Files
discord_bot/fjerkroa_bot/ai_responder.py
T

366 lines
14 KiB
Python
Raw Blame History

This file contains invisible Unicode characters
This file contains invisible Unicode characters that are indistinguishable to humans but may be processed differently by a computer. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
import asyncio
import json
import logging
import os
import pickle
import random
import re
import time
from functools import lru_cache, wraps
from io import BytesIO
from pathlib import Path
from pprint import pformat
from typing import Any, Dict, List, Optional, Tuple, Union
from .persistence import PersistentStore
def pp(*args, **kw):
if "width" not in kw:
kw["width"] = 300
return pformat(*args, **kw)
@lru_cache(maxsize=300)
def parse_json(content: str) -> Dict:
# Strict JSON only — model output is schema-enforced (ENV-18/19),
# history entries are json.dumps products.
return json.loads(content.strip())
def sanitize_external_text(text: str, max_len: int = 4000) -> str:
"""Neutralize attacker-influenced text before it enters a prompt (SAF-03)."""
text = re.sub(r"[\x00-\x08\x0b\x0c\x0e-\x1f\x7f]", "", text)
text = text.replace("@everyone", "@everyone").replace("@here", "@here")
return text[:max_len]
def exponential_backoff(base=2, max_delay=60, factor=1, jitter=0.1, max_attempts=None):
"""Generate sleep intervals for exponential backoff with jitter.
Args:
base: Base of the exponentiation operation
max_delay: Maximum delay
factor: Multiplication factor for each increase in backoff
jitter: Additional randomness range to prevent thundering herd problem
Yields:
Delay for backoff as a floating point number.
"""
attempt = 0
while True:
sleep = min(max_delay, factor * base**attempt)
jitter_amount = jitter * sleep
sleep += random.uniform(-jitter_amount, jitter_amount)
yield sleep
attempt += 1
if max_attempts is not None and attempt > max_attempts:
raise RuntimeError("Max attempts reached in exponential backoff.")
def async_cache_to_file(filename):
cache_file = Path(filename)
cache = None
if cache_file.exists():
try:
with cache_file.open("rb") as fd:
cache = pickle.load(fd)
except Exception:
cache = {}
def decorator(func):
@wraps(func)
async def wrapper(*args, **kwargs):
if cache is None:
return await func(*args, **kwargs)
key = json.dumps((func.__name__, list(args[1:]), kwargs), sort_keys=True)
if key in cache:
return cache[key]
result = await func(*args, **kwargs)
cache[key] = result
with cache_file.open("wb") as fd:
pickle.dump(cache, fd)
return result
return wrapper
return decorator
class AIMessageBase(object):
def __init__(self) -> None:
self.vars: List[str] = []
def __str__(self) -> str:
return json.dumps({k: v for k, v in vars(self).items() if k in self.vars})
class AIMessage(AIMessageBase):
def __init__(self, user: str, message: str, channel: str = "chat", direct: bool = False, historise_question: bool = True) -> None:
self.user = user
self.message = message
self.urls: Optional[List[str]] = None
self.channel = channel
self.direct = direct
self.historise_question = historise_question
self.vars = ["user", "message", "channel", "direct", "historise_question"]
class AIResponse(AIMessageBase):
def __init__(
self,
answer: Optional[str],
answer_needed: bool,
channel: Optional[str],
staff: Optional[str],
picture: Optional[str],
picture_edit: bool,
hack: bool,
) -> None:
self.answer = answer
self.answer_needed = answer_needed
self.channel = channel
self.staff = staff
self.picture = picture
self.picture_edit = picture_edit
self.hack = hack
self.vars = ["answer", "answer_needed", "channel", "staff", "picture", "hack"]
class AIResponderBase(object):
def __init__(self, config: Dict[str, Any], channel: Optional[str] = None) -> None:
super().__init__()
self.config = config
self.channel = channel if channel is not None else "system"
class AIResponder(AIResponderBase):
def __init__(self, config: Dict[str, Any], channel: Optional[str] = None) -> None:
super().__init__(config, channel)
self.history: List[Dict[str, Any]] = []
self.memory: str = "I am an assistant."
