- pptx-wizard: add "🔍 이미지 분석" button in Step 2 image panel
* calls /api/pptx/analyze-image per figure via mistral vision
* shows per-image progress (1/10, 2/10...)
* "제외" caption → dimmed card + filtered from outline prompt
* captions shown below each thumbnail in the grid
- server: POST /api/pptx/analyze-image endpoint
* resolves /api/files/ path to workspace file, thumbnails if >300KB
* calls orchestration.secondary vision model (mistral-large-3:675b-cloud)
* caches results in project/.image-captions.json (skips on re-analysis)
- pptx-wizard: generateOutline() includes captions in imgList
* "제외" images filtered out before sending to LLM
* format: "/api/files/... (figure_p2_1) — 카플란-마이어 생존곡선 3그룹"
- skills-manager: fix getModelOverrideForUser() trigger filtering
* was returning skill model for ALL messages regardless of trigger keywords
* now only overrides when message matches skill triggers (same logic as buildPromptContext)
* presenter model (mistral) now only applies for 발표/pptx/슬라이드 keywords
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- config: primary model → gpt-oss:120b-cloud (faster general chat, 2.72s median)
- server-v2: fix vision regex to exclude gpt-oss (gpt(?!-oss) negative lookahead)
* gpt-oss was incorrectly matching /gpt/ → image data sent to non-vision model
* applied at all 3 _supportsVision check sites (lines 5204, 6362, 7832)
- server-v2: vision fallback routing — when primary lacks vision support and user
message contains images, automatically route to orchestration.secondary
(mistral-large-3:675b-cloud, 4.27s avg, fastest vision cloud model)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- skills-manager: conditional skill injection via triggers frontmatter field
(skills only sent to model when message matches trigger keywords)
- skills: added triggers to all skill SKILL.md files; cleaned up headers/disclaimers
- anthropic-adapter: toAnthropicContent() converts OpenAI image_url → Anthropic image blocks
- multi-agent: vision flag in SecondaryProfile; secondarySupportsVision() checks config
- config: secondary set to mistral-large-3:675b-cloud with vision:true
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
The 💻 코드 sidebar tab now opens a dedicated code.html page in a new tab,
mirroring how the slide wizard opens pptx-wizard.html.
- Add web-ui/code.html: standalone code editor page reusing styles.css and
the same CodeMirror/xterm/Monaco head as the main app.
- Add web-ui/code.js: self-contained code-view logic — inlined shared helpers
(api/authHeaders/escHtml/getCookie) + the extracted code-editor block +
a bootstrap that inits Monaco and loads models/dir. It mirrors the code
block in app.js (kept duplicated so the main app stays untouched).
- index.html: tab-code now window.open('/code.html'). The in-page code mode
(top mode button) is unchanged, so chat→editor live streaming still works.
Verified with Playwright: Monaco initializes, editor/sidebar/toolbar/Coder AI
render, no JS errors.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The slide editor draws a title slide's image as a full-bleed darkened
background, but pptx_gen.py's make_title_slide ignored image_path entirely
and only drew the skin background — so photo backgrounds showed in the
editor but vanished from the generated PPTX / preview.
- make_title_slide now renders image_path full-bleed (cover) with a dark
overlay and white title/subtitle, matching the editor preview.
- Add set_shape_fill_opacity() that injects OOXML <a:alpha>, since
python-pptx's fill.transparency is a silent no-op (the overlay was
rendering as fully opaque black). Apply it to the fullscreen bottom bar too.
- Pass project_dir/workspace_path/warnings into make_title_slide.
- Bump version to 2.8.2.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- Fix pptx-wizard slide editor modal opening off-screen (header clipped)
when a saved drag position is restored on a fresh page load: clear the
CSS centering transform and clamp coordinates into the viewport.
- Add pptx-wizard.html and python-runner.html web UIs.
- Bump version to 2.8.1.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- All skill toggles are now per-user (except multi-agent-orchestrator
which remains global). Admin toggle without a target username defaults
to their own per-user state instead of the global file.
- GET /api/skills defaults to the requesting user's own state (previously
returned global state when no ?username= param was given).
- Add SkillsManager.initUserSkillsState(): creates per-user
skills_state.json from global defaults if missing — called on user
creation and on every login as a safety net for existing accounts.
