design §3.7b 의 thin layer (ParsedBlock 류) 가 4 parser 중 1개 (markdown) 만 lift 를 경유하는 현실 — fan-in/fan-out 모두 1 → layer 의미 잃음. kebab-normalize (1097 LOC) + kebab-parse-types (98 LOC) 둘을 kebab-parse-md 로 흡수. 설계: docs/superpowers/specs/2026-05-26-normalize-absorption-spec.md 플랜: docs/superpowers/plans/2026-05-26-normalize-absorption-plan.md HOTFIXES: tasks/HOTFIXES.md 의 2026-05-26 entry (design deviation) - 5 사용 type + 3 forward-declared struct → kebab-parse-md::types module 의 pub explicit re-export. - build_canonical_document + derive_title + warning_agent → kebab-parse-md::normalize module. - 4 hard-coded agent literal (lib.rs:122/128/134/153) + warning_agent body return + tracing target literal 모두 보존 — stage label 일관성. - kebab-app callsite (lib.rs:51 use + :1119 context string) + Cargo.toml 의 2 dep (regular + dead) 제거. - kebab-chunk + kebab-store-sqlite 의 [dev-dependencies] kebab-normalize → 제거 (kebab-parse-md 로 갈음). 통합 test source 의 use shift. - test file 이동 (kebab-normalize/tests/normalize_snapshot.rs → kebab-parse-md/tests/). - workspace Cargo.toml: Hunk (a) members 2 entry 삭제 + Hunk (b) version 0.18.0 → 0.19.0 (frozen contract 변경). - design §3.7b 4-단락 재작성 (원래 intent 보존 + 현재 상태 + 보존된 surface + future re-extraction trigger). - design §8 graph 갱신 (3 edge 제거 + 2 forbidden bullet 의미 갱신 + commentary). - ARCHITECTURE.md crate graph + directory tree mechanical 갱신. - tasks/INDEX.md L169 closure mention + "Future work / deferred" 섹션 신설 (image/pdf normalize integration entry). - tasks/HOTFIXES.md 신규 entry (4-block — design deviation Symptom). - HANDOFF.md cross-link 한 줄. - 3 dead struct (ParsedImageRegion / ParsedPdfPage / ParsedAudioSegment) 는 보존 — v0.20+ image/pdf normalize integration 의 future surface (spec §11). Wire / surface impact: 0건. CLI / TUI / MCP / --json 출력 / config / XDG path / parser_version 모두 unchanged. wire-invisible provenance.events[].agent + tracing target literal "kb-normalize" 도 보존 — old DB row 와 new DB row 의 audit log 일관성. Verification: cargo test --workspace --no-fail-fast -j 1 → 1313 passed / 0 failed (172 result blocks). cargo clippy --workspace --all-targets -j 1 -- -D warnings → 0 warning (5m 46s). cargo metadata --no-deps --format-version 1 | jq '.workspace_members | length' = 22. cargo tree -p kebab-app --depth 2 | grep -E "kebab_(parse_types|normalize)" = 0 줄.
16 KiB
title, source, date
| title | source | date |
|---|---|---|
| KB 작업 단위 인덱스 | kebab_local_rust_report.md | 2026-04-27 |
KB 작업 단위 인덱스
kebab_local_rust_report.md 의 Phase 로드맵을 아키텍처 수준 작업 단위로 분해. 각 task 문서는 독립적으로 착수/검수 가능한 단위.
의존 그래프
P0 ── P1 ── P2 ── P3 ── P4 ── P5
│
├─ P6 (image)
├─ P7 (pdf)
├─ P8 (audio)
└─ P9 (TUI/desktop)
P0P5 는 직렬. P6P9 는 P5 이후 병렬 가능.
