kebab/docs/ARCHITECTURE.md

Architecture

kebab 의 내부 구조 — crate 의존성, 디렉토리, 핵심 기술 결정. 사용자 사용법은 README.md, 진척도는 HANDOFF.md, frozen 설계 계약은 docs/superpowers/specs/2026-04-27-kebab-final-form-design.md, 머지 후 발견된 deviation 은 tasks/HOTFIXES.md.

한 줄

Cargo workspace, 함수 호출 기반 모듈러 모놀리스. UI binary (kebab-cli, kebab-tui, 미래 kebab-desktop) 가 facade crate (kebab-app) 만 참조. 도메인 / 파이프라인 / 저장소 / 외부 어댑터가 명확한 boundary 로 분리.

핵심 기술 결정 (lock 됨)

결정	값
언어	Rust 2024 (resolver=3, edition 2024)
repo	Cargo workspace (single repo, 함수 호출 기반 모듈러 모놀리스)
원본 저장	filesystem + blake3 content-addressable copy (대용량은 reference + checksum)
metadata	SQLite + FTS5 (lexical search)
vector	LanceDB (embedded, model 별 분리 table)
Markdown parser	pulldown-cmark
embedding	fastembed-rs (multilingual-e5-small, 384d)
LLM	Ollama HTTP (default gemma4:e4b ─ OCR / caption 와 family 통일. 사용자가 더 큰 variant gemma4:26b 등으로 override 가능)
음성 ASR	whisper.cpp (via whisper-rs) — P8 보류, 시스템 dep brainstorm 후
OCR (image)	Ollama vision LM (default gemma4:e4b) — OcrEngine trait 으로 Tesseract / Apple Vision 등 future swap (HOTFIXES P6-2)
OCR (PDF, v0.20.0+)	Ollama vision LM (default qwen2.5vl:3b) — post-extract enrichment via kebab-app::pdf_ocr_apply (H-1 resolution). DCTDecode-only v1 (FlateDecode/CCITTFax skip + warning). family asymmetry vs image OCR: PoC alnum 94.79% (qwen2.5vl) >> 27% (gemma4:e4b 받침), 본 단계에서 PDF OCR 만 qwen2.5vl.
Image caption	Ollama vision LM, runtime gate image.caption.enabled (default OFF)
PDF parser	lopdf per-page 텍스트 + scanned-page image extract (page_image::extract_dctdecode_page_image, v0.20.0). chunker_version = "pdf-page-v1" 하드코딩 (HOTFIXES P7-3). parser_version = "pdf-text-v1" 보존 (v0.20 OCR 후에도) — provenance event 로 OCR 사용 차별화. force-reingest 가 v0.19 indexed scanned PDF 의 재처리에 필요.
code parser	tree-sitter + tree-sitter-rust / tree-sitter-python / tree-sitter-typescript / tree-sitter-javascript / tree-sitter-go / tree-sitter-java / tree-sitter-kotlin-ng — parser-side (kebab-parse-code), chunker-side 아님 (design §6.3). chunker versions: Rust = code-rust-ast-v1, Python = code-python-ast-v1, TypeScript = code-ts-ast-v1, JavaScript = code-js-ast-v1, Go = code-go-ast-v1, Java = code-java-ast-v1, Kotlin = code-kotlin-ast-v1. ast_chunk_max_lines = 200 상수 고정 (HOTFIXES 2026-05-19 — Chunker trait 이 per-medium config 미노출). Kotlin grammar 은 tree-sitter-kotlin-ng 사용 — bare tree-sitter-kotlin 은 tree-sitter 0.21–0.23 에 고착되어 있어 사용 불가. Tier 2 (p10-2): YAML/k8s → serde_yaml + k8s-manifest-resource-v1 (apiVersion+kind per resource), Dockerfile → dockerfile-file-v1 (whole-file), Cargo.toml/go.mod/.json/.xml/.groovy → manifest-file-v1 (whole-file). Tier 2 chunkers live in kebab-chunk; no tree-sitter grammar needed (structure from file type, not AST). Tier 3 (p10-3): shell scripts (.sh/.bash/.zsh) direct → code-text-paragraph-v1 (blank-line paragraph segmentation + 80-line / 20-overlap line-window for oversize). Same chunker also serves as fallback when Tier 1/2 emit 0 chunks or Err — non-k8s YAML / invalid YAML / AST extractor failures all picked up. symbol = None; lang preserved from input doc. Tier 1 family complete (p10-1D): C (tree-sitter-c, code-c-ast-v1, .c/.h) + C++ (tree-sitter-cpp, code-cpp-ast-v1, .cpp/.cc/.cxx/.hpp/.hh/.hxx). C symbol = function name only; C++ symbol = namespace::Class::method (recursive nesting). .h 가 C++ syntax 만나면 tree-sitter-c parse 실패 → Tier 3 fallback.
1B symbol path	workspace path → module path: Python = dotted prefix (kebab_eval.metrics.compute_mrr), TypeScript/JavaScript = slash-style prefix (src/Foo.Foo.search). Rust 1A-2 는 file-scope nesting 만 (workspace prefix 없음, 비일관 수용 — HOTFIXES 2026-05-20).
TUI	Ratatui + crossterm — P9-1 Library 패널, P9-2/3/4 진행 예정
Desktop	Tauri 2 + pdfjs-dist (native PDF render backend 금지) — P9-5
citation 형식	URI fragment (path#L12-L34 / path#p=12 / path#xywh=0,0,100,50, W3C Media Fragments)
ID 생성	blake3(canonical_json(tuple))[..32] hex
RRF fusion_score	[0, 1] 정규화 — 2 / (k_rrf + 1) 로 나눠 mode 간 비교 가능 (post-merge hotfix)
layout	XDG (~/.local/share/kebab/, ~/.config/kebab/, …)

