최종 갱신: 2026-04-11 세션 19 (코드베이스 기준)
SSOT: docs/reference/dataModelRevised.md
기능
상태
비고
Project CRUD
done
list/create/get/hide(soft-delete) + 같은 경로 복원
Conversation CRUD
done
list/create/get/delete + custom_label rename
Message CRUD
done
list/create/append/update_status/delete_pair
Message FTS5 search
done
messages_fts (v15) + 트리거 동기화 + search_messages command + CenterPanel SearchBox
Branch 기본
done
list/create/adopt/delete/rename + parent_branch_id + git_branch
Branch depth navigation
done
드로어 네비게이터, breadcrumb, VS Code 트리
Memo CRUD
done
list/create/delete + branch_brief
Artifact CRUD
done
list/create/update_status/delete + subtask link + provenance
Skill 로딩
done
~/.tunaflow/skills/ 스캔 + vendor 그룹핑 + 검색/필터
DB migrations v1-v46
done
v26 plan.slug → v46 agent_jobs background worker. 세부 버전 매핑은 §DB 스키마 표 참조
HTTP API (axum, localhost:19840)
기능
상태
비고
Health / Projects / Conversations / Messages
done
GET, Bearer token auth
Plans / Plan Events / Artifacts / Agents Status
done
GET, read-only
Create Project / Conversation / Branch
done
POST
Send Message (dryRun + agent exec)
done
POST, background agent via spawn_blocking
Delete Conversation / Branch
done
POST/DELETE
Branch Archive / Adopt / Rename
done
POST, adopt summary = all assistant messages
Roundtable Run
done
POST, sequential multi-engine, spawn_blocking
RT Cancel
done
POST, CancelRegistry integration
WebSocket Events
done
agent:completed, agent:error, roundtable:* bridge
Auth
done
UUID v4 Bearer token (per startup)
Multi-Agent (5-engine parity — UI, + OpenCode backend-only) UI ENGINE_CONFIGS 기준으로 5 엔진 디스패치 (claude / codex / gemini / ollama / lmstudio). OpenCode 는 백엔드(start_opencode_run, roundtable executor, skills/identity) 는 완성돼 있으나 프론트 ENGINE_CONFIGS 맵에는 등재 안 돼 있음 → 사용자 노출 경로 없음. 복귀 시 src/lib/engineConfig.ts 에 한 줄 추가하면 활성화.
기능
상태
비고
Claude send + stream
done
ContextPack, resume token, background start_claude_stream
Codex send + stream
done
normalized prompt, JSONL synthetic streaming, background start_codex_run
Gemini send + stream
done
normalized prompt, stream-json, background start_gemini_stream
Ollama send + stream
done
OpenAI-compat (start_openai_compat_stream), UI 디스패치 O
LM Studio send + stream
done
OpenAI-compat (start_openai_compat_stream), UI 디스패치 O
OpenCode send (backend)
done
start_opencode_run, opencode models discovery, roundtable executor 연결됨 — UI 디스패치 X (ENGINE_CONFIGS 미등재)
Roundtable run/followup
done
Sequential + Deliberative, background execution. RT 참가자로는 OpenCode 포함 가능
Unified send factory
done
runtimeSlice.sendWithEngine() — ENGINE_CONFIGS map, 중복 제거
Identity framing
done
## Identity 블록, profile/engine/persona 3층 분리, 혼합 표현 금지
Message author attribution
done
per-message author 태그 [assistant:Profile (engine)]
Agent cancel
done
CancelRegistry thread-aware, stream/background 모두 지원
기능
상태
비고
Normalized prompt assembly
done
build_normalized_prompt_with_budget(), 전 엔진 공용 (claude/codex/gemini/ollama/lmstudio + opencode)
Mode system (Lite/Standard/Full)
done
mode-specific ModeProfile caps + resolution
Auto mode heuristic
done
신호 기반 scoring → profile 자동 선택 + trace reason 기록
Context budget control
done
Settings UI (mode selector + total cap slider) + appStore + backend override
Identity + Persona injection
done
## Identity + ## Persona, persona_fragment 주입
Recent context with authors
