agentfield/.coverage-gate.toml at main · Agent-Field/agentfield · 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
# Coverage gate configuration for AgentField.
#
# This file is read by scripts/coverage-gate.py and by the Coverage Summary
# GitHub Actions workflow (.github/workflows/coverage.yml). Update it in the
# same PR as any deliberate coverage regression so reviewers see *why* the
# floor changed.
#
# Thresholds follow the convention used by most top-tier open source Python/
# Go/JS libraries:
#
#   * per-surface floor  — absolute minimum any tracked surface may hit
#   * aggregate floor    — absolute minimum for the weighted repo total
#   * max drop           — maximum per-surface regression vs baseline, in pp
#
# The gate is intentionally strict on the *aggregate* and more forgiving
# per-surface, because individual surfaces can legitimately fluctuate as
# code is deleted or refactored.

[thresholds]
# Any single surface below this number hard-fails the gate.
# Current baseline (see coverage-baseline.json): lowest surface is
# control-plane at 87.30%. This floor gives ~1.3 pp of headroom.
# Raise this value in its own PR as surfaces improve.
min_surface      = 86.0

# The weighted aggregate across all tracked surfaces must stay at or above
# this number. This is the value that backs the README coverage badge.
# Current aggregate is 89.01%; 88.5 gives ~0.5 pp of headroom.
min_aggregate    = 88.5

# A PR may not regress any surface by more than this many percentage points
# against the baseline without explicitly updating coverage-baseline.json.
max_surface_drop = 1.0

# A PR may not regress the aggregate by more than this many percentage points.
max_aggregate_drop = 0.5

# Minimum coverage required on lines the PR actually touches (patch coverage).
# Enforced in CI by `diff-cover` against origin/main, see
# .github/workflows/coverage.yml → "Run patch coverage gate". Patch coverage
# is the most effective regression signal in top-tier OSS projects (codecov,
# vitest, rust-lang): aggregates drift slowly, but untested new code shows up
# here immediately.
min_patch = 80.0

[weights]
# Weights used to compute the aggregate. These match the relative source size
# of each surface (lines / statements), so a tiny helper package hitting 100%
# can't move the needle.
#
# Regenerate via: ./scripts/coverage-summary.sh && jq -r '.surfaces[] | [.name, .coverage_percent] | @tsv' test-reports/coverage/summary.json
control-plane  = 24326   # Go statements, control-plane/internal/...
sdk-go         = 3527    # Go statements, sdk/go/...
sdk-python     = 1884    # Python lines, sdk/python/agentfield/...
sdk-typescript = 5540    # TS lines, sdk/typescript/src/...
web-ui         = 41693   # TS lines, control-plane/web/client/src/...

[report]
# Files the gate writes under test-reports/coverage/ for humans and agents.
markdown = "gate-report.md"
json     = "gate-status.json"