core/config.py

"""Runtime configuration for the content generation agent.

All settings can come from environment variables (for Docker / server deploys)
or be overridden at runtime from the Streamlit UI (so a user can paste their
own API key and choose a model before running a command).
"""
from __future__ import annotations

import os
try:
    from dotenv import load_dotenv
    load_dotenv()
except ImportError:
    pass
from dataclasses import dataclass, field, asdict
from typing import Optional

# OpenAI-compatible chat-completion endpoints for the supported providers.
PROVIDER_ENDPOINTS = {
    "groq": "https://api.groq.com/openai/v1/chat/completions",
    "openrouter": "https://openrouter.ai/api/v1/chat/completions",
}

# Sensible default models per provider.
DEFAULT_MODELS = {
    "groq": "llama-3.3-70b-versatile",
    "openrouter": "meta-llama/llama-3.3-70b-instruct",
}

# A short menu the UI can present per provider.
MODEL_CHOICES = {
    "groq": [
        "llama-3.3-70b-versatile",
        "llama-3.1-8b-instant",
        "openai/gpt-oss-20b",
    ],
    "openrouter": [
        "meta-llama/llama-3.3-70b-instruct",
        "openai/gpt-4o-mini",
        "anthropic/claude-3.5-sonnet",
        "google/gemini-flash-1.5",
    ],
}


def _env_bool(name: str, default: bool) -> bool:
    val = os.getenv(name)
    if val is None:
        return default
    return val.strip().lower() in {"1", "true", "yes", "on"}


@dataclass
class Settings:
    """Resolved settings used by the pipeline."""

    provider: str = field(default_factory=lambda: os.getenv("LLM_PROVIDER", "groq"))
    api_key: Optional[str] = field(default_factory=lambda: os.getenv("LLM_API_KEY"))
    model: Optional[str] = field(default_factory=lambda: os.getenv("LLM_MODEL"))
    # Mock mode lets the whole pipeline run with no network / no API key.
    mock_mode: bool = field(default_factory=lambda: _env_bool("MOCK_MODE", False))
    temperature: float = field(default_factory=lambda: float(os.getenv("LLM_TEMPERATURE", "0.7")))
    max_tokens: int = field(default_factory=lambda: int(os.getenv("LLM_MAX_TOKENS", "1200")))

    # Crawl controls
    crawl_max_pages: int = field(default_factory=lambda: int(os.getenv("CRAWL_MAX_PAGES", "40")))
    crawl_max_depth: int = field(default_factory=lambda: int(os.getenv("CRAWL_MAX_DEPTH", "2")))
    crawl_timeout: int = field(default_factory=lambda: int(os.getenv("CRAWL_TIMEOUT", "10")))

    def resolved_model(self) -> str:
        if self.model:
            return self.model
        return DEFAULT_MODELS.get(self.provider, DEFAULT_MODELS["groq"])

    def endpoint(self) -> str:
        return PROVIDER_ENDPOINTS.get(self.provider, PROVIDER_ENDPOINTS["groq"])

    def is_live(self) -> bool:
        """True when we should make real API calls."""
        return (not self.mock_mode) and bool(self.api_key)

    def to_dict(self) -> dict:
        d = asdict(self)
        # Never echo the full key back.
        if d.get("api_key"):
            d["api_key"] = d["api_key"][:4] + "…"
        return d


def get_settings(**overrides) -> Settings:
    """Build settings from env, then apply any UI overrides (non-None only)."""
    s = Settings()
    for k, v in overrides.items():
        if v is not None and hasattr(s, k):
            setattr(s, k, v)
    return s