File size: 2,198 Bytes
3eb9552
4b77608
3eb9552
 
4b77608
3eb9552
4b77608
3eb9552
 
 
4b77608
3eb9552
4b77608
3eb9552
 
4b77608
 
3eb9552
4b77608
 
 
 
3eb9552
 
 
4b77608
 
 
3eb9552
 
4b77608
 
3eb9552
4b77608
 
 
 
3eb9552
4b77608
3eb9552
4b77608
 
 
 
3eb9552
4b77608
 
3eb9552
4b77608
 
 
 
3eb9552
4b77608
3eb9552
4b77608
 
 
 
3eb9552
 
 
4b77608
3eb9552
4b77608
 
 
 
 
 
 
 
3eb9552
 
 
4b77608
 
 
3eb9552
4b77608
 
3eb9552
 
 
4b77608
 
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
# OpenEnv Configuration
# Real-world email triage environment for AI agent training

environment:
  name: "openenv-email-triage"
  version: "1.0.0"
  description: "Autonomous email triage based on intent and urgency."
  authors: ["OpenEnv Team"]
  license: "MIT"
  repository: "https://github.com/yourusername/OpenEnv"
  tags: ["natural-language", "triage", "classification", "email"]
  openenv_version: "1.0"

tasks:
  easy:
    name: "Basic Email Triage"
    description: "Triage 10 simple emails (spam, urgent, or generic)."
    config:
      num_emails: 10
      spam_ratio: 0.3
      urgent_ratio: 0.2
      confounding_ratio: 0.0
    grader:
      success_threshold: 0.7
      criteria:
        - name: "accuracy"
          weight: 0.8
        - name: "critical_safety"
          weight: 0.2
  medium:
    name: "Moderate Triage with Ambiguity"
    description: "Triage 20 emails with some internal ambiguity."
    config:
      num_emails: 20
      spam_ratio: 0.3
      urgent_ratio: 0.2
      confounding_ratio: 0.2
    grader:
      success_threshold: 0.8
      criteria:
        - name: "accuracy"
          weight: 0.7
        - name: "critical_safety"
          weight: 0.3
  hard:
    name: "Complex Dynamics Triage"
    description: "Triage 50 complicated, ambiguous, or noisy emails."
    config:
      num_emails: 50
      spam_ratio: 0.4
      urgent_ratio: 0.1
      confounding_ratio: 0.4
    grader:
      success_threshold: 0.9
      criteria:
        - name: "accuracy"
          weight: 0.6
        - name: "critical_safety"
          weight: 0.4

# Observation space
observation:
  dimensions: 4
  components:
    - name: "emails_remaining"
      description: "Remaining count"
    - name: "is_spam"
      description: "Is the current email marked as spam?"
    - name: "is_urgent"
      description: "Is the current email marked as urgent?"
    - name: "is_internal"
      description: "Is the current email from an internal sender?"

# Action space
action:
  dimensions: 1
  type: "discrete"
  bounds: [0, 4]
  components:
    - name: "action_type"
      description: "0=Ignore, 1=Reply, 2=Forward, 3=Archive, 4=Delete"

# Rendering
rendering:
  mode: ["human"]
  fps: 1