Release/v0.6.0

[devops-thiago]

No description provided.

[thrillhousebot]

🤖 ThrillhouseBot — changes since the last review

• New findings this round: 2
• Previous findings resolved: 0
• Previous findings still open: 16

[thrillhousebot]

🟡 MEDIUM — Clear directive regex hardcodes @thrillhousebot instead of configured bot logins (medium confidence — verify before acting)

The CLEAR_DIRECTIVE pattern is hardcoded to '@thrillhousebot resolved' (Pattern.compile("@thrillhousebot\s+resolved\b(?!\s*\?)", Pattern.CASE_INSENSITIVE)). The repository convention (project instructions) states bot login(s) are configurable via thrillhousebot.github.bot-logins and that bot identity must be derived from BotIdentity/TriggerDetector, never a hardcoded literal. On a deployment using a different bot login, a maintainer's '@ resolved : — <title>' comment will not match isClearDirective, so the ack is not posted and the threadless-finding clear (#548) silently fails — and because mention detection is driven by the configured login, such comments may not even reach this code. The regex should be built from the configured bot identities rather than a fixed literal.

[thrillhousebot]

🔵 LOW — Patchless files are disclosed as exceeding the review budget

DiffBudgetPlanner now classifies a changed file with no/blank patch (binary or oversized text diff) as a coverage gap and omits it by name: '... there is nothing for the model to read — so packing it would count the file as fully reviewed ... Omit it by name like an unclippable file'. However, the only rendered reason for omitted files on the verdict/disclosure surfaces is 'because the diff exceeded the review budget' / '(omitted — exceeded the review call budget; not analyzed)'. That reason is false for patchless files — no budget increase would make a binary or huge-display-truncated file reviewable. Users are pointed at the wrong knob. Add a distinct 'no patch provided' omission class (or disclose patchless files with a neutral reason) so the reason matches the cause.

[thrillhousebot]

🤖 ThrillhouseBot — changes since the last review

• New findings this round: 1
• Previous findings resolved: 0
• Previous findings still open: 17

[thrillhousebot]

🟡 MEDIUM — withheldMaterialNotice does a linear contains() per changed file — O(n^2) per review

withheldMaterialNotice builds var reviewable = ReviewDiffFormatter.namesOf(ctx.reviewableFiles()); (a List) and then, inside for (var file : ctx.files()), tests reviewable.contains(file.filename()). Both collections are sized by the PR's changed-file list, so the membership test inside the loop is the canonical dimension-5 quadratic shape: O(files × reviewable). A PR at GitHub's file-listing cap (~3000 files) costs millions of list scans on every review, on the ack-thread path. The fix is a hashed membership collection; Set.copyOf(...) is unsafe here because a FileDiff can carry a null filename (the codebase guards that elsewhere via ReviewDiffFormatter.namesContain), so a HashSet is the right replacement — null-tolerant and O(1) lookup.

Suggested change

	var reviewable = ReviewDiffFormatter.namesOf(ctx.reviewableFiles());
	var rows = new ArrayList<String>();
	for (var file : ctx.files()) {
	if (reviewable.contains(file.filename())) {
	var reviewable = new java.util.HashSet<>(ReviewDiffFormatter.namesOf(ctx.reviewableFiles()));
	var rows = new ArrayList<String>();
	for (var file : ctx.files()) {
	if (reviewable.contains(file.filename())) {

[sonarqubecloud]

Quality Gate passed

Issues
83 New issues
0 Accepted issues

Measures
0 Security Hotspots
99.8% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[thrillhousebot]

🤖 ThrillhouseBot — changes since the last review

• New findings this round: 1
• Previous findings resolved: 0
• Previous findings still open: 16

[thrillhousebot]

🔵 LOW — .env.example still lists /add-docs as line-capped though the command now batches by tokens

The updated .env.example comment reads: "# Optional: line cap on single-call diff renders (/add-docs, replies, base comparison," / "# budgeting-disabled review). Token-budgeted reviews and the batched commands — /improve," / "# /describe, /changelog, /generate-tests — ignore it (DiffBudgetPlanner owns coverage by" — i.e. it still names /add-docs as one of the renders the REVIEW_MAX_DIFF_LINES line cap applies to. The README updated in the same PR says the opposite: "Token-budgeted reviews and the batched commands — /improve, /describe, /changelog, /generate-tests, /add-docs — ignore it (the planner owns coverage by tokens)". The implementation agrees with the README: DocGenerationService now extends AbstractPrSuggestionGenerator, plans batches through planBatches, and no longer calls buildDiffStringWithStats for the model (its comment says "The whole-PR render is deliberately absent"), so /add-docs is token-budgeted like its siblings and never consults the line cap. This is the "comment left describing the behavior this very change replaced" class: the .env.example statement is false after this PR. Risk low — no configuration value breaks, but an operator tuning /add-docs coverage against the .env.example reference gets the wrong mechanism.

Suggested change

	# Optional: line cap on single-call diff renders (/add-docs, replies, base comparison,
	# budgeting-disabled review). Token-budgeted reviews and the batched commands — /improve,
	# /describe, /changelog, /generate-tests — ignore it (DiffBudgetPlanner owns coverage by
	# tokens); 0 = off
	# Optional: line cap on single-call diff renders (replies, base comparison,
	# budgeting-disabled review). Token-budgeted reviews and the batched commands — /improve,
	# /describe, /changelog, /generate-tests, /add-docs — ignore it (DiffBudgetPlanner owns
	# coverage by tokens); 0 = off