done
load_recent_messages_with_author, author attribution 태그
Conversation retrieval (FTS5)
done
pair/anchor/brief chunk 재조립, scoring + dedup + overlap suppression
Compressed conversation memory
done
v17 conversation_memory, 12+ msg threshold, 구조화 요약
Memory status model
done
not_generated/fresh/stale + provenance (source_count, timestamps)
Unified memory policy
done
priority 고정 + budget-aware fallback + overlap suppression
Threshold tuning
done
mode-aware threshold (retrieval/compressed)
Section budget breakdown
done
per-section chars 기록 → contextHash JSON → TracePanel 표시
rawq post-processing
done
confidence filter(0.4+), dedup(±5줄), 300자 snippet, scope/confidence 표시
rawq multi-resolution
done
top 2 full, next 2 skeleton, rest one-line reference
Import block folding
done
3+ 연속 import → [N imports folded]
Jaccard block folding
done
cross-session 유사 블록 접기 (0.8+ threshold)
Markdown lightening
done
bold/italic 제거, 코드 블록 보존
Typed compression
done
section별 압축 목표 (context=800, cross-session=600, findings=400)
context metadata
done
trace_log에 context_mode/sections/length/truncated 기록 (전 엔진 공통)
Trace visibility
done
TracePanel pills (active/skipped), mode badge, budget breakdown, RuntimeStatusBar hint
기능
상태
비고
CLI integration
done
health/search/get, source policy (bundled/local/private only)
Settings UI
done
검색 input, 결과 리스트, 문서 미리보기
Explicit handoff
done
Copy / Send to Context / Save as Artifact
Agent Profiles & Personas
기능
상태
비고
Profile CRUD
done
engine/model/personaId/defaultSkills, appStore persistence
ProfileSelector
done
드롭다운 + Custom fallback
Persona 7종 built-in
done
General/Reviewer/Tester/Architect/Implementer/Debugger/UX Critic
Persona Settings UI
done
priorities/behaviors/constraints/tone/outputStyle/promptFragment
Runtime persona binding
done
promptFragment → ## Persona section, 전 엔진 공통
Applied config visibility
done
message.persona DB 저장 → MessageMeta 표시
기능
상태
비고
Plan CRUD
done
create/get/list/update_status/delete + subtask CRUD
Plan → ContextPack
done
active plan 요약 주입
Plan → Artifact link
done
artifacts.subtask_id
Evaluation backend
done
eval_runs + eval_results, run_eval_agent (실제 에이전트 실행)
Evaluation UI
done
CreateRunForm, ExecuteButton, result cards, cancel/retry
기능
상태
비고
Save as Artifact
done
MessageActions + RT 카드 → SaveArtifactDialog
Artifacts 탭
done
필터/정렬 + 통합 리스트
Artifact 상세 모달
done
content + status + copy/forward/delete
Provenance
done
source conversation/branch/RT + jumpToSource
기능
상태
비고
trace_log write
done
모든 엔진 + RT + context metadata
OTel span metadata
done
trace_id/span_id/parent_span_id/operation/engine/duration_ms/status
Context metadata in trace
done
context_mode/sections/length/truncated + section sizes (contextHash)
Memory policy trace
done
active/skipped pills, budget breakdown, Auto reason 표시
Compressed memory status
done
TracePanel Brain 아이콘 + not_generated/fresh/stale badge
usage_status
done
exact/unavailable/unknown per engine (v16)
Trace export (JSON)
done
list_traces + export_traces_otel
기능
상태
비고
Linear-inspired layout
done
Sidebar + CenterPanel 5-tab + RuntimeStatusBar
CenterPanel 5-tab
done
Chat/Plan/Artifacts/Review/Test
Settings 4-section (분리)
done
settings/ 폴더: AgentsSection, PersonasSection, RuntimeSection
Project-first startup
done
