2.5 KiB
Native CLI acceptance
Acceptance run: 2026-10-06.
Clean build
The committed tree was exported with git archive HEAD into a new temporary
directory and built without network access:
cargo build --release --locked --offline --bin balatro-seed-search
balatro-seed-search 0.1.0
Required workflows
All commands used the release binary and disabled periodic progress output. The
final done summary still went to stderr as designed.
One joker on or before ante 2:
$ balatro-seed-search find-jokers --joker Blueprint --before-ante 2 \
--max-seeds 1000 --max-matches 1 --threads 2 --progress 0
1111111Q rank=25 score=1/1 jokers=Blueprint mode=all before_ante=2
Two jokers with AND:
$ balatro-seed-search find-jokers --joker Joker --joker "Greedy Joker" \
--mode all --before-ante 2 --max-seeds 5000 --max-matches 1 \
--threads 4 --progress 0
11111116 rank=5 score=2/2 jokers=Joker,Greedy Joker mode=all before_ante=2
Two jokers with OR and seed-only output:
$ balatro-seed-search find-jokers --joker Blueprint --joker Brainstorm \
--mode any --before-ante 2 --max-seeds 1000 --max-matches 2 \
--threads 4 --output seeds --progress 0
11111117
1111111Q
Advanced non-joker filter, bounded parallel search, and JSONL output:
$ balatro-seed-search search --filter examples/boss-filter.json \
--max-seeds 100 --max-matches 2 --threads 3 --output jsonl --progress 0
{"query":"filter=examples/boss-filter.json","rank":3,"score":1,"seed":"11111114","total_clauses":1}
{"query":"filter=examples/boss-filter.json","rank":6,"score":1,"seed":"11111117","total_clauses":1}
Native seed inspection:
$ balatro-seed-search inspect 1111111Q \
--filter examples/boss-filter.json --output jsonl
{"ok":true,"seed":"1111111Q","clauses":[{"index":0,"matched":true,"detail":"ante 1 boss = The Hook (wanted The Hook)"}],"matched":1,"total":1}
Quality gates
cargo test --release --all-targets --locked PASS (34 engine, 16 CLI)
cargo clippy --all-targets --locked -- -D warnings PASS
cargo run --release --example verify -- 10000 PASS (0 hard failures)
Known limitation
The regression suite proves deterministic internal behavior and published-pool sanity, but the repository does not contain an external per-seed oracle for a complete bit-for-bit comparison against every Balatro/Immolate version. The README describes this limitation; results should be verified in the relevant game version when that distinction matters.