- Create initial per-user skills_state.json for cherry and jasmine.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
New skill:
- .smallclaw/skills/meteorologist/SKILL.md — 기상 전문가 스킬 (Windy iframe 규칙 포함)
New weather tools (src/tools/weather.ts):
- weather_airpollution: OpenWeather 대기질 (AQI, PM2.5/PM10/O3/NO2)
- weather_openmeteo: Open-Meteo 시간별 예보 (ECMWF/GFS, 키 불필요)
- weather_kma: 기상청 공식 API (초단기실황·단기예보)
- weather_airkorea: 에어코리아 실시간 대기질
- weather_nasa_power: NASA POWER 기후 데이터 (MERRA-2, 키 불필요)
- weather_era5: ERA5 재분석 과거 데이터 (Open-Meteo Historical, 키 불필요)
- weather_cds: Copernicus CDS 정식 API (ERA5·CMIP6 SSP 시나리오)
- weather_cmip6: CMIP6 기후 모델 (Open-Meteo Climate API, 키 불필요)
Scripts:
- scripts/era5_cds_fetch.py — CDS/CMIP6 Python 헬퍼 (ZIP→NetCDF 처리 포함)
web-ui:
- Windy URL → iframe 자동 렌더링 (standalone URL + 마크다운 링크 모두 감지)
- windyToEmbedUrl: windy.com comma 형식 URL 파싱 개선
server-v2.ts:
- BROWSER RULE: "열어줘" 제거 → 오용 방지
- weather context: 9개 도구 선택 기준 추가
- JSON schema 8개 추가
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
These skills only added prompt instructions — no gateway tools or
unique capabilities. The system prompt already covers their domains.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
- Remove blanket regex that stripped ALL image markdown from AI responses,
which was breaking dental-dict images. The [IMAGE COMPLETE] prompt reminder
already tells the model not to duplicate image_edit/image_read results.
- Remove coder skill (SKILL.md)
- Remove dental-dict skill (now handled by MCP server description alone)
- Remove broken external URL for 볼튼분석 (Open-i link dead)
- investor skill enabled in skills_state
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
- Fix settings save bug: primaryModel captured before _modelSettingsLoaded guard → wiped model roles when models tab not opened
- Fix server: filter empty-string roles before saving to prevent role wipe
- Add Voice and Session settings tabs to UI
- Add OpenWeather API key field to Credentials tab
- Add GET/POST endpoints: /api/settings/voice, /api/settings/session, /api/settings/openweather
- Fix Mistral parallel tool calls: model embeds tool calls as text instead of tool_calls array; parseEmbeddedToolCalls() detects and converts
- Add Anthropic provider adapter (anthropic-adapter.ts)
- Weather tool vault key resolution fix
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
사이드바에 💻 코드 탭 추가:
- Monaco Editor (vs-dark/light 테마 자동 연동)
- 파일 탭 관리 (새 파일, 전환, 닫기), Ctrl+S 저장
- 오른쪽 AI 채팅 패널 (Coder 세션, SSE 스트리밍)
- "선택 코드 → AI 전송" 버튼
- "▶ 실행" → 하단 터미널 패널로 명령 전송
- 에디터/AI 패널 수직 드래그 리사이즈
- 스킬 탭 이모지 버그 수정 (조건 반전, 커스텀 이모지 표시 안 되던 문제)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
풀스택 시니어 엔지니어 어시스턴트:
- 도구: python_eval, bash_eval, file_read/write, web_search
- 코드 실행 후 결과 검증 원칙
- 공식 문서 우선 참조 사이트 (MDN, PyPI, React, FastAPI 등)
- 코드 리뷰 체크리스트 (보안→정확성→성능→가독성)
- Python/JS/백엔드/DB/DevOps 언어별 핵심 포인트
- skills_state.json에 coder: true 등록
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- investor skill: TradingView 실시간 주가 차트 렌더링 지원
- ```tradingview SYMBOL|INTERVAL``` 블록 → 차트 위젯 자동 생성
- JSON 형식 출력 시 HTML 언이스케이프 후 파싱 (symbol/interval 추출)
- renderTVChart(): tv.js 동적 로드, 다크/라이트 테마 연동, RSI·MACD 기본 포함
- musician skill: MIDI 다운로드 사이트 4개 추가 (freemidi, bitmidi, piano-midi, midiworld)
- MIDI 탭 닫기 버그: window.open에 noopener,noreferrer 추가
- beforeunload/pagehide 시 midiRouter.stop() 호출
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- music-renderer: saveMidiFile export (JSON, no LilyPond), getRenderFile prefers original.midi, fix self-copy bug
- server-v2: POST /api/music/midi now saves to workspace/music/<name>/ and returns JSON immediately; new SSE endpoint /api/music/midi-score for on-demand LilyPond
- index.html: buildMidiPickerWidget with color-coded buttons (green/purple/amber), MIDI download moved inside player tab, score panel default closed
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- web-ui/midi-editor.html: full piano roll MIDI editor (new tab)
- all tracks simultaneously, color-coded per channel
- note add/delete/move/resize, multi-select, copy/paste
- velocity lane, track panel (mute/solo/instrument), undo/redo
- GM reset + bank select on playback, MIDI output device select
- export to .mid, open via ?url= param or file picker
- web-ui/index.html: add ✏ 편집 button to MIDI widget (opens editor)
- GM SysEx reset + bank select before note playback
- 200ms note delay for synth init time
- getRenderFile: prefer original.midi over LilyPond output.midi
- src/gateway/music-renderer.ts: getRenderFile returns original.midi first
- LilyPond SVG render overwrites output.midi (no Program Changes)
- original.midi preserved from upload, served via /api/music/file/:id/midi
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- session.ts: Replace fixed maxMessages limit with token-based trimHistory;
resolveNumCtx() now queries Ollama API for actual model context length (cached).
maxMessages raised to 2000 as safety ceiling only.