ProjectStartup 화면, projectless chat 불가
주 모니터 중앙 배치
done
primary_monitor() 기준, window-state 복원
Branch 드로어
done
모든 branch/RT는 드로어, 드로어 네비게이터
RT 드로어 렌더링
done
BranchThreadPanel → RoundtableView
FTS 검색 UI
done
CenterPanel SearchBox + 결과 드롭다운 + branch 결과 → 드로어
사이드바 트리
done
VS Code 스타일, status dot, 삭제/인디케이터 위치 정리
Markdown/CodeBlock
done
react-markdown + remark-gfm + 15줄 collapse + copy + lang label
FileViewer
done
inline 경로 감지 + 모달 preview
기능
상태
비고
rawq sidecar
done
daemon mode, background indexing, RawqIndexing guard
OpenCode discovery
done
opencode models CLI + ~/.opencode/bin/ 경로
Gemini discovery
done
npm root -g 기반
Codex discovery
done
~/.codex/models_cache.json
window-state
done
CloseRequested 시 명시적 save
Vite watch ignore
done
docs/**, README*.md, CLAUDE.md 제외
기능
Claude
Codex
Gemini
OpenCode
Normalized ContextPack
O (system prompt)
O (inline)
O (inline)
O (inline)
Identity framing
O
O
O
O
Author attribution
O
O
O
O
Skills injection
O
O
O
O
Plan/Findings/Artifacts
O
O
O
O
rawq code context
O
O
O
O
Retrieval (FTS5 chunks)
O
O
O
O
Compressed memory
O
O
O
O
Cross-session context
O
O
O
O
Thread inheritance
O
O
O
O
Continuation
O (native resume)
O (context replay)
O (context replay)
O (context replay)
Streaming
O (native)
O (JSONL synthetic)
O (native)
partial (progress only)
Background execution
O
O
O
O
Token/cost tracking
exact
exact
streaming exact
N/A
Model discovery
fallback only
cache file
npm node
CLI opencode models
영역
테스트 수
도구
Rust unit
84
cargo test --lib
Frontend smoke/integration
96
vitest + jsdom
Total 180
버전
내용
v1
Core tables
v2
resume_token
v3
plans + plan_subtasks
v4
artifacts.subtask_id
v5
eval_runs + eval_results
v6
trace_log OTel columns
v7
plan_subtasks agent ownership
v8
branches.mode
v9
branches.subtask_id
v10
agent_jobs
v11
trace_log context metadata
v12
(reserved)
v13
projects.hidden
v14
branch shadow conversation fix
v15
messages_fts FTS5 + triggers
v16
usage_status (trace_log + conversations)
v17
conversation_memory
v18
plan_events + plans 6개 컬럼 확장
v19
plans.version_major/minor
v20
plans.revision
v21
session_links + conversation_memory topic/provenance/model_used
v22
conversation_chunks (벡터 임베딩 BLOB)
v23
trace_log.message_id
v24
plan_subtasks.depends_on + parallel_group
v25
plans.parent_plan_id
v26
plans.slug
v27
failure_lessons (FTS5)
v28
artifacts.plan_id
v29
insight_sessions / findings / reports
v30
vec_chunks (sqlite-vec vec0)
v31
document_edges + document_index_status (doc graph)
v32
bge-m3 migration (vec_chunks 384→1024dim + embed_model 컬럼)
v33
projects.meta_conversation_id + onboarding_done (meta agent)
v34
insight_findings.review_branch_id
v35
trace_log.cache_read_tokens + cache_creation_tokens
v36
defensive repair (v35 cache 컬럼 누락 idempotent 재적용)
v37
per-project conventions sync toggle
v38
meta_notifications 테이블
v39
conversation_chunks 스테일 row 정리
v40
branches.adopt_message_id (mobile δ-Branch)
v41
ws_event_log 테이블
v42
meta_notifications 복구 (v38 부분 실패 케이스 대응)
v43
agent_session_audit 테이블
v44
query_cache 테이블
v45
messages_fts standalone FTS5 (external content 제거) + content_tokenized
v46
agent_jobs background worker 컬럼 3건 (metaAgent Phase 4)
현재 우선순위는 CLAUDE.md §11 다음 우선순위 참조 (세션 단위 갱신). 본 문서는 기능별 구현 현황 SSOT 에 집중하며, 로드맵은 단일 소스로 통합됐음.
머지된 P0/P1/P2 항목 이력은 docs/reference/sessionHistory.md 의 세션별 기록 참조.