self.rate_limit_backoff = exponential_backoff()
self.store: Optional[PersistentStore] = None
if "history-directory" in self.config:
directory = Path(self.config["history-directory"]).expanduser()
self.store = PersistentStore(directory / "bot.db")
# Legacy pickles import once, then live on as *.migrated (PER-03)
self.store.migrate_pickles(self.channel, directory / f"{self.channel}.dat", directory / f"{self.channel}.memory")
self.history = self.store.load_history(self.channel)
stored_memory = self.store.load_memory(self.channel)
if stored_memory is not None:
self.memory = stored_memory
logging.info(f"memmory:\n{self.memory}")
def message(self, message: AIMessage, limit: Optional[int] = None) -> List[Dict[str, Any]]:
messages = []
system = self.config.get(self.channel, self.config["system"])
system = system.replace("{date}", time.strftime("%Y-%m-%d")).replace("{time}", time.strftime("%H:%M:%S"))
news_feed = self.config.get("news")
if news_feed and os.path.exists(news_feed):
with open(news_feed) as fd:
news_feed = fd.read().strip()
system = system.replace("{news}", sanitize_external_text(news_feed))
system = system.replace("{memory}", self.memory)
messages.append({"role": "system", "content": system})
if limit is not None:
while len(self.history) > limit:
self.shrink_history_by_one()
for msg in self.history:
messages.append(msg)
if not message.urls:
messages.append({"role": "user", "content": str(message)})
else:
content: List[Dict[str, Union[str, Dict[str, str]]]] = [{"type": "text", "text": str(message)}]
for url in message.urls:
content.append({"type": "image_url", "image_url": {"url": url}})
messages.append({"role": "user", "content": content})
return messages
async def draw(self, description: str) -> BytesIO:
if self.config.get("leonardo-token") is not None:
return await self.draw_leonardo(description)
return await self.draw_openai(description)
async def draw_leonardo(self, description: str) -> BytesIO:
raise NotImplementedError()
async def draw_openai(self, description: str) -> BytesIO:
raise NotImplementedError()
async def post_process(self, message: AIMessage, response: Dict[str, Any]) -> AIResponse:
# Envelope arrives schema-validated (ENV-19); .get defaults keep old
# history entries and hand-built test dicts working.
answer = response.get("answer")
answer_needed = bool(response.get("answer_needed", False))
if answer is None:
answer_needed = False
else:
answer = str(answer)
answer = re.sub(r"@\[([^\]]*)\]\([^\)]*\)", r"\1", answer)
answer = re.sub(r"\[[^\]]*\]\(([^\)]*)\)", r"\1", answer)
if message.direct or message.user in message.message:
answer_needed = True
response_message = AIResponse(
answer,
answer_needed,
response.get("channel"),
response.get("staff"),
response.get("picture"),
bool(response.get("picture_edit", False)),
bool(response.get("hack", False)),
)
if response_message.staff is not None and response_message.answer is not None:
response_message.answer_needed = True
if response_message.channel is None:
response_message.channel = message.channel
return response_message
def short_path(self, message: AIMessage, limit: int) -> bool:
if message.direct or "short-path" not in self.config:
return False
for chan_re, user_re in self.config["short-path"]:
chan_ma = re.match(chan_re, message.channel)
user_ma = re.match(user_re, message.user)
if chan_ma and user_ma:
self.history.append({"role": "user", "content": str(message)})
while len(self.history) > limit:
self.shrink_history_by_one()
return True
return False
async def chat(self, messages: List[Dict[str, Any]], limit: int) -> Tuple[Optional[Dict[str, Any]], int]:
raise NotImplementedError()
async def memory_rewrite(self, memory: str, message_user: str, answer_user: str, question: str, answer: str) -> str:
raise NotImplementedError()
async def translate(self, text: str, language: str = "english") -> str:
raise NotImplementedError()
@staticmethod
def _entry_channel(item: Dict[str, Any]) -> Optional[str]:
try:
return parse_json(item["content"]).get("channel")
except Exception:
return None
def shrink_history_by_one(self) -> None:
if not self.history:
return
