.gitignore

# Rust / Cargo build artifacts
/target/
**/target/
target_*/

# Cargo lock is committed (workspace contains binary crate fsqlite-cli)
# Cargo.lock

# Editor and IDE files
*.swp
*.swo
*~
.vscode/
.idea/
*.iml
.zed/

# OS-generated files
.DS_Store
Thumbs.db
Desktop.ini

# Rust-specific
**/*.rs.bk

# Common accidental compiler output
a.out

# Backup files
*.bak

# Profiling and coverage
*.profraw
*.profdata
lcov.info
tarpaulin-report.html
coverage/

# Raw Linux perf recordings are machine-local and large. The curated
# tests/artifacts/perf policy lives with the artifacts/ rules below.
tests/artifacts/perf/**/*.data

# Benchmarking output
criterion/

# Fuzz corpus and artifacts
fuzz/corpus/
fuzz/artifacts/

# Debug and log files
*.log

# Snapshot testing review artifacts
*.snap.new

# Source PNGs (webp versions are committed)
*.png
!site/spec-evolution/og-image.png
!site/spec-evolution/twitter-image.png

# Python tool caches
.ruff_cache/
__pycache__/

# Claude Code / AI agent ephemeral files
.claude/

# Local analysis artifacts
spec_review_findings.md
COMPREHENSIVE_BREAKDOWN_OF_FRANKENSQLITE_PERFORMANCE_AND_CONFORMANCE_ASSURANCE_PROCESS__CODEX.md
COMPREHENSIVE_BREAKDOWN_OF_FRANKENSQLITE_PERFORMANCE_AND_CONFORMANCE_ASSURANCE_PROCESS__CC.md
#
# Spec-evolution visualization sources are intentionally tracked under
# site/spec-evolution/. Root copies are generated deployment detritus.
/ANALYSIS_OF_SPEC_DOC_DIFFS.md
/frankensqlite_diagram.webp
/frankensqlite_illustration.webp
/frankentui.css
/levenshtein_bytes.wasm
/spec_evolution_classification_v1.json
/spec_evolution_data_v1.json.gz
/spec_evolution_v1.sqlite3
/spec_evolution_v1.sqlite3.config.json
/visualization_of_the_evolution_of_the_frankensqlite_specs_document_from_inception.html

# Local scratch SQLite DBs (never tracked)
storage.sqlite3
storage.sqlite3-*
*.sqlite3.bak
*.sqlite3-wal.bak
*.sqlite3.corrupt-*

# E2E fixture corpus (local only; DB bytes never tracked)
sample_sqlite_db_files/golden/*
!sample_sqlite_db_files/golden/.gitignore
!sample_sqlite_db_files/golden/.gitkeep
sample_sqlite_db_files/working/*
!sample_sqlite_db_files/working/.gitignore

# bv (beads viewer) local config and caches
.bv/

# Spec evolution visualization artifacts (tracked)

# Node / Playwright (local only)
node_modules/
test-results/

# Cloudflare Pages build output
dist/

# Wrangler config (downloaded, not source of truth)
wrangler.toml

# Cloudflare Pages source config lives under site/spec-evolution/.
# Root copies are generated/deployment detritus.
/_headers
/_routes.json

# Agent/compiler ephemeral files
.plan.md
screenshots_for_diagnostics/
rustc-ice-*.txt
reports/
.agent-mail.yaml

# Profiling and benchmark data
perf.data
perf.data.*
flamegraph*.svg

# FastEmbed model caches (downloaded at runtime)
.fastembed_cache/

# Codex build artifacts and ephemeral directories
target-codex*/
.codex-target/
.codex-target-arc/
.codex-local-check/
.codex-bench/
.codex-tmp/

