Repository navigation
Expand file tree
/
Copy path.gitignore
More file actions
136 lines (119 loc) · 5.31 KB
/
Copy path.gitignore
File metadata and controls
136 lines (119 loc) · 5.31 KB
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
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
# Runtime campaign state — local saves, never committed
/state/
# Dashboard-play runtime — saves, chat log, move sink, and DM transcripts for games
# launched via scripts/play.sh / worldos-play.command. One dir per game; regenerable.
/play-state/
# QA harness runtime artifacts — sandboxed playtest state, transcripts, logs.
# The harness scripts/rubric/config are committed. Local machine overrides and run
# outputs are ignored; the portable MCP template is qa/qa.mcp.example.json.
/qa/state/
/qa/state-oc/
/qa/transcripts/
/qa/runs/
/qa/*.err
/qa/RRI.json
/qa/smoke*.json
/qa/qa.mcp.json
# AI playtester (issue #324): run artifacts (screenshots, a11y dumps, bug/console/network
# logs, summaries) are regenerable and bulky — never committed (public repo). The harness
# (qa/ui_playtest.sh), palette MCP server (qa/playwright/palette_server.js + package.json),
# persona briefs, and scorer ARE committed; only the run outputs + node_modules are ignored.
/qa/ui_playtest_runs/
# player smoke (#1443): qa/player_smoke.sh run artifacts (screenshots + smoke_result.json) — same
# regenerable-and-bulky rationale as ui_playtest_runs/ above; the runner + driver ARE committed.
/qa/player_smoke_runs/
# agent-play (qa/agent_play.sh) per-run artifacts: session.json, DM stream-json, quest trace,
# dry-run command log. Regenerable + LLM-run output — same rationale as the smoke/playtest dirs.
/qa/agent_play_runs/
/qa/playwright/node_modules
/qa/playwright/_*.js
# The on-demand findings ledger (qa/collect_findings.py output) — a regenerable digest of
# local QA run scores, derived from the ignored qa/transcripts/. The collector script IS
# committed; its output is not.
/qa/findings.jsonl
# HV2 harvest-loop extraction output (qa/export_campaign_artifacts.py) — regenerable per
# campaign from snapshot.json + transcripts. The extractor script and the schema handshake
# (data/library/artifact_schema.json) ARE committed; the extracted artifacts are not.
/qa/artifacts_out/
# HV3/HV5 promotion + auto-nomination runtime state. /qa/nominations.jsonl is the nomination queue —
# hand-bootstrapped for HV3's first batch (PR #1338), now appended to per scored run by HV5's
# qa/closeout.py -> qa/nominate.py (this PR) and consumed by HV3's tools/library/promote.py --batch.
# /library/.promoted.jsonl is promote.py's per-library idempotency marker log. Both are regenerable
# runtime state (from scored runs + qa/artifacts_out), never committed. The promote.py / nominate.py /
# closeout.py / library_lint.py scripts AND the promoted library/ pack itself ARE committed.
/qa/nominations.jsonl
/library/.promoted.jsonl
# Privately-imported, user-owned adventures — NEVER commit (may be copyrighted).
# Only original / CC-licensed content belongs under content/campaigns/.
/content/campaigns/_imported/
# Private/user-owned campaigns kept for personal/local play only (documented _private path).
/content/campaigns/_private/
# Private world seeds for personal/internal play (e.g. Forgotten Realms / post-BG3).
# Loaded by the same IP-agnostic start_world loader; kept out of the root-licensed
# tree so the public repo stays clean. NEVER commit.
/content/worlds/_private/
# Raw wiki-ingest cache (tools/ingest) — fetched source, regenerable; not committed.
/tools/ingest/.cache/
# UI-audit screenshots — reproduce rendered output from _private (BG3 portraits, FR maps,
# city scene art). These are © under the same Fan Content / wiki-image policy as the
# _private dirs above. Regenerable via `qa/owshot.sh <screen> <out>.png 8799`.
/docs/ui-audit/screenshots/*.png
# Python
__pycache__/
*.py[cod]
.venv/
venv/
.uv/
*.egg-info/
# Native macOS app build outputs
/dist/
/macos/WorldOSApp/.build/
/macos/WorldOSApp/.swiftpm/
# TTS models / audio caches (large, regenerated locally)
*.onnx
*.pt
*.bin
# Unity renderer model weights are intentionally committed through Git LFS.
!/extensions/renderers/unity/**/*.onnx
!/extensions/renderers/unity/**/*.pt
.cache/
/state/audio/
# Local operator / agent artifacts. Public repo guidance should live in tracked docs;
# machine-specific Claude/Codex notes stay local.
/CLAUDE.md
.claude/local.json
.claude/settings.local.json
.claude/*.local.json
.claude.local/
.aider*
*.har
*.trace
*.tmp
# OS / editor
.DS_Store
.idea/
.vscode/
*.swp
.claude/launch.json
# fast-gate rotation counter (Tier 1)
/qa/.fast_gate_iter
# fast-gate OPT-IN green-cache (WORLDOS_FASTGATE_CACHE=1) — cached GREEN verdicts keyed on
# (git HEAD + persona + gate args). Pure inner-loop convenience; regenerable; never committed.
/qa/.cache/
# pytest-cov data files (coverage is report-only in CI; the .coverage db is never committed)
.coverage
.coverage.*
servers/*/.coverage
.qa-archive/
# Asset registry (gfx M-C, #1195): ONLY the manifest (registry.json) is committed.
# Any co-located binary assets (.fbx/.glb/.png/.wav/...) are regenerable via the
# gen_recipe rows and live on the Unity box's Assets/ tree, never in git.
/data/asset-registry/**
!/data/asset-registry/
!/data/asset-registry/registry.json
# Local nested worktree checkouts (each is its own full git working tree with its own .git
# file) — never repo content, and left unignored they get scanned as ordinary files by
# tools like GitNexus, ballooning the node count past the embeddings cap.
/WorldOS-worktrees/
# HV1 transient calibration db (final rows land in qa/scores.db artifacts table)
qa/artifact_calibration.db