전체 frozen 설계는 docs/superpowers/specs/2026-04-27-kebab-final-form-design.md 12 sections 참조.

crate 의존성 그래프

그룹 단위 view + 컴포넌트별 상세는 docs/components/.

flowchart TB
    subgraph UI ["UI binary"]
        cli["kebab-cli"]
        tui["kebab-tui"]
        mcp["kebab-mcp<br/>(P9-FB-30)"]
        desktop["kebab-desktop<br/>(P9-5)"]
    end
    app["kebab-app<br/>(facade)"]
    subgraph Ingest ["ingest pipeline"]
        srcfs["kebab-source-fs"]
        pmd["kebab-parse-md"]
        ppdf["kebab-parse-pdf"]
        pimg["kebab-parse-image"]
        paud["kebab-parse-audio<br/>(P8 보류)"]
        pcode["kebab-parse-code<br/>(P10-1A-2 + P10-1B + P10-1C-Go + P10-1C-JK + P10-2 + P10-3 + P10-1D)"]
        chunk["kebab-chunk"]
    end
    subgraph Persist ["persistence"]
        sqlite["kebab-store-sqlite"]
        vector["kebab-store-vector"]
    end
    subgraph Adapters ["traits + adapters"]
        embed["kebab-embed<br/>(trait)"]
        embedlocal["kebab-embed-local<br/>(fastembed)"]
        llm["kebab-llm<br/>(trait)"]
        llmlocal["kebab-llm-local<br/>(Ollama)"]
        search["kebab-search"]
        rag["kebab-rag"]
        nli["kebab-nli<br/>(NLI verifier, fb-41)"]
    end
    eval["kebab-eval"]
    config["kebab-config"]
    core["kebab-core<br/>(domain types)"]

    cli --> app
    tui --> app
    mcp --> app
    desktop --> app

    app --> srcfs
    app --> pmd
    app --> ppdf
    app --> pimg
    app --> paud
    app --> pcode
    app --> chunk
    app --> sqlite
    app --> vector
    app --> embedlocal
    app --> llmlocal
    app --> search
    app --> rag
    app --> eval
    app --> config

    pmd --> core
    ppdf --> core
    pimg --> core
    paud --> core
    pcode --> core
    embedlocal --> embed
    llmlocal --> llm
    rag --> search
    rag --> llm
    rag --> sqlite
    rag --> nli
    app --> nli
    nli --> config
    search --> sqlite
    search --> vector
    search --> embed
    eval --> app

    config --> core
    embed --> core
    llm --> core
    sqlite --> core
    vector --> core
    chunk --> core
    search --> core
    rag --> core
    srcfs --> core
    eval --> core