# Benchmark reports (generated by comprehensive_bench)
benchmark_report_*.json
benchmark_report_*.html

# Ephemeral agent scratch files
bead_comment.md
new_version
old_version
visible

# Beads history snapshots
.beads/.br_history/

# Beads lock files (write/sync/bv/br advisory flocks)
.beads/*.lock

# Beads transient metadata
.beads/last-touched

# Multi-agent swarm coordination scratch state
.swarm/
.swarm_marching_orders.md

# Ephemeral files (agent-generated, build artifacts, test detritus)
ci-artifacts/
proptest-regressions/
# These checked-in proptest seeds are intentional regression fixtures.
!crates/fsqlite-btree/proptest-regressions/
crates/fsqlite-btree/proptest-regressions/*
!crates/fsqlite-btree/proptest-regressions/cursor.txt
!crates/fsqlite-parser/proptest-regressions/
crates/fsqlite-parser/proptest-regressions/*
!crates/fsqlite-parser/proptest-regressions/parser.txt
test_meminfo

# Benchmark database files
bench_*.db

# One-off test scripts at repo root
/test_*.rs

# Agent session ephemeral files
SESSION_TODO.md
SESSION_REPORT.md
SESSION_*.md
/verify_*.sh
!/scripts/verify_*.sh
newfile.txt
/repro_*.rs

# Nested crate build artifacts
crates/*/target/

# Archived mailbox state snapshots
**/archived_mailbox_states/
.env
.claude/settings.local.json

# Alternative target directories
target-local/
target-fresh/
target-obs/
target-wasm/
**/target-obs/
**/target-wasm/

# Agent-generated Python scripts (one-off analysis/fix tools)
/find_*.py
/fix_*.py
/get_*.py
/parse_*.py
/test_*.py

# Clippy analysis output
clippy_out.*
clippy_output.*

# MCP config files (contain local bearer tokens)
*.mcp.json

# Artifacts directories (CI/E2E/validation run outputs)
artifacts/
# Perf evidence should stay compact in Git. Run artifacts under
# tests/artifacts/perf/ stay local by default.
#
# bd-oyaig WARNING: because of the blanket rule below, a plain `git add` of
# anything under tests/artifacts/perf/ is SILENTLY SKIPPED — commits that
# cite an artifact path here have shipped with the evidence dropped and no
# error. Any artifact a commit, bead, or README cites MUST be made durable
# with a per-directory unignore block (pattern below, same as the
# crates/*/artifacts blocks) covering its compact provenance files
# (README/verdict/manifest/checksums/rerun script); heavy run data
# (tarballs, raw logs, dumps) stays local.
!tests/artifacts/
!tests/artifacts/perf/
tests/artifacts/perf/**/*
!tests/artifacts/perf/**/
# Tiny replay evidence JSONs with explicit commands are retained; ad hoc
# artifacts remain ignored unless deliberately force-added.
!crates/fsqlite-mvcc/artifacts/
crates/fsqlite-mvcc/artifacts/*
!crates/fsqlite-mvcc/artifacts/bd-2g5.1/
crates/fsqlite-mvcc/artifacts/bd-2g5.1/*
!crates/fsqlite-mvcc/artifacts/bd-2g5.1/txn_slots_e2e_artifact.json
!crates/fsqlite-planner/artifacts/
crates/fsqlite-planner/artifacts/*
!crates/fsqlite-planner/artifacts/bd-1as.4/
crates/fsqlite-planner/artifacts/bd-1as.4/*
!crates/fsqlite-planner/artifacts/bd-1as.4/planner_index_selection_e2e_artifact.json

# One-off test shell scripts at repo root
/run_test.sh

# Test output files
/test_out.txt
/test_*.db
/test_*.db-wal
/test_*.sql
unwraps.txt

# Core dumps and agent config backups
core.*
core
*.mcp.json.*.bak
cline.mcp.json
opencode.json
windsurf.mcp.json
clippy.txt
clippy.json
clippy_results.json
clippy_unwrap.json
clippy_unwrap_clean.txt
check_output.txt
.next_backup*/
storage.sqlite3
storage.sqlite3-shm
storage.sqlite3-wal
storage.sqlite3.bak
storage.sqlite3.corrupt-*
storage.sqlite3-shm.corrupt-*
storage.sqlite3-wal.corrupt-*

# ACFS ephemeral patterns (auto-generated agent artifacts)
codex.mcp.json
cursor.mcp.json
gemini.mcp.json
.opencode.json.*.bak
.windsurf.mcp.json.*.bak
.agent-mail-project-id
*.sqlite3.corrupt*
.beads/*.corrupt*
.beads/*.pre_repair_*
.beads/*.repair_candidate_*
.beads/*.bak-corrupt
tmp_golden/
tmp_rch_probe/