작업 단위
| # | 코드 | 제목 | 핵심 산출 crate | 선행 |
|---|---|---|---|---|
| P0 | phase-0-skeleton.md | Workspace 뼈대 + 도메인 계약 | kebab-core, kebab-parse-types, kebab-config, kebab-app, kebab-cli | – |
| P1 | phase-1-markdown-ingestion.md | Markdown ingestion 파이프라인 | kebab-source-fs, kebab-parse-md, kebab-normalize, kebab-chunk, kebab-store-sqlite | P0 |
| P2 | phase-2-lexical-search.md | SQLite FTS5 lexical 검색 + citation | kebab-search (lexical) | P1 |
| P3 | phase-3-vector-hybrid.md | Local embedding + LanceDB + hybrid | kebab-embed, kebab-embed-local, kebab-store-vector, kebab-search | P2 |
| P4 | phase-4-local-llm-rag.md | Local LLM + RAG + grounded answer | kebab-llm, kebab-llm-local, kebab-rag | P3 |
| P5 | phase-5-evaluation.md | Golden query / regression eval | kebab-eval | P4 |
| P6 | phase-6-image.md | 이미지 ingestion (OCR + caption) | kebab-parse-image | P5 |
| P7 | phase-7-pdf.md | PDF text + page citation | kebab-parse-pdf | P5 |
| P8 | phase-8-audio.md | 음성 transcription + timestamp citation | kebab-parse-audio | P5 |
| P9 | phase-9-ui.md | TUI + desktop app | kebab-tui, kebab-desktop | P5 |
| P10 | p10/INDEX.md | Code ingest framework + AST chunkers | kebab-parse-code, kebab-source-fs (code walk) | P5 |
Component task decomposition (per phase)
각 phase 의 component-level 분해. AI sub-agent 1세션 = 1 task 가 sweet spot.
-
P0 — p0/ — 1 component
-
P1 — p1/ — 6 components
-
P2 — p2/ — 2 components
-
P3 — p3/ — 5 components
-
P4 — p4/ — 3 components
-
P5 — p5/ — 2 components
-
P6 — p6/ — 4 components
-
P7 — p7/ — 3 components
-
P8 — p8/ — 2 components
-
P9 — p9/ — 5 components + 도그푸딩 피드백
-
p9-dogfooding 피드백 인덱스 — 사용자가 직접 돌려보며 수집한 UX 잡음 → p9-fb-01 ~ 20 으로 분해 + 구현 20/20 ✅ (2026-05-03 완료)
-
p9-fb-22 cursor mid-string editing + Ask follow-tail (post-도그푸딩)
-
p9-fb-24 status bar + Library header + page scroll (post-도그푸딩)
-
p9-fb-25 config workspace.include 제거 + 지원 형식 가시성 (post-도그푸딩)
-
⏳ fb-32 ~ fb-42: 백로그 only — 미구현 + brainstorm 선행 필요. spec 작성 시 superpowers:brainstorming 부터 시작. status: open. 다른 세션에서 이 그룹 손대기 전 사용자 확인 필요. 번호 = release 순서 — 작은 번호일수록 먼저 작업 (2026-05-06 renumber).
🎯 0.3.x — agent foundation (MCP + introspection) ✅ 완료
- p9-fb-26 ingest 로그 출력 일관성 — ✅ 머지 (2026-05-07)
- p9-fb-27 introspection + structured error wire — ✅ 머지 + v0.3.0 cut (2026-05-07)
- p9-fb-28 agent invocation flags (--readonly / --quiet) — ✅ 머지 (2026-05-07)
- p9-fb-29 HTTP daemon (
kebab serve) — 🚫 deferred (2026-05-07) — fb-30 stdio MCP 가 동일 가치 제공, daemon 복잡도 회피. P+ 재개 trigger 는 spec 참조. - p9-fb-30 MCP server — ✅ 머지 + v0.3.1 cut (2026-05-07)
- p9-fb-31 single-file / stdin ingest — ✅ 머지 + v0.3.2 cut (2026-05-07)
🎯 0.4.0 — agent surface refinement (additive only)
- p9-fb-32 stale doc indicator — ✅ 머지 + v0.4.0 cut 후보 (2026-05-09)
- p9-fb-33 streaming ask (ndjson delta) — ✅ 머지 + v0.5.0 cut 후보 (2026-05-09)
- p9-fb-34 output budget controls — ✅ 머지 + v0.5.0 cut 후보 (2026-05-09)
- p9-fb-35 verbatim fetch — ✅ 머지 + v0.5.0 cut 후보 (2026-05-09)
- p9-fb-36 search filter args — ✅ 머지 (2026-05-10)
- p9-fb-37 trace + stats — ✅ 머지 (2026-05-10)
