harbor-bot/config.yaml at main · parsewave/harbor-bot · GitHub

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
# Harbor Bot Configuration
# This file defines whitelisted repositories and check parameters

# Whitelisted repositories - bot will only respond to these repos
whitelist:
  - owner: "laude-institute"
    repo: "terminal-bench-2"

# Polling configuration
polling:
  interval_seconds: 30
  command_prefix: "/bot"
  commands:
    - "check"

# Concurrency settings
concurrency:
  max_concurrent_jobs: 3
  max_jobs_per_pr: 1

# Database configuration
database:
  pool_size: 20

# Check workflow configuration
checks:
  # Basic checks (warnings only, don't block)
  basic:
    instruction_max_words: 500
    solution_max_lines: 1000
    tests_max_lines: 1000
    task_max_size_mb: 10

  # AI Detection via GPTZero
  ai_detection:
    enabled: true
    threshold: 0.70
    files_to_check:
      - "instruction.md"
      - "solution/solve.sh"

  # Oracle/Nop Agent validation
  oracle_nop:
    oracle_required_accuracy: 1.0   # 100% required
    nop_required_accuracy: 0.0      # 0% required
    oracle_attempts: 3
    nop_attempts: 1

  # Similarity check against TB2 tasks
  similarity:
    enabled: true
    tb2_repo_path: "/data/terminal-bench-2"
    tb2_repo_url: "https://github.com/laude-institute/terminal-bench-2.git"
    threshold: 0.80
    fail_fast: false

  # TB Check (harbor tasks check)
  tb_check:
    model: "openrouter/openai/gpt-5.2"
    allowed_failures: 3
    fail_fast: true

  # TB Run Small
  tb_run_small:
    agent: "terminus-2"
    model: "openrouter/openai/gpt-oss-120b:exacto"
    n_attempts: 1
    n_concurrent: 1

  # TB Run Large
  tb_run_large:
    agent: "terminus-2"
    model: "openrouter/openai/gpt-5.2"
    n_attempts: 3
    stop_on_first_success: true

  # TB Debug
  tb_debug:
    model: "openrouter/openai/gpt-5.2"

# Paths configuration
paths:
  runs_dir: "./runs"    # Directory for job output files
  temp_dir: "/tmp"      # Directory for temporary files
  artifacts_repo: "https://github.com/parsewave/harbor-bot-artifacts.git"
  artifacts_branch: "job-artifacts"

# Logging configuration
logging:
  level: "INFO"
  format: "%(asctime)s - %(name)s - %(levelname)s - %(message)s"
  structured: false  # Set true for JSON logging in production