File size: 3,056 Bytes
fee8744
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
name: EmailTriage-v1
version: 1.0.0
description: |
  Email triage environment for training agents to classify and route emails.
  Agents learn to categorize emails (spam/normal/urgent/billing), route them
  to appropriate teams, and set priority levels based on email content and
  sender attributes.

tasks:
  - name: spam_detection
    description: Binary classification task - identify spam vs legitimate emails
    difficulty: easy
    num_instances: 10

  - name: multi_class_routing
    description: Multi-class classification with team routing
    difficulty: medium
    num_instances: 12

  - name: context_aware_triage
    description: Complex routing with context awareness and escalation
    difficulty: hard
    num_instances: 20

action_space:
  type: object
  properties:
    classification:
      type: string
      enum:
        - spam
        - normal
        - urgent
        - billing
      description: Email classification category
    team:
      type: string
      enum:
        - support
        - sales
        - billing
        - none
      description: Target team for email routing
    priority:
      type: integer
      minimum: 0
      maximum: 3
      description: Priority level (0=low, 1=normal, 2=high, 3=urgent)
  required:
    - classification
    - team
    - priority

observation_space:
  type: object
  properties:
    current_email:
      type: object
      properties:
        email_id:
          type: string
          description: Unique email identifier
        subject:
          type: string
          description: Email subject line
        body:
          type: string
          description: Email body content
        sender_domain:
          type: string
          description: Domain of email sender
        timestamp:
          type: string
          format: date-time
          description: When email was received
        is_vip_sender:
          type: boolean
          description: Whether sender is VIP customer
        sla_hours:
          type: [integer, "null"]
          description: SLA response time in hours (if applicable)
    inbox_state:
      type: object
      properties:
        pending:
          type: integer
          description: Number of emails pending processing
        spam:
          type: integer
          description: Count of detected spam emails
        urgent:
          type: integer
          description: Count of urgent emails
        processed:
          type: integer
          description: Number of emails processed
    step_count:
      type: integer
      description: Current step in episode
    task_name:
      type: string
      description: Name of current task

reward:
  type: number
  minimum: 0.0
  maximum: 1.0
  description: |
    Reward signal for classifier/routing decision.
    Combines partial credit for:
    - Correct classification (40%)
    - Correct team routing (30%)
    - Correct priority setting (30%)

metadata:
  author: Meta Hackathon Participant
  license: MIT
  tags:
    - email-processing
    - classification
    - routing