modules/chat.py

import json
import logging
import random
import os

from modules import ai_provider
from modules import lexicon
import config

logger = logging.getLogger(__name__)


# Load prompts from .ilang files (prompts/ if exists, else prompts_demo/)
def _load_prompt(name):
    for d in ("prompts", "prompts_demo"):
        path = os.path.join(os.path.dirname(os.path.dirname(__file__)), d, name)
        if os.path.exists(path):
            with open(path, "r", encoding="utf-8") as f:
                content = f.read()
                logger.info("Loaded prompt: " + path + " (" + str(len(content)) + " chars)")
                return content
    logger.warning("Prompt not found: " + name)
    return ""


SYSTEM_PROMPT = _load_prompt("persona.ilang")
ANTISPAM_TEXT_PROMPT = _load_prompt("antispam.ilang")
VISION_PROMPT = _load_prompt("vision.ilang")

if not SYSTEM_PROMPT:
    SYSTEM_PROMPT = "You are TelegramGuard, a helpful AI assistant on Telegram. Reply concisely in the user's language. JSON format: {\"intent\": \"chat\", \"reply\": \"your text\"}"
    logger.warning("Using fallback system prompt")

GROUP_WELCOME = (
    "I-Lang Guard is here\n\n"
    "I auto-clean spam. No config needed.\n"
    "@ me if you need anything.\n\n"
    "Just give me admin permissions (delete messages + ban users)."
)


def _parse(raw):
    if not raw:
        return ("chat", "...")
    t = raw.strip()
    if t.startswith("```"):
        nl = t.find("\n")
        t = t[nl + 1:] if nl > 0 else t[3:]
    if t.endswith("```"):
        t = t[:-3].strip()
    try:
        d = json.loads(t)
        return (d.get("intent", "chat"), d.get("reply", t))
    except json.JSONDecodeError:
        pass
    last_brace = t.rfind("}")
    while last_brace >= 0:
        start = t.rfind("{", 0, last_brace)
        if start >= 0:
            try:
                d = json.loads(t[start:last_brace + 1])
                return (d.get("intent", "chat"), d.get("reply", t))
            except json.JSONDecodeError:
                pass
        last_brace = t.rfind("}", 0, last_brace)
    for line in t.split("\n"):
        line = line.strip()
        if line and not line.startswith("{") and not line.startswith("taint") and not line.startswith("The "):
            return ("chat", line)
    return ("chat", "...")


def _ctx(history, info):
    parts = []
    if info:
        parts.append("[ctx] " + info)
    if history:
        for h in history[-8:]:
            r = "user" if h["role"] == "user" else "bot"
            parts.append(r + ": " + h["text"])
    return "\n".join(parts)


def _deflect():
    lines = [
        "That's a tough one. What else can I help with?",
        "Let's try a different angle. What do you need today?",
        "That's beyond my range. What else is on your mind?",
    ]
    return random.choice(lines)


def _clip(text, limit):
    """Keep the head AND the tail of an over-long message, not just the head.

    The old text[:1000] / caption[:500] cut the message before the judge ever saw
    it, while Telegram allows 4096 characters in one message — so pasting a news
    article in front and appending the payload at the end hid the payload
    completely. Taking half from each end guarantees an ad tacked onto the tail
    lands inside the judged range.
    """
    t = text or ""
    if len(t) <= limit:
        return t
    half = limit // 2
    return t[:half] + "\n…(middle omitted)…\n" + t[-half:]


def _is_spam(raw):
    """Parse a spam-judge reply into True / False / None (couldn't parse it).

    Fixes the old `"spam" in result` substring bug (a wordy 'not spam' counted as
    spam) and the fail-open that replaced it: anything not starting with spam/yes
    silently meant "clean", so a model that prefixes its answer ("Based on the
    above, yes") or gets truncated inside that preamble was a free pass.
    None means unparseable — callers fall back to the lexicon instead of letting
    the message through.
    """
    s = (raw or "").strip().lower().lstrip("\"'`*# 　")
    if not s:
        return None
    if s.startswith(("spam", "yes", "y,", "违规", "是", "有")) or s == "y":
        return True
    if s.startswith(("ok", "no", "n,", "not ", "clean", "正常", "否", "不是", "无")) or s == "n":
        return False
    return None


def _lexicon_fallback(text):
    """Used when the verdict is unparseable or the API call failed: fall back to
    a lexicon hard hit rather than silently letting the message through."""
    try:
        s, _ = lexicon.score(text or "")
        return s >= getattr(config, "LEXICON_HARD_THRESHOLD", 6)
    except Exception as e:
        logger.warning("lexicon fallback failed: " + str(e))
        return False


