Repository navigation
Releases
Releases · alesierraalta/tsp
tsp 0.9.1
What's Changed
- docs: list TSP in the Gentle Hub community catalog with the Gentle-AI banner by @alesierraalta in #441
- fix(bench): a renamed build under test is still the tsp the agent runs by @alesierraalta in #449
- fix(skills): no skill cites a reference file that is not shipped by @alesierraalta in #450
- fix(plan): gaps names every owner of a layer without stray quote marks (#455) by @alesierraalta in #456
- bench: u*, w* and x* corpora with planted defects at three scales (#442, #447, #451) by @alesierraalta in #454
- docs(measurements): surface inventory is flat to worse at three scales (#446) by @alesierraalta in #452
- chore(release): 0.9.1 by @alesierraalta in #458
Full Changelog: v0.9.0...v0.9.1
tsp 0.9.0
What's Changed
- test(pi): prove in a live session that a Gentle session runs TSP without a declared gate (#418) by @alesierraalta in #421
- feat(gate): refuse the stop once when an edited test is unrecorded by @alesierraalta in #422
- feat(bench): record whether the nudge made the run record by @alesierraalta in #423
- feat(pi): keep the record nudge marker on the branch by @alesierraalta in #424
- docs(bench): record deviation D9 before the registered Haiku rerun (#416) by @alesierraalta in #417
- fix(plan): run witnesses from the repository root, and four defects the TSP run pinned by @alesierraalta in #426
- fix(plan): give every Findings row a witness that can be run by @alesierraalta in #428
- feat(bench): the chain arm pins another review's findings by @alesierraalta in #429
- fix(pin): run the witness from the repository root by @alesierraalta in #431
- fix(bench): name the plan's path in the chain arm's instructions by @alesierraalta in #432
- docs(bench): the chain arm's measurement by @alesierraalta in #433
- ci: shard the suite into three and answer merge groups by @alesierraalta in #434
- fix(pi): show the debt on the status line while the offer is open by @alesierraalta in #436
- ci: run the corpus fixture check only when the corpus can change by @alesierraalta in #438
- chore(release): 0.9.0 by @alesierraalta in #439
Full Changelog: v0.8.1...v0.9.0
tsp 0.8.1
What's Changed
- feat(plan): evidence-based security candidate dispositions (#359) by @alesierraalta in #365
- feat(admit): bound sandboxed evidence resources (#362) by @alesierraalta in #364
- feat(plan): independent security candidate reviews (#361) by @alesierraalta in #367
- docs(skills): boundary-specific security guidance (#363) by @alesierraalta in #368
- feat(bench): Gentle AI native review arm and --as-change (#370) by @alesierraalta in #373
- feat(bench): score review findings with the plan matcher (#371) by @alesierraalta in #374
- docs(testing): pre-register the RDD vs TSP protocol (#372) by @alesierraalta in #375
- feat(bench): empty-base changes for controls and audits (#372) by @alesierraalta in #376
- test(pi): prove the TSP offer in a live Pi session (#83) by @alesierraalta in #377
- fix(bench): end escaped descendants of every bounded run (#378) by @alesierraalta in #380
- fix(bench): fail a gentle-review run that never captured a reviewer (#381) by @alesierraalta in #382
- docs(testing): RDD vs TSP study results (#372) by @alesierraalta in #383
- feat(plan): refuse a report that names a defect the plan never recorded (#384) by @alesierraalta in #389
- fix(bench): give claude bench runs a HOME of their own (#390) by @alesierraalta in #391
- fix(bench): keep claude bench runs to their own settings and memory (#392) by @alesierraalta in #393
- docs(testing): TSP on Luna rerun and the study erratum (#388) by @alesierraalta in #394
- feat(plan): refuse a witness that passes on the code it is said to pin (#385) by @alesierraalta in #395
- feat(skill): read the change's diff before sweeping the repository (#386) by @alesierraalta in #397
- feat(tsp): add killmatrix command and an auto-run toggle for required runs by @alesierraalta in #396
- feat(pi): ask a run that pinned a test to record its finding (#398) by @alesierraalta in #403
- test(pi): prove the recording nudge in a live session, and see tests (#398) by @alesierraalta in #405
- feat(plan): record whether a defect came with the change (#402) by @alesierraalta in #406
- feat(plan): a finding may carry its reproduction, and check replays it (#399) by @alesierraalta in #407
- feat(plan): take the request as input, and hold criteria to rows (#400) by @alesierraalta in #408
- feat(pin): hold another reviewer's findings to a test that fails (#401) by @alesierraalta in #409
- feat(bench): ask a run once to record what it found (#398) by @alesierraalta in #412
- docs(testing): the record nudge closes the detection gap (#398) by @alesierraalta in #413
- feat(bench): a reviewer arm that runs on any model (#410) by @alesierraalta in #414
- docs(testing): reviewing versus pinning on one model, without Claude (#410) by @alesierraalta in #415
- fix(pi): pass the Gentle signal to tsp check so a Gentle session is required to run TSP (#418) by @alesierraalta in #419
