"""Ollama VM123 + fallback — Spec 029 T0/T1."""
from __future__ import annotations

import os

import httpx

OLLAMA_BASE_URL = os.getenv("OLLAMA_BASE_URL", "http://10.10.10.123:11434").rstrip("/")
AGENTIC_LLM_MODEL = os.getenv("AGENTIC_LLM_MODEL", "qwen2.5:7b-instruct")
AGENTIC_EMBED_MODEL = os.getenv("AGENTIC_EMBED_MODEL", "nomic-embed-text")
AGENTIC_LLM_ENABLED = os.getenv("AGENTIC_LLM_ENABLED", "false").lower() in ("1", "true", "yes")


def ollama_available() -> bool:
    try:
        with httpx.Client(timeout=3.0) as c:
            return c.get(f"{OLLAMA_BASE_URL}/api/tags").status_code == 200
    except Exception:
        return False


def _chat(prompt: str, *, system: str | None = None, max_tokens: int = 800) -> tuple[str, str]:
    if not AGENTIC_LLM_ENABLED or not ollama_available():
        return ("", "t0")
    messages = []
    if system:
        messages.append({"role": "system", "content": system})
    messages.append({"role": "user", "content": prompt})
    try:
        with httpx.Client(timeout=120.0) as c:
            r = c.post(
                f"{OLLAMA_BASE_URL}/api/chat",
                json={"model": AGENTIC_LLM_MODEL, "messages": messages, "stream": False},
            )
        if r.status_code == 200:
            txt = (r.json().get("message") or {}).get("content", "").strip()
            if txt:
                return txt, AGENTIC_LLM_MODEL
    except Exception:
        pass
    return ("", "t0-fallback")


def advise_human_action(
    *, finding_title: str, finding_detail: str, kb_snippets: list[str] | None = None
) -> tuple[str, str]:
    prompt = (
        "Advisor Agentic Ops Ligbox. Português BR, máx 6 frases. O que fazer AGORA?\n"
        f"Problema: {finding_title}\nDetalhe: {finding_detail}\n"
        f"KB: {'---'.join(kb_snippets or [])[:2500] or 'N/A'}"
    )
    txt, model = _chat(prompt)
    if txt:
        return txt, model
    return (f"Investigar manualmente: {finding_title}", "t0")


def chat_context(
    *,
    question: str,
    kb_snippets: list[str] | None = None,
    findings_summary: str | None = None,
    user_role: str = "technician",
) -> tuple[str, str]:
    """T1 — resposta contextual para janela Desk / bot interno."""
    system = (
        "És o copiloto Agentic Ops da Ligbox (VM112 wizard, VM122 Desk, VM123 finance). "
        "Responde em português BR, objectivo, com passos acionáveis. "
        "Nunca inventes credenciais. Se não souberes, diz o que verificar."
    )
    ctx = []
    if findings_summary:
        ctx.append(f"Findings abertos:\n{findings_summary[:2000]}")
    if kb_snippets:
        ctx.append("KB:\n" + "\n---\n".join(kb_snippets[:6])[:4000])
    prompt = (
        f"Perfil utilizador: {user_role}\n"
        f"Contexto ops:\n{chr(10).join(ctx) or 'N/A'}\n\n"
        f"Pergunta: {question}"
    )
    txt, model = _chat(prompt, system=system)
    if txt:
        return txt, model
    return (
        "Modo T0 activo — LLM indisponível. Consulte findings e audit log no painel Agentic Ops.",
        "t0",
    )