email_utils.py · laya

1

"""Email helpers for RL Agent: clean raw emails into a compact state and a ready-made set of email questions.

2

3

Jev-style models lose accuracy on long, noisy state, and RL Agent reads at most max_len (512) tokens,

4

so strip quoted replies, signatures and disclaimers in code before asking questions.

5

"""

6

import re

7

8

_QUOTE_HEADERS = [

9

re.compile(r"^\s*On .{0,300}wrote:\s*$", re.I),

10

re.compile(r"^\s*-{2,}\s*(Original|Forwarded) Message\s*-{2,}", re.I),

11

re.compile(r"^\s*_{8,}\s*$"),

12

re.compile(r"^\s*From:\s.+$", re.I),

13

]

14

_SIGNATURE_MARKERS = [

15

re.compile(r"^\s*--\s*$"),

16

17

re.compile(r"^\s*sent from my (iphone|android|mobile|ipad)", re.I),

18

]

19

_DISCLAIMER = re.compile(r"(confidential|intended (solely )?for the (use of the )?(named )?(addressee|recipient)|"

20

r"if you (have )?received this (e-?mail|message) in error)", re.I)

21

22

23

def clean_email_body(body, max_chars=3000):

24

"""Remove quoted history, signature and legal disclaimer; collapse whitespace; truncate."""

25

text = (body or "").replace("\r\n", "\n").replace("\r", "\n").replace("\\n", "\n")

26

lines = []

27

for line in text.split("\n"):

28

if any(p.match(line) for p in _QUOTE_HEADERS) and lines:

29

break # everything below is the previous thread

30

if line.lstrip().startswith(">"):

31

continue

32

lines.append(line.rstrip())

33

# a sign-off only counts near the end (last 40%, or last 8 lines of a short email) and must be a short line

34

cut = len(lines)

35

for i in range(max(1, min(int(len(lines) * 0.6), len(lines) - 8)), len(lines)):

36

if len(lines[i].strip()) <= 40 and any(p.match(lines[i]) for p in _SIGNATURE_MARKERS):

37

cut = i

38

break

39

lines = lines[:cut]

40

paragraphs = [p for p in re.split(r"\n\s*\n", "\n".join(lines)) if not _DISCLAIMER.search(p)]

41

text = re.sub(r"[ \t]+", " ", "\n\n".join(p.strip() for p in paragraphs if p.strip()))

42

return text[:max_chars]

43

44

45

def email_state(subject, body, sender=None, clean=True, **extra):

46

"""Build the state dict the email questions refer to (`subject`, `body`, optional `from`)."""

47

state = {"subject": (subject or "").strip(), "body": clean_email_body(body) if clean else (body or "")}

48

if sender:

49

state["from"] = sender

50

state.update({k: v for k, v in extra.items() if v is not None})

51

return state

52

53

54

def email_questions(categories=None):

55

"""A default fan-out of email questions. `categories` = {key: description} for your own routing labels."""

56

categories = categories or {

57

"billing": "invoices, payments, refunds", "technical": "bugs, outages, integrations",

58

"sales": "pricing, demos, new purchases", "account": "login, access, profile changes",

59

"hr": "hiring, leave, payroll", "other": "none of the above",

60

}

61

return {

62

        "category": {"type": "choice", "instructions": "Which team should handle the email in `body`?", "criteria": categories},

63

"is_spam": {"type": "noul", "instructions": "Is this email unsolicited spam or bulk marketing?"},

64

        "is_phishing": {"type": "noul", "instructions": "Is this email a phishing or scam attempt to steal money, credentials, or personal data?",

65

"criteria": {"true": "phishing, scam, or fraud", "false": "a legitimate email"}},

66

"urgency": {"type": "score", "instructions": "How urgent is the issue described in `body`?",

67

"criteria": ["no time pressure", "needs attention soon", "blocking issue or hard deadline"]},

68

"needs_reply": {"type": "noul", "instructions": "Does the sender expect a reply?"},

69

"sentiment": {"type": "score", "instructions": "What is the sender's tone in `body`?",

70

"criteria": ["angry or very negative", "negative", "neutral", "positive"]},

71

}

72