async def ai_text(text, history=None, context_info=""):
    try:
        c = _ctx(history, context_info)
        prompt = c + "\nuser: " + text if c else "user: " + text
        raw = await ai_provider.generate_text(prompt, system=SYSTEM_PROMPT)
        if not raw:
            return ("chat", _deflect())
        logger.info("AI raw[" + str(len(raw)) + "]: " + raw[:200])
        return _parse(raw)
    except Exception as e:
        logger.warning("AI text exception: " + str(e))
        return ("chat", _deflect())


async def ai_vision(image_bytes, caption="", history=None, context_info=""):
    try:
        c = _ctx(history, context_info)
        prompt = VISION_PROMPT + "\n" + c
        if caption:
            prompt += "\nuser: " + caption
        raw = await ai_provider.generate_vision(prompt, image_bytes, system=SYSTEM_PROMPT)
        return _parse(raw)
    except Exception as e:
        logger.warning("AI vision: " + str(e))
        return ("chat", "Couldn't read that image. Try another one?")


async def ai_voice(audio_bytes, mime_type="audio/ogg", history=None, context_info=""):
    try:
        c = _ctx(history, context_info)
        prompt = c + "\nUser sent a voice message:" if c else "User sent a voice message:"
        fmt = "ogg"
        if "/" in mime_type:
            fmt = mime_type.split("/", 1)[1].split(";")[0] or "ogg"
        raw = await ai_provider.generate_audio(prompt, audio_bytes, fmt=fmt, system=SYSTEM_PROMPT)
        return _parse(raw)
    except Exception as e:
        logger.warning("AI voice: " + str(e))
        return ("chat", "Didn't catch that. Try again or type it out.")


async def ai_judge_group_message(text, sender_context=""):
    try:
        prompt = ANTISPAM_TEXT_PROMPT
        if sender_context:
            prompt += "\n\nSender context: " + sender_context
        prompt += "\n\nMessage content: " + _clip(text, getattr(config, "JUDGE_TEXT_LIMIT", 1800))
        # max_tokens 32, not 8: eight tokens is enough to truncate the answer
        # inside a preamble ("Based on the above,"), which leaves nothing to
        # parse and drops the judgement down to the lexicon for no reason.
        raw = await ai_provider.generate_text(prompt, max_tokens=32, temperature=0.0)
        verdict = _is_spam(raw)
        if verdict is None:  # unparseable — don't let it through silently
            logger.warning("spam text verdict unparseable: " + repr((raw or "")[:80]) + " — falling back to lexicon")
            return _lexicon_fallback(text)
        return verdict
    except Exception as e:
        logger.warning("spam text judge failed, falling back to lexicon: " + str(e))
        return _lexicon_fallback(text)


async def ai_judge_group_image(image_bytes, caption="", sender_context=""):
    try:
        prompt = ANTISPAM_TEXT_PROMPT + "\n\nJudge this image. Reply ONLY: spam or ok."
        if sender_context:
            prompt += "\nSender context: " + sender_context
        if caption:
            prompt += "\nCaption: " + _clip(caption, getattr(config, "JUDGE_CAPTION_LIMIT", 900))
        raw = await ai_provider.generate_vision(prompt, image_bytes, max_tokens=32, temperature=0.0)
        verdict = _is_spam(raw)
        if verdict is None:  # same as above — no silent pass
            logger.warning("spam image verdict unparseable: " + repr((raw or "")[:80]) + " — falling back to lexicon")
            return _lexicon_fallback(caption)
        return verdict
    except Exception as e:
        logger.warning("spam image judge failed, falling back to lexicon: " + str(e))
        return _lexicon_fallback(caption)


async def ai_group_vision(image_bytes, caption="", history=None):
    try:
        ctx = _ctx(history, "GROUP_CHAT: User shared an image and @mentioned you. Comment naturally in 1-2 sentences.")
        prompt = SYSTEM_PROMPT + "\n" + ctx
        if caption:
            prompt += "\nuser: " + caption
        else:
            prompt += "\nuser: [shared an image]"
        raw = await ai_provider.generate_vision(prompt, image_bytes, system=SYSTEM_PROMPT)
        raw = (raw or "").strip()
        if not raw:
            return _deflect()
        intent, reply = _parse(raw)
        if reply in ("...", ""):
            return _deflect()
        return reply
    except Exception:
        return _deflect()


async def ai_group_reply(text, history=None):
    try:
        ctx = _ctx(history, "GROUP_CHAT: You were @mentioned in a group. Reply directly, 1-2 sentences.")
        prompt = ctx + "\nuser: " + text
        raw = await ai_provider.generate_text(prompt, system=SYSTEM_PROMPT)
        raw = (raw or "").strip()
        if not raw:
            return _deflect()
        intent, reply = _parse(raw)
        if reply in ("...", ""):
            return _deflect()
        return reply
    except Exception:
        return _deflect()