# Beads ephemeral artifacts (recovery, backup, bad/rebuild snapshots)
.beads/.br_recovery/
.beads_recovery_*/
.beads_snapshot_*/
.beads/*.pre_*
.beads/*.backup.*
.beads/*.bad_*
.beads/*.backup_*
.beads/*.rebuild*
.beads/*.reindex_test_*
.beads/*.vacuum_test_*
.beads/recovery*/
# br fsqlite WAL-cert sidecars — per-machine runtime, like -wal/-shm.
.beads/*.db-wal-cert*

# Remote compilation helper local cache. Keep the project transfer contract,
# but ignore every other machine-local RCH file.
.rch/*
!.rch/config.toml
.rch-target/
# Suffixed rch target shards (correctness lane, retained-dml lane, per-job
# job-id + nanos shards) — created by concurrent `rch` cargo invocations.
.rch-target-*/
# The shared target pool is named PLURAL with no separator, so neither
# `.rch-target/` nor `.rch-target-*/` above matches it. It is a symlink into
# /data/tmp/rch-targets-pool.
.rch-targets/
.rch-correctness-target/
.rch-retained-dml-target/

# Session extension binary test artifacts
__CHANGESET__
__PATCHSET__
.codex-target-arc/

# Ephemeral agent-generated scratch files at repo root
.beads_recovery/
/C[0-9]*
/D[0-9]*
/db300_*.toml
librust_out.rlib

# Codex scratch directory — self-documenting "do not commit" marker, appears during codex sessions
.codex_tmp_do_not_commit/
.codex_tmp_do_not_commit

# Oracle / verification evidence dumps (regenerated per run, can be huge)
evidence/

# NTM agent session directories
.ntm/

# Benchmark/test numeric output at repo root (timing values, row counts, etc.)
/[0-9]*

# Bare test binary at repo root
/test
.bench-history/

# Codex CLI ephemeral files and directories
.codex
.codex/

# Heaptrack profiling dumps
heaptrack.*.zst

# Beads recovery artifacts (timestamped directories at repo root)
.beads.recovery_*/

# Per-bead progress reports — belong under docs/progress/, not repo root
/progress.md
/progress_bd-*.md
/bd-*.md
# br doctor per-run artifacts
.doctor/

# Ephemeral run artifacts / stray temp dumps
/tmp_total.txt
tests/artifacts/parity-status/

# Library-updater progress file (machine-local, transient)
claude-upgrade-progress.json
# Agent scratch test files (machine-local, never tracked).
#
# The `zz_` prefix is the swarm's own marker for a throwaway probe: these files
# open with `//! EPHEMERAL bench. Not for commit.` and are deleted once the
# measurement they exist for is recorded. The narrower `zz_scratch_*` pattern
# never matched what the agents actually write (`zz_<subject>_bench.rs`,
# `zz_<subject>_probe.rs`), so thirty of them sat in `git status` indefinitely.
# No `zz_` file has ever been tracked here.
**/zz_*.rs

# === multi-repo ephemeral patterns 2026-08-05 ===
# Root scratch SQLite left by local probes / doctor runs
/state.db
/state.db-shm
/state.db-wal
# One-off probe binary dropped at repo root
/lint_probe
# Beads local machine markers (never canonical tracker state)
.beads/.local_version
.beads/.migration-hint-ts
# WAL certification sidecars (`*.db-wal` does not match beads.db-wal-cert)
.beads/*.db-wal-cert
.beads/*.db-wal-cert-head
# br / cleanup session markers
.last-cleanup
# RCH virtual FS workload workspaces (compile offload scratch)
.rch-vfs-*/

# Empty / scaffolding oracle stubs left ignored (not real e2e cases)
crates/fsqlite-e2e/tests/*_ignored.rs

# Skill-loop / repo-janitor scratch
.skill-loop-progress.md
.skill-loop-progress-*.md
# rustc metadata leftover at repo root (`librust_out.rmeta`); not cargo output.
/*.rmeta