Full Changelog: v0.8.0...v0.8.1
tsp 0.8.0
What's Changed
- bench: record gpt-6-luna reading on the 35 real-bug cases by @alesierraalta in #349
- fix: read a turn's shell changes and commits from the repo by @alesierraalta in #350
- fix: a findings registry that cannot be read is reported, not read as none by @alesierraalta in #351
- feat: the Pi TSP offer keeps its state, reaches headless sessions, and shows what is owed by @alesierraalta in #352
- feat: the Pi host is visible to doctor, repair and uninstall by @alesierraalta in #353
- fix: judge a JavaScript suite at base, and keep the record's ignored witnesses runnable by @alesierraalta in #356
- fix: read a vitest json report from the file it writes by @alesierraalta in #357
- chore(release): 0.8.0 by @alesierraalta in #358
Full Changelog: v0.7.6...v0.8.0
tsp 0.7.6
What's Changed
- feat(skill): state the invariants a run is held to by @alesierraalta in #332
- feat(doctor): report skills modified since install by @alesierraalta in #333
- fix(bench): rescore runs of source-referenced cases by @alesierraalta in #334
- bench: record gpt-6-luna and real-bug readings on r01-r05 by @alesierraalta in #335
- feat(skills): run TSP on every change to tsp itself by @alesierraalta in #336
- fix(feedback): keep code references when generalizing stored text by @alesierraalta in #337
- feat(plan): take the next free id when add-finding or add-evidence gets none by @alesierraalta in #338
- ci: run plan check on the repository plan by @alesierraalta in #339
- fix(check): ask for one TSP run per branch even when the change brought a test by @alesierraalta in #340
- fix(pi): ask for feedback only when the TSP run has not recorded one by @alesierraalta in #341
- chore(release): 0.7.6 by @alesierraalta in #344
Full Changelog: v0.7.5...v0.7.6
tsp 0.7.5
What's Changed
- docs(bench): a case without a corrected file measures nothing by @alesierraalta in #323
- feat(skill): add measure-change to decide the reading a change owes by @alesierraalta in #324
- feat(test-strategy): refuse a rejection and a fake as evidence by @alesierraalta in #326
- docs(measurements): record the h-case control and what it cannot say by @alesierraalta in #328
- fix(bench): the gate refuses a baseline from another instrument by @alesierraalta in #329
- fix(skill): leave the plan the reading measures by @alesierraalta in #330
- chore(release): 0.7.5 by @alesierraalta in #331
Full Changelog: v0.7.4...v0.7.5
tsp 0.7.4
What's Changed
- docs: explain what counts as a defect witness by @alesierraalta in #319
- chore(ci): run the bench on the operator's machine, not in CI by @alesierraalta in #320
- feat(bench): record the index in the history a run appends by @alesierraalta in #321
- chore(release): 0.7.4 by @alesierraalta in #322
Full Changelog: v0.7.3...v0.7.4
tsp 0.7.3
What's Changed
- feat(findings): remember the defect classes a repository has already paid for by @alesierraalta in #318
Full Changelog: v0.7.2...v0.7.3
tsp 0.7.2
What's Changed
- feat(quality): dead code on changed lines, and an arch corpus that says what it proves by @alesierraalta in #309
- fix(bench): the nightly run uses pi, not an invented API key by @alesierraalta in #310
- bench: record the baseline the nightly gate compares against by @alesierraalta in #311
- fix(bench): CI authenticates pi with the provider keys pi already reads by @alesierraalta in #312
- feat(findings): a TSP loop that fixes, re-attacks and stops by itself by @alesierraalta in #313
- feat(bench): families, a readiness index, and a probe flow that covers a matrix by @alesierraalta in #314
- fix(findings): judge each round on the code it actually attacked by @alesierraalta in #315
- fix(bench): keep a module-less fixture contract readable by @alesierraalta in #317
- chore(release): 0.7.2 by @alesierraalta in #316
Full Changelog: v0.7.1...v0.7.2
tsp 0.7.1
What's Changed
- fix(check): a commit range can carry a waiver by @alesierraalta in #296
- feat(bench): a control only counts when the run did the work by @alesierraalta in #297
- feat(sync): the Gemini CLI and the Codex CLI fire the gate by @alesierraalta in #298
- feat(scan): semgrep reads rules over the same change by @alesierraalta in #299
- feat(lang): one inventory names the stacks and missing tools by @alesierraalta in #300
- fix(mutate): a Python change is a refusal, not a nothing-to-do by @alesierraalta in #301
- feat(arch): TypeScript changes answer to the same layer rules by @alesierraalta in #302
- feat(mutate): Python changes are measured by mutmut by @alesierraalta in #303
- feat(mutate): JavaScript changes are measured by Stryker by @alesierraalta in #304
- feat(bench): a run that measured nothing cannot pass the gate by @alesierraalta in #305
- docs: the bench gate is what a scheduled run needs by @alesierraalta in #306
- refactor: the harness's own complexity, measured by the harness by @alesierraalta in #307
- release: 0.7.1, and the help names what the binary can do by @alesierraalta in #308
Full Changelog: v0.7.0...v0.7.1