UI → store/llm/parse 직접 의존 금지. 모든 user-facing 진입은 kebab-app facade 만 통한다 (frozen 설계 §8). kebab-cli 가 --config <path> flag 를 honor 하려면 kebab_app::*_with_config(cfg, …) companion 을 통해 Config 을 명시적으로 thread 하는 패턴 — 자세한 이유는 tasks/HOTFIXES.md 의 --config 항목.

kebab-parse-code 의 외부 tree-sitter grammar crate 의존: P10-1A-2 에서 tree-sitter-rust 추가, P10-1B 에서 tree-sitter-python / tree-sitter-typescript / tree-sitter-javascript 추가, P10-1C-Go 에서 tree-sitter-go 추가, P10-1C-JK 에서 tree-sitter-java / tree-sitter-kotlin-ng 추가, P10-1D 에서 tree-sitter-c / tree-sitter-cpp 추가. 모두 kebab-parse-code 에만 격리 (facade 룰 — UI crate / chunker 가 직접 import 금지). Kotlin 은 tree-sitter-kotlin-ng 사용 (bare tree-sitter-kotlin 은 tree-sitter 0.21–0.23 에 고착 — 사용 불가). v0.18.0+ 부터 kebab-source-fs 는 자체 code_meta 모듈 (lang detect + skip helpers + BUILTIN_BLACKLIST) 을 보유, kebab-parse-code 와 분리 (refactor 2026-05-26). v0.19.0 부터 kebab-parse-md 가 kebab-parse-types (parser intermediate types) + kebab-normalize (CanonicalDocument lift) 두 crate 를 흡수 — 24 → 22 crates, design §3.7b 재작성 (HOTFIXES 2026-05-26).