cap = self.config.get("history-per-channel", 3)
counts: Dict[Optional[str], int] = {}
for item in self.history:
chan = self._entry_channel(item)
counts[chan] = counts.get(chan, 0) + 1
for index, item in enumerate(self.history):
if counts[self._entry_channel(item)] > cap:
del self.history[index]
return
del self.history[0]
def update_history(self, question: Dict[str, Any], answer: Dict[str, Any], limit: int, historise_question: bool = True) -> None:
if not isinstance(question["content"], str):
question["content"] = question["content"][0]["text"]
if historise_question:
self.history.append(question)
self.history.append(answer)
while len(self.history) > limit:
self.shrink_history_by_one()
def update_memory(self, memory) -> None:
self.memory = memory
async def _persist_history(self) -> None:
if self.store is not None:
await asyncio.to_thread(self.store.save_history, self.channel, list(self.history))
async def _persist_memory(self) -> None:
if self.store is not None:
await asyncio.to_thread(self.store.save_memory, self.channel, self.memory)
async def handle_picture(self, response: Dict) -> bool:
if not isinstance(response.get("picture"), (type(None), str)):
logging.warning(f"picture key is wrong in response: {pp(response)}")
return False
if response.get("picture") is not None:
response["picture"] = await self.translate(response["picture"])
return True
def _parse_answer(self, answer: Dict[str, Any]) -> Optional[Dict[str, Any]]:
# Schema-enforced output should always parse; anything else is a
# failed attempt — no repair model (ENV-18).
try:
return parse_json(answer["content"])
except Exception as err:
logging.error(f"failed to parse the answer: {pp(err)}\n{repr(answer['content'])}")
return None
async def memoize(self, message_user: str, answer_user: str, message: str, answer: str) -> None:
self.memory = await self.memory_rewrite(self.memory, message_user, answer_user, message, answer)
self.update_memory(self.memory)
await self._persist_memory()
async def memoize_reaction(self, message_user: str, reaction_user: str, operation: str, reaction: str, message: str) -> None:
quoted_message = message.replace("\n", "\n> ")
await self.memoize(
message_user, "assistant", f"\n> {quoted_message}", f"User {reaction_user} has {operation} this raction: {reaction}"
)
async def send(self, message: AIMessage) -> AIResponse:
# Get the history limit from the configuration
limit = self.config["history-limit"]
# Check if a short path applies, return an empty AIResponse if it does
if self.short_path(message, limit):
await self._persist_history()
return AIResponse(None, False, None, None, None, False, False)
# Number of retries for sending the message; failed attempts are
# spaced by exponential backoff (ENV-12 / D1)
retries = 3
backoff = exponential_backoff(max_delay=10)
async def failed_attempt() -> None:
nonlocal retries
retries -= 1
if retries > 0:
await asyncio.sleep(next(backoff))
while retries > 0:
# Get the message queue
messages = self.message(message, limit)
logging.info(f"try to send this messages:\n{pp(messages)}")
# Attempt to send the message to the AI
answer, limit = await self.chat(messages, limit)
if answer is None:
await failed_attempt()
continue
# Attempt to parse the AI's response (strict — ENV-18)
response = self._parse_answer(answer)
if response is None or not await self.handle_picture(response):
await failed_attempt()
continue
# Post-process the message and update the answer's content
answer_message = await self.post_process(message, response)
answer["content"] = str(answer_message)
# Update message history; persistence runs off the loop (PER-05)
self.update_history(messages[-1], answer, limit, message.historise_question)
await self._persist_history()
logging.info(f"got this answer:\n{str(answer_message)}")
# Update memory
if answer_message.answer is not None:
await self.memoize(message.user, "assistant", message.message, answer_message.answer)
# Return the updated answer message
return answer_message
# Raise an error if the process failed after all retries
raise RuntimeError("Failed to generate answer after multiple retries")