- session.ts: Boot sessions (boot-*) never saved to disk — ephemeral by design.
- session.ts: New sessions not written to disk until first message is added.
- session.ts: Add cleanupEmptySessions() called at startup to auto-remove empty files.
- session.ts: migrateGlobalSessionsToUser() deletes global original after copying to user dir.
- server-v2.ts: Call cleanupEmptySessions() at startup.
- server-v2.ts: User creation API accepts optional telegramId for immediate channel mapping.
- config.json: Remove explicit maxMessages (now uses dynamic token-based default).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Replace piper TTS with edge-tts (Microsoft neural Korean voice)
- piper ko_KR-kss-medium uses pygoruut phoneme type unsupported by C++ binary,
causing Chinese-sounding output; edge-tts solves this via online neural TTS
- Added src/tools/edge_tts_synth.py Python helper script
- Server TTS endpoint now dispatches by provider (edge_tts vs piper)
- Config updated: voice.tts.provider = "edge_tts", voice = "ko-KR-SunHiNeural"
- Add TTS toggle switch to web UI (chat input bar)
- Toggles visibility of all 🔊 buttons and stops active playback
- State persisted in localStorage (default: on)
- Mic input now auto-submits after speech recognition completes
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- STT: Replace MediaRecorder+Whisper with browser Web Speech API (ko-KR)
Whisper base model hallucinated English for Korean speech; Chrome's
built-in Google STT is far more accurate for Korean
- TTS: Skip ffmpeg OGG conversion, return WAV directly from Piper
Avoids OGG/Opus codec compatibility issues in browsers
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- POST /api/voice/stt: Whisper-cpp transcription of webm/ogg audio
- POST /api/voice/tts: Piper Korean TTS returning ogg audio
- Web UI: 🎤 mic button in chat input; click to start/stop recording
MediaRecorder → /api/voice/stt → auto-fills chat input
- Web UI: 🔊 TTS button on each AI message; click to play/stop
- config.json: add absolute model/binary paths for whisper-cli and piper
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Luckysheet modal: 8-direction resize handles (edge + corner), title bar drag-to-move
- First drag/resize converts from flex-centered to absolute positioning
- Minimize/close reset to centered state on reopen
- New xlsx-viewer.html: standalone Luckysheet page for opening xlsx in separate window
- Added "↗ 새 창" button on xlsx cards to open viewer in a resizable popup window
- Viewer shares session cookies so auth works transparently
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Skills:
- Add accountant (CPA) and investor (CFA) skills with site: hints
- Add site: hints to lawyer, psychiatrist, musician skills
- Allow multiple skills active simultaneously (remove exclusive mode)
Tools:
- Add excel_read / excel_write tools (openpyxl-based)
- Fix python_eval packages: retry with --break-system-packages on PEP 668 failure
- Fix email_read: spaces→underscores in filenames, return /api/files/ links,
remove absolute savedPaths from data, send SSE 'files' event for attachments
Server:
- FILE_OP v2: fall through to primary when secondary model unavailable
- Add PUT /api/files/{*filePath} endpoint for xlsx editor save
UI:
- xlsx viewer: Luckysheet modal (lazy CDN load, ~3MB on first use)
with edit mode, 💾 save button, Nanum Gothic font
- Email attachment preview: SSE 'files' event appends file links to final reply
- ✏️ 편집 button (blue) for xlsx cards
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
스캔 PDF에서 텍스트 추출 실패 시 자동으로 OCR 실행.
PyMuPDF로 300DPI 이미지 렌더링 후 tesseract kor+eng 적용.
ocr: true 파라미터로 강제 OCR 가능. method 필드로 추출 방법 표시.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
사용자가 업로드한 PDF는 이미 갖고 있으므로 저장 버튼 불필요.
renderMarkdown/renderLinkButton에 hideDownload 옵션 추가,
유저 메시지 렌더링 시 적용. 미리보기 버튼은 유지.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
스캔 악보 PDF 처리 절차 추가:
pdf_read 시도 → 실패 시 pdf_extract_images로 페이지 추출 → image_read로 시각 분석.