🎯 0.5.0 — RAG quality (cascade 동반: V00X + reindex)
- p9-fb-38 score semantics — ✅ 머지 (2026-05-10)
- p9-fb-39 retrieval precision 튜닝 — ✅ 머지 (2026-05-10) — eval foundation only, lever 적용 deferred
- p9-fb-39b embedding upgrade — ✅ 머지 (2026-05-10) — multilingual-e5-large default
- p9-fb-40 fact-grounded answer — ✅ 머지 (2026-05-10)
🎯 0.6.0 또는 P+ — reasoning
- p9-fb-41 multi-hop reasoning — ✅ 머지 (v0.18.0, 2026-05-26). 5 sub-PR (PR #176-180) + NLI verification (mDeBERTa-v3 XNLI ONNX). spec:
docs/superpowers/specs/2026-05-25-p9-fb-41-finalize-spec.md. plan:docs/superpowers/plans/2026-05-25-p9-fb-41-finalize-plan.md. - p9-fb-42 bulk multi-query + re-rank hint — ✅ 머지 (2026-05-10) — bulk only, rerank hint deferred
-
P10 — p10/ — code ingest (multi-task, sub-indexed in p10/INDEX.md)
- p10-1A-1 code ingest framework — ✅ 머지
- p10-1A-2 Rust AST chunker — ✅ 머지
- p10-1B Python + TS/JS AST chunkers — 🟡 PR 오픈 (코드 완성, 머지 대기)
- p10-1C-Go Go AST chunker — 🟡 PR 오픈 (v0.12.0,
code-go-ast-v1) - p10-1C-JavaKotlin Java + Kotlin AST chunkers — 🟢 PR 오픈 (v0.13.0,
code-java-ast-v1/code-kotlin-ast-v1) - p10-1D C + C++ AST chunkers — ✅ 머지 (v0.16.0,
code-c-ast-v1+code-cpp-ast-v1) - p10-2 Tier 2 resource-aware — ✅ 머지 (v0.14.0,
k8s-manifest-resource-v1/dockerfile-file-v1/manifest-file-v1) - p10-3 Tier 3 paragraph + line-window fallback — ✅ 머지 (v0.15.0,
code-text-paragraph-v1)
🎯 P10 Dogfooding Feedback (v0.17.0)
도그푸딩 round 2 (2026-05-22) 에서 발견된 follow-up 셋. spec + plan:
docs/superpowers/specs/2026-05-22-korean-trigram-tokenizer-design.md,docs/superpowers/plans/2026-05-22-korean-trigram-tokenizer.md. release: v0.17.0.- PR-A 한국어 trigram FTS5 tokenizer + lexical builder + hint — ✅ 머지 (#159, 2026-05-24).
chunks_fts가 V007 migration 으로unicode61→trigram.lexical.rs::build_match_stringtrigram-aware 재설계 (whole-phrase OR token-AND, 3자 미만 토큰 drop, raw FTS5 mode 유지).SearchResponse.hintadditive 필드 + CLI/TUI 안내. 영어 lexical 도 substring 매칭으로 동작 변경. - PR-B C typedef alias unit + parser_version cascade — ✅ 머지 (#160, 2026-05-24).
type_definition분기 — top-level typedef-wrapped anonymous struct/enum/union 의 alias 이름으로 synthetic unit.PARSER_VERSION code-c-v1→code-c-v2bump + same-workspace_path orphan purge cascade. - PR-C
code_lang_chunk_breakdownadditive wire field — ✅ 머지 (#161, 2026-05-24).schema.v1.stats에 chunk 수 집계 sister 필드 + 기존code_lang_breakdown/repo_breakdownJSON schema description 정정 ("chunk count" 오기재 → "doc count").
v0.17.1 post-dogfood polish (release: v0.17.1):
- PR #162
[models.llm] request_timeout_secsconfig + 권장 모델 가이드 — ✅ 머지 (2026-05-25). 8B+ 모델 CPU 추론 시 5분 hard timeout 회피용 노브. additive serde default + env override + 0-edge doc. README + SMOKE 에 CPU only / ≤16GB RAM ⇒ ≤4B Q4 모델 권장 한 단락. - PR #163 sudo 없이 ollama 설치 + ask --stream 권장 (docs only) — ✅ 머지 (2026-05-25). README + SMOKE 에 tarball + OLLAMA_MODELS env 설치 패턴 + cold start 긴 모델은 progressive 토큰 권고 (p9-fb-33 surface).
v0.18.0 fb-41 multi-hop RAG + NLI verification ship (release: v0.18.0):
- PR #176 PR-9a kebab-nli crate skeleton — ✅ 머지 (2026-05-25).
