forge-loop/forge-loop.yaml at trunk · hadamrd/forge-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
# forge-loop dogfood configuration.
#
# Operator-local overrides belong in `forge-loop.local.yaml` or env vars.
# Keep this file conservative: it should support bounded self-dogfood runs
# without depending on a private machine layout.

repo:
  github: hadamrd/forge-loop
  base_branch: trunk
  worktree_root: /tmp

deploy:
  task: ""

scheduling:
  parallel: 1
  tick_interval_s: 60
  worker_timeout_s: 3600
  max_ticks: 1
  # Operational-convergence axis: the maintenance role runs the GC arcs (epic-sweep,
  # and — once wired — branch/worktree GC + checkout-restore). 0 disabled it, which
  # is how the loop accreted ~120 branches + 17 epics. Run it every N ticks.
  maintenance_every_n_ticks: 5

labels:
  ready: "loop:ready"
  triage: "loop:triage"
  blocked: "loop:blocked"
  risk_gate: "risk:high"

agent:
  provider: codex

worker:
  provider: codex
  model: ""
  thinking: medium
  # Worker environment contract — DECLARE + PREFLIGHT the toolchain so a
  # worker never silently inherits the orchestrator's ambient env (the
  # 2026-06-05 silent-toolchain incident). See forge_loop.worker_env.
  env:
    path_prepend: [".venv/bin"]
    vars: { VIRTUAL_ENV: ".venv" }
    require: [python, ruff, pytest, pyright]
  verify: ["ruff check src tests", "pyright src/forge_loop", "python -m pytest -q"]

po:
  enabled: false
  provider: codex
  model: ""
  thinking: high
  timeout_s: 600

critic:
  provider: codex
  model: ""
  thinking: off

lumen:
  top_k: 3
  test_pattern: "tests/test_*.py"