sigmap 8.32.1 → 8.34.0

package/CHANGELOG.md

@@ -10,6 +10,31 @@ Format: [Semantic Versioning](https://semver.org/)
 
 ---
 
+## [8.34.0] — 2026-09-13
+
+### Added
+- **Lua extractor** (Tier 3) — global and local functions, module-table functions (`function M.name`, `function M:name`), assigned functions, `require` module hints, and LDoc `---` doc comments as first-sentence hints (#540, PR #550) — thanks **@zerone0x**. The contribution shipped its own fixture and expected output, which is exactly what the #588 fixture guard now requires, contributed before that guard existed. Brought up to the disclosure convention it predates: its caps now report the true overflow rather than truncating silently, with the ceiling unchanged
+- Call-graph definitions for **Kotlin and Scala** (#586, PR #602). `extractDefs` returned null for `.kt`/`.scala` and the walk did not collect them, so the graph was empty for both — and an empty graph returns no error, so `--impact` and blast radius silently read zero. That is the failure mode that cost a release in #561, and akka (Scala) sits in the gated JVM corpus, so CI exercised Scala in a way that could never detect it. Adds `jvmBodyRange` for expression bodies (`fun f() = expr`), which Java does not have. On a small Kotlin+Scala tree: 0 symbols / 0 edges → 8 / 2. Scope is pinned by test — same-file calls resolve, cross-file *receiver* calls do not yet, because receiver typing is Java-shaped
+
+### Fixed
+- Generated `context-*.md` splits from a previous `strategy` no longer pollute retrieval (#555, PR #601). Splits are discovered by filename pattern at read time, not by consulting the config, so a file left by an earlier strategy — or by a module since dropped from `srcDirs` — kept being merged into the index and steering every query. On a 524-file Java repo a stale 376 KB split held ranks 1, 3 and 4 with generated entities, one unrelated to the query, while both files implementing the feature fell outside the top 6 — with `sig-index.json` correctly holding zero entries for it. Each strategy now declares what it wrote and the rest is pruned, and the prune is reported rather than silent
+
+---
+
+## [8.33.0] — 2026-09-13
+
+### Added
+- The generated artifact now **says when the token budget left files out**. The `[sigmap] budget: dropped N file(s)` warning only ever went to stderr, which an agent reading the file never sees — on flask that meant 25 of 51 files present with no indication the other 26 existed, indistinguishable from a 25-file repo. `applyTokenBudget` attaches a non-enumerable summary to the array it returns, so all four call sites (full, per-module, hot-cold) gain the footer without a signature change. The wording states that the omitted files are still in the retrieval index and reachable via `sigmap ask`, and a test asserts it does not overstate the loss — replacing a misleading silence with a misleading warning would be no improvement (#587, PR #599)
+- A test fixture for **every** language extractor, plus a guard that fails when one is missing (#588, PR #595). The 8 languages named in the issue were only part of it: the guard immediately found a ninth, `typescript_react` — `.tsx`, every React component, and the extractor that shipped a silent truncation cap in v8.32.0. It also caught a stale `test/expected/vue.txt` that outlived the `vue.js` deleted in v8.32.1, which had been making `--diagnose-extractors` print a silent `SKIP`. That diagnostic went from 21 passing with an unnoticed SKIP to 32 passing with none
+
+### Fixed
+- NestJS route paths now compose the `@Controller` prefix. `@Controller('cats')` + `@Get(':id')` emitted `:id` rather than `/cats/:id`, so the route pseudo-signature matched nothing a user would ask about — defeating the purpose of the feature, which exists so a route-worded query can reach a controller whose signatures never mention the path. The prefix is attributed per controller rather than per file, since a file may declare several. The other six claimed frameworks were verified unchanged (#585, PR #598)
+
+### Changed
+- Extractor resolution has **one source of truth**. Three places decided which extractor module to load and two had drifted: `src/eval/analyzer.js` carried a dead duplicate `.vue` key, and the `--diagnose-extractors` map still pointed at `vue.js` after that module was deleted — which is how an unreachable extractor survived unnoticed in the first place. The hand-maintained copy was also incomplete, with no `.gd` entry, so gdscript was never actually diagnosed despite having both a fixture and recorded expected output. `language-detector.js` and `dashboard.js` are deliberately **not** merged in: they map `.tsx` for language *statistics* and display *labels*, not resolution, and a test pins that distinction so a future dedup cannot quietly break language counts (#591, PR #597)
+
+---
+
 ## [8.32.1] — 2026-09-13
 
 

package/README.md

@@ -64,7 +64,7 @@ That map is exactly what agentic grep is worst at: reproducible, auditable conte
 - **62.9% task-success proxy** — modeled from retrieval tiers, not measured LLM sessions
 - **1.58 prompts per task** — down from 2.84 (44.4% fewer retries, modeled)
 <!--/SM:whyMetrics-->
-- **<!--SM:languages-->32<!--/SM:languages--> languages supported** — TypeScript, Python, Go, Rust, Java, R, and more
+- **<!--SM:languages-->33<!--/SM:languages--> languages supported** — TypeScript, Python, Go, Rust, Java, R, and more
 - **No vendor lock-in** — works with any AI assistant or local LLM
 - **No API costs** — use local models (Ollama, llama.cpp, vLLM) with zero token fees
 - **Full privacy** — keep your code and context on your machine
@@ -122,8 +122,8 @@ Ask → Rank → Context → Validate → Judge → Learn
 
 <!--SM:benchmarkBlock-->
 ```
-Benchmark : sigmap-v8.32-main (21 repositories, including R language)
-Date      : 2026-09-12
+Benchmark : sigmap-v8.34-main (21 repositories, including R language)
+Date      : 2026-09-13
 
 Hit@5          : 78.9%   (grep-agent baseline 44.0%  — 1.73× lift)
 Token reduction: 96.8%   (across 21 repos)
@@ -332,7 +332,7 @@ sigmap --health
 | Benchmark methodology | [benchmark.html](https://sigmap.io/guide/benchmark.html) |
 | Config reference | [config.html](https://sigmap.io/guide/config.html) |
 | Roadmap | [roadmap.html](https://sigmap.io/guide/roadmap.html) |
-| <!--SM:languages-->32<!--/SM:languages--> languages | [generalization.html](https://sigmap.io/guide/generalization.html) |
+| <!--SM:languages-->33<!--/SM:languages--> languages | [generalization.html](https://sigmap.io/guide/generalization.html) |
 
 ---
 
@@ -389,7 +389,7 @@ See [.github/PULL_REQUEST_TEMPLATE.md](.github/PULL_REQUEST_TEMPLATE.md) for the
 
 ---
 
-## <!--SM:languages-->32<!--/SM:languages--> languages
+## <!--SM:languages-->33<!--/SM:languages--> languages
 
 TypeScript · JavaScript · Python · Java · Kotlin · Go · Rust · C# · C/C++ · Ruby · PHP · Swift · Dart · Scala · Vue · Svelte · HTML · CSS/SCSS · YAML · Shell · SQL · GraphQL · Terraform · Protobuf · Dockerfile · TOML · XML · Properties · Markdown · R · GDScript