NliVerifiertrait +NliScoresstruct (XNLI 3-channel: entailment / neutral / contradiction) +OnnxNliVerifierplaceholder. workspace.dependencies 에 ort 2.0-rc.9, tokenizers 0.21 (default-features=false, onig), hf-hub 0.4, ndarray 0.16. - PR #177 PR-9b OnnxNliVerifier ONNX inference + model download — ✅ 머지 (2026-05-25). hf-hub lazy download (XDG
model_dir/nli/<sanitized>) + ortSession::commit_from_file+ tokenizersOnlyFirsttruncation (max_length=512, premise 끝부터 잘림 — hypothesis 보전).--ignoredintegration test 5 cases manual smoke (EN self-entailment / EN unrelated / KR entailment / long premise truncation / empty hypothesis err). - PR #178 PR-9c-1 core types + wire scaffolding — ✅ 머지 (2026-05-26).
RefusalReason::NliVerificationFailed+NliModelUnavailable(serde rename_all snake_case, wire = identical strings).Answer.verification: Option<VerificationSummary>additive minor wire.NliCfg+RagCfg.nli_threshold(default 0.0) + env override.RagPipeline.verifierfield +with_verifierbuilder. wire schemas +docs/ARCHITECTURE.mdMermaid 갱신. - PR #179 PR-9c-2 pipeline integration + mock test + SKILL.md — ✅ 머지 (2026-05-26). ★ 첫 user-visible behavior.
ask_multi_hopstep 8.5 NLI hook (empty answer 가드 +truncate_for_nli+ verifier.score + verification field + refusal 분기) +App::open_with_config의 NliVerifier construction + 5 mock multi-hop tests + SKILL.md NLI 안내 한 단락. - PR #180 PR-9d dogfood retest + HOTFIXES closure + corpus 보존 — ✅ 머지 (2026-05-26). 동일 dogfood corpus 의 S7/S1/S3/S10 multi-hop retest — S7 PR-8 baseline
grounded=true + Adam hallucination→ PR-9nli_verification_failed, nli_score 0.0035(HALLUCINATION FIXED 확정).docs/dogfood/v0.18.0/신규 — sanitized SUMMARY + 4 sample wire JSON 보존. - PR #181 chore: workspace-wide cleanup + post-PR9 refactor — ✅ 머지 (2026-05-26). v0.18.0 cut 전 마지막 정리.
[workspace.lints.clippy] pedantic = warn+ 의도적 30+ allow (각 rationale inline). 128 files mechanical clippy --fix. OMC teampost-pr9-refactor가 추가 H1 ([models.nli].modelconfig wiring —DEFAULT_MODEL_ID제거 + provider 분기) + H2 (truncate_for_nlistub_hypothesis제거) + H3 (was_truncatedtracing::debug! surface) + D (MCP test flake fix) + E (HOTFIXES cross-link) + 9 new tests (T1-T4). post-refactor dogfood = PR-9d byte-identical (deterministic 확인). system-architect 의 component-level review 결론 = pre-cut nothing, all v0.18.1+ defer (kebab-normalize 흡수 — v0.19.0 closure, see HOTFIXES.md 2026-05-26; Extractor dispatch unification; kebab-source-fs dep lightening 등).
Post-merge 핫픽스
머지 후 발견된 버그들과 그 follow-up PR들은 HOTFIXES.md에 dated 로그로 기록한다. 원래 task spec은 frozen 상태로 두고, post-merge 동작 변경은 HOTFIXES.md를 source of truth로 본다.
Future work / deferred
- v0.20+ image/pdf normalize integration — design §3.7b intent 미구현 (3 dead struct
ParsedImageRegion/ParsedPdfPage/ParsedAudioSegment보존). PR #186 (normalize-absorption) 의 spec §11 (docs/superpowers/specs/2026-05-26-normalize-absorption-spec.md) 참조.
모든 task 공통 규약
- 의존성 경계 (
Allowed/Forbidden) 위반 금지. report §19 참조. - citation 없는 검색 결과 / RAG 응답 금지.
- 원본 파일 파괴 금지. 파생물만 재생성.
- 모든 record 에 version (parser/chunker/embedding/index/prompt) 기록.
- 각 phase 완료 =
cargo check --workspace && cargo test --workspace통과 + 해당 phase 의 완료 조건 CLI 데모 통과.