stet v1.3.0
Fixes
stet benchmark reads the repository's .review/config.toml, so it benchmarks the configured provider and model (#32)stet benchmark: the typical-hunk estimate for OpenAI-compatible providers accounts for prompt size (#31)- Comment-only edits in
.jsx files no longer trigger a re-review (#26)
- JS/TS symbol lookup finds repo functions named like built-in methods (
get, set, log, find, ...) when they are called directly or imported; only member calls such as x.get( are skipped (#28)
- CLI tests drain captured output while commands run and close their pipes, so large output can no longer deadlock a test (#29)
Maintenance
test_output.txt is ignored, and the release check fails if one is tracked (#30)- The JS/TS extension list is read-only for callers:
jsts.Extensions() returns a copy (#27)
Upgrade note
- Semantic hunk IDs for
.jsx files change once. Existing sessions re-review .jsx hunks one time after upgrading.
Documentation
- README:
stet benchmark configuration source and end-to-end estimate
docs/review-process-internals.md updated for v1.3.0 (.jsx hunk IDs, JS RAG built-in method handling, jsts.Extensions())
Install
curl -fsSL https://raw.githubusercontent.com/beettlle/stet/main/install.sh | STET_RELEASE_TAG=v1.3.0 bash