디렉토리 구조

kebab/
├── README.md                                       # 사용자 첫 stop (사용법 / Quick start / Mermaid)
├── HANDOFF.md                                      # 진척도 (phase status / 다음 task)
├── kebab_local_rust_report.md                      # 최초 설계 보고서 (방향성 + 근거)
├── docs/
│   ├── ARCHITECTURE.md                             # 이 파일
│   ├── SMOKE.md                                    # 로컬 워크스페이스 직접 돌려보는 절차
│   ├── superpowers/
│   │   ├── specs/
│   │   │   └── 2026-04-27-kebab-final-form-design.md  # frozen design (12 sections)
│   │   └── plans/
│   │       └── 2026-04-27-task-decomposition.md    # task 분해 implementation plan
│   └── wire-schema/v1/                             # JSON Schema 7 (citation, search_hit, answer, …)
├── tasks/
│   ├── INDEX.md                                    # phase 인덱스 + component task 트리
│   ├── HOTFIXES.md                                 # post-merge dated fix 로그
│   ├── _template.md                                # task spec 작성 템플릿
│   ├── phase-0-skeleton.md … phase-9-ui.md         # phase epic (high-level)
│   ├── p0/p0-1-skeleton.md                         # component task (1)
│   ├── p1/p1-1 … p1-6                              # (6)
│   ├── p2/p2-1, p2-2                               # (2)
│   ├── p3/p3-1 … p3-5                              # (5 — p3-5 = app-wiring, post-spec 추가)
│   ├── p4/p4-1 … p4-3                              # (3)
│   ├── p5/p5-1, p5-2                               # (2)
│   ├── p6/p6-1 … p6-4                              # (4 — p6-4 = image-ingest-wiring 후속 추가)
│   ├── p7/p7-1 … p7-3                              # (3 — p7-3 = pdf-ingest-wiring 후속 추가)
│   ├── p8/p8-1, p8-2                               # (2 — 보류)
│   └── p9/p9-1 … p9-5                              # (5)
├── crates/
│   ├── kebab-core/  kebab-config/                     # 도메인 + 설정 (P0)
│   ├── kebab-source-fs/                               # 워크스페이스 walk + checksum (P1-1)
│   ├── kebab-parse-md/                                # Markdown frontmatter + blocks + types + ParsedBlock → CanonicalDocument lift (P1-2/3/4 — v0.19.0 흡수)
│   ├── kebab-chunk/                                   # heading-aware + pdf-page-v1 + code-*-ast-v1 (Tier 1) + k8s-manifest-resource-v1 + dockerfile-file-v1 + manifest-file-v1 + tier2_shared (P10-2) + code-text-paragraph-v1 (P10-3) chunker (P1-5, P7-2, P10-1A-2, P10-1B, P10-1C-Go, P10-1C-JK, P10-2, P10-3, P10-1D)
│   │   └── src/
│   │       ├── code_*_ast_v1.rs              # Tier 1 AST chunkers (rust/python/ts/js/go/java/kotlin/c/cpp)
│   │       ├── code_c_ast_v1.rs              # Tier 1 (p10-1D): C top-level fn / struct / enum / union
│   │       ├── code_cpp_ast_v1.rs            # Tier 1 (p10-1D): C++ namespace::Class::method (recursive nesting)
│   │       ├── k8s_manifest_resource_v1.rs   # Tier 2 (p10-2): YAML multi-doc, apiVersion+kind per resource
│   │       ├── dockerfile_file_v1.rs         # Tier 2 (p10-2): whole-file Dockerfile
│   │       ├── manifest_file_v1.rs           # Tier 2 (p10-2): whole-file Cargo.toml / go.mod / .json / .xml / .groovy
│   │       ├── code_text_paragraph_v1.rs     # Tier 3 (p10-3): blank-line paragraph + 80/20 line-window fallback
│   │       └── tier2_shared.rs               # Tier 2 (p10-2): shared oversize fallback + Chunk builder helpers
│   ├── kebab-store-sqlite/                            # SQLite + FTS5 (V001/V002/V003) (P1-6, P2-1, P3-3)
│   ├── kebab-search/                                  # Lexical + Vector + Hybrid retriever (P2-2, P3-4)
│   ├── kebab-embed/  kebab-embed-local/                  # Embedder trait + fastembed adapter (P3-1, P3-2)
│   ├── kebab-store-vector/                            # LanceDB VectorStore (P3-3, P7-3 follow-up)
│   ├── kebab-llm/  kebab-llm-local/                      # LanguageModel trait + Ollama adapter (P4-1, P4-2)
│   ├── kebab-rag/                                     # RAG pipeline (P4-3)
│   ├── kebab-nli/                                     # NLI verifier (mDeBERTa-v3 XNLI, fb-41 PR-9a/9b/9c-1)
│   ├── kebab-eval/                                    # golden query runner + metrics (P5-1, P5-2)
│   ├── kebab-parse-image/                             # ImageExtractor + Ollama OCR + caption (P6)
│   ├── kebab-parse-pdf/                               # lopdf per-page text extractor (P7-1)
│   ├── kebab-parse-code/                              # tree-sitter AST extractors: Rust (P10-1A-2), Python + TypeScript + JavaScript (P10-1B), Go (P10-1C-Go), Java + Kotlin (P10-1C-JK — java.rs + kotlin.rs), C + C++ (P10-1D — c.rs + cpp.rs); chunker lives in kebab-chunk
│   ├── kebab-app/                                     # facade (P0 시그니처 + P3-5/P6-4/P7-3 본체)
│   ├── kebab-tui/                                     # Ratatui shell + Library 패널 (P9-1)
│   ├── kebab-mcp/                                     # stdio MCP server — tools: schema, doctor, search, ask (P9-FB-30)
│   └── kebab-cli/                                     # binary (P0 → 핫픽스로 --config flag wiring 강화)
├── migrations/                                     # SQLite refinery V001/V002/V003
└── fixtures/                                       # 테스트 fixture 트리

외부 AI 통합

--json 플래그 가 모든 명령에 붙어 frozen wire schema v1 (schema_version 항상 포함) 을 출력. 외부 도구는 wire 만 의존하면 됨:

1. Claude Code / Codex skill — 얇은 wrapper (kebab search --json / kebab ask --json 호출). ~50 lines.
2. MCP server — kebab 를 stdio MCP server 로 wrap. 모든 LLM client 가 자동으로 사용.
3. HTTP wrapper — kebab serve --bind 127.0.0.1:7711 (P+, local-only 가치 깨므로 신중).

wire schema 자체는 docs/wire-schema/v1/.

비-목표 (frozen design §11 / §0)

• 다중 사용자 SaaS, K8s 배포, 원격 vector DB
• enterprise RBAC/ABAC, 실시간 협업
• 모든 파일 포맷의 완벽한 parsing
• agent 가 임의로 파일을 수정하는 자동화
• multi-workspace (P+ 후순위)
• LLM-as-judge eval (rule-based must_contain 만)
• visual embedding (CLIP) — P+
• desktop app kebab:// protocol handler — P+