"""

from __future__ import annotations

import os
from dataclasses import dataclass
from pathlib import Path


@dataclass
class Settings:
    """Container for frequently used paths and model configuration."""

    project_root: Path
    zip_path: Path
    cve_root: Path
    corpus_path: Path
    chroma_dir: Path
    chroma_collection: str
    embed_model: str
    rerank_model: str
    hf_token: str  # Optional, only needed for downloading models from HF Hub
    embed_batch_size: int
    device: str  # "cpu" or "cuda" for GPU

def load_settings() -> Settings:
    root = Path(__file__).resolve().parents[1]
    data_root = root / "data" / "cvelistV5-main"
    artifacts_dir = root / "rag_mixedbread" / "artifacts"
    chroma_dir = root / "rag_mixedbread" / "index"

    hf_token = os.environ.get("HF_API_TOKEN", "")

    settings = Settings(
        project_root=root,
        zip_path=root / "testing" / "cvelistV5-main.zip",
        cve_root=data_root,
        corpus_path=artifacts_dir / "cve_corpus.jsonl",
        chroma_dir=chroma_dir,
        chroma_collection=os.environ.get("RAG_CHROMA_COLLECTION", "cve_chunks"),
        embed_model=os.environ.get(
            "RAG_EMBED_MODEL", "mixedbread-ai/mxbai-embed-large-v1"
        ),
        rerank_model=os.environ.get(
            "RAG_RERANK_MODEL", "mixedbread-ai/mxbai-rerank-base-v2"
        ),
        hf_token=hf_token,  # Optional, only needed for private models or rate-limited downloads
        embed_batch_size=int(os.environ.get("RAG_EMBED_BATCH", "8")),
        device=os.environ.get("RAG_DEVICE", "cpu"),  # "cpu" or "cuda"
    )

    artifacts_dir.mkdir(parents=True, exist_ok=True)
    chroma_dir.mkdir(parents=True, exist_ok=True)
    return settings


__all__ = ["Settings", "load_settings"]