engineering-loop/model-policy.yml at main · AS215932/engineering-loop · 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
version: 1
defaults:
  provider: openrouter
  model: minimax/minimax-m3
  tier: cheap
roles:
  systems_engineer:
    provider: openrouter
    model: moonshotai/kimi-k2.6
    tier: mid
  devops_netops:
    provider: openrouter
    model: moonshotai/kimi-k2.6
    tier: mid
  security_auditor:
    provider: anthropic
    model: claude-sonnet-4-6
    tier: strong
  network_architect:
    provider: anthropic
    model: claude-sonnet-4-6
    tier: strong
  finops_integrity:
    provider: openrouter
    model: moonshotai/kimi-k2.6
    tier: mid
  virtual_lab_chaos:
    provider: openrouter
    model: moonshotai/kimi-k2.6
    tier: mid
  implementation_writer:
    provider: openrouter
    model: moonshotai/kimi-k2.6
    tier: mid
risk_overrides:
  high:
    min_tier: strong
  critical:
    min_tier: frontier
retry_escalation:
  after_failures: 1
  max_tier: frontier
tier_fallbacks:
  cheap:
    provider: openrouter
    model: minimax/minimax-m3
  mid:
    provider: openrouter
    model: moonshotai/kimi-k2.6
  strong:
    provider: anthropic
    model: claude-sonnet-4-6
  frontier:
    provider: openai
    model: gpt-5.5
backends:
  # Executor selection follows the implementation-writer tier, so the same
  # risk/retry escalation that picks a stronger model can also pick a
  # stronger harness. mock = deterministic offline backend (no API keys).
  default: mock
  tiers:
    cheap: mock
    mid: mock
    strong: mock
    frontier: mock
  # To go live, point a tier at a real harness and (optionally) override
  # the argv template; {prompt} and {max_iterations} are expanded:
  # tiers:
  #   strong: claude-code
  # definitions:
  #   claude-code:
  #     command:
  #       - claude
  #       - -p
  #       - "{prompt}"
  #       - --output-format
  #       - json
  #       - --permission-mode
  #       - acceptEdits
  #       - --max-turns
  #       - "{max_iterations}"