repowise-bench/.gitignore at master · repowise-dev/repowise-bench · 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
# Python
__pycache__/
*.py[cod]
*.egg-info/
.venv/
.env

# OS / editor
.DS_Store
.idea/
.vscode/

# Cloned target repositories — fetched on demand by scripts/download_benchmarks.py
/repos/

# Generated repowise indexes — built on demand by the harness on first run.
# Exception: the flask48-v2 full index is committed so the C2 arm reproduces
# without paying doc-generation cost (gpt-5.4-nano docs + openai embeddings,
# flask @ 7ee9ceb). Only the LLM-cost artifacts are tracked — parse/centrality
# caches and job state are regenerable.
/indexes/*
!/indexes/pallets_flask_full/
/indexes/pallets_flask_full/*
!/indexes/pallets_flask_full/.repowise/
/indexes/pallets_flask_full/.repowise/*
!/indexes/pallets_flask_full/.repowise/wiki.db
!/indexes/pallets_flask_full/.repowise/config.yaml
!/indexes/pallets_flask_full/.repowise/lancedb/

# Per-run logs and raw outputs — useful locally, not part of the published artifact
/logs/

# Scratch / working directories
/scratch_*/
/tmp/

# Result archives — only the canonical results/swe_qa_flask48/ and
# results/swe_qa_sklearn48/ directories are tracked. Anything else under
# results/ is treated as ad-hoc and ignored.
/results/*
!/results/swe_qa_flask48/
!/results/swe_qa_flask48/swe_qa.jsonl
!/results/swe_qa_flask48/experiment_meta.json
!/results/swe_qa_sklearn48/
!/results/swe_qa_sklearn48/swe_qa.jsonl
!/results/swe_qa_sklearn48/experiment_meta.json
!/results/swe_qa_flask48_v2/
!/results/swe_qa_flask48_v2/swe_qa.jsonl
!/results/swe_qa_flask48_v2/experiment_meta.json
!/results/health_defect_*/
!/results/health_defect_*/correlation.json
!/results/health_defect_*/joined_data.json
!/results/health_defect_*/defect_counts.json
!/results/calibration/
!/results/calibration/*.json
!/results/archive-pre-t0/
# Top-level aggregate outputs of the tool-comparison scripts — tracked so the
# comparison and resolution-time analyses reproduce from cache.
!/results/codescene_headtohead.json
!/results/codescene_paired_deltas.json
!/results/resolution_time.json
!/results/pr_effort_signals.json
# ICSE paper evidence artifacts — small JSON summaries the paper's numbers cite
# (the canonical wall, effort curve, TOST, and the R&R experiment gates). Tracked
# so every headline figure reproduces from cache; the heavy per-repo inputs stay local.
!/results/error_analysis_21repo.json
!/results/error_analysis_21repo_fixedcuts.json
!/results/within_band_ci.json
!/results/f4_curve.json
!/results/bootstrap_tost.json
!/results/prior_defect_ablation.json
!/results/positive_control.json
!/results/refit_bootstrap_tost.json

# Heavy per-repo extraction caches — regenerable from the repo checkouts, too
# large for git (70+ MB each across the corpus). The derived scorecards and
# per-hunk aggregates above ARE tracked.
/results/health_defect_*/hunk_dataset.json
/results/health_defect_*/hunk_tokens.json

# Local experiment scratch — Ollama modelfiles, smoke logs, ad-hoc dumps
Modelfile.*
logs_smoke*.txt
*_index.log
tool_health_dump.json

# opencode source clone (inspection only, local)
opencode-src/