도구 목록에 pdf_extract_images, image_read 추가.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Settings that were silently lost on reinstall:
- Track .smallclaw/heartbeat/config.json (active hours 8-22, interval 30m)
- Track .smallclaw/skills_state.json (multi-agent-orchestrator enabled)
- Fix gitignore: heartbeat/ → heartbeat/runs/ (only ignore run logs, not config)
- Fix gitignore: remove skills_state.json from ignored list
Vault (contains channel tokens, passwords) cannot go in git — instead:
- Add scripts/backup-vault.sh: backs up vault+credentials to ~/.smallclaw-backup/
keeping last 5 snapshots. Run manually or via cron (0 3 * * *).
- Run initial backup now → ~/.smallclaw-backup/vault_20260518_232441
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
When no explicit output path is given and source is JPEG/WebP, output
defaults to PNG (lossless). This avoids cumulative quality loss on
repeated edits (brightness → contrast → filter → ...).
Exception: 'convert' operation keeps the original format so explicit
format-change requests still work. Explicit output path always honored.
Also fix quality default mismatch: TS params was 85, Python was 92.
Now both are 92 for when JPEG/WebP output is explicitly requested.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Add per-user session/workspace paths to .gitignore so runtime data
(sessions, uploads, memory, attachments, pubmed) is no longer tracked
- Narrow ollama_web_search description to prevent it being chosen over
web_search for news/info queries — it's now explicitly a fallback only
- Add scripts/install-image-deps.sh to install Pillow/torch/rembg deps
- Remove stray root files (1, pdf-pptx.txt, eng.traineddata)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Add Korean search keywords to WEB category (뉴스, 검색해, etc.) so
Korean news queries trigger web_search instead of ollama_web_search
- Show ✓ SearXNG in startup banner when searxng_url is configured
- Update dental dict databases scripts and DB
- Clean up stale sessions and workspace runtime files
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
image_edit:
- speech_bubble operation with Korean font, rounded rect, tail
- sketch quality: CLAHE pre-processing for better line contrast
- output filename with timestamp to prevent overwrites
- anime/painting stylize timeout 30s→600s (AnimeGANv2 model download)
- remove_bg PNG output fix
server-v2.ts:
- resolveToolImageContent: embed edited image as base64 in tool results
so vision models can see the output (3 execution paths covered)
- _activeModelName: fix vision support detection for Ollama-hosted models
(kimi/gemini run via Ollama — provider='ollama' was masking vision capability)
- Synthetic tool call ID generation to prevent Gemini function_response empty name error
- image_edit path hint format changed to English to prevent Gemini hallucination
- userRequestedImageEdit: expanded keyword list (풍선, 달아, 붙여 등)
- image_read OCR failure returns success:true with visual description hint
- TOOL_BLOCKS photo: image_edit workflow documentation updated
web-ui/index.html:
- renderFileDownloads: skip download buttons for image file extensions
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Reorganize .smallclaw/databases/ around a single entry point with
subcommands (add-terms / add-images / verify / reassign / stats),
absorb pmc_reassign.py into reassign, and route every image source
through inline vision-model verification before commit.
Layout
config.py paths, endpoints, API-key locations (no more
hard-coded absolute paths in source files)
manage.py argparse dispatcher
workflow_terms.py seed-JSON import (JSONC supported, dedupes on
korean/english)
workflow_images.py renamed from dental_image_workflow.py; main()
converted to run(verify=True, ...)
workflow_verify.py validate_image() / verify_db() + verify_updates()
gate used by every add-images source
seeds/ recovered seed_all.json, seed_periodontics.json
scratch/ ad-hoc work area replacing the /tmp habit
(only README.md is tracked)
docs/howto.md moved + expanded
docs/archive_image_rounds/ one-off round scripts + their results
docs/archive_validation/ model-comparison + validation history
Verification
- VERIFY_MODEL defaults to qwen3.5:397b-cloud (50-image test on
procedure-heavy sample: 9.8s/img avg, zero timeouts, Kimi-level
rigor — see archive_validation/).
- Prompt strengthened: in-image text/captions are not valid grounds
for "적합"; technique terms require visible procedure steps, not
generic device/anatomy photos.
- All 9 add-images sources gated through verify_updates() before
commit; --no-verify escape hatch for bulk runs.
.gitignore: API key files, dental_images/, __pycache__/, scratch/*
(README.md kept).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Python stdout already includes the correct markdown download link.
Appending extra bare URL and absolute path caused the model to generate
two download links in its response. Keep only Python's stdout.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
The v2.2.3 hint "ALWAYS use /api/files/uploads/filename" was too broad,
causing the model to format PPTX download links as uploads/file.pptx
instead of the correct project_folder/file.pptx path returned by the tool.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Playwright's bundled browser install fails on Ubuntu 26.04 with
"Playwright does not support chromium on ubuntu26.04-x64".
Fall back to system Chromium (/usr/bin/chromium-browser etc.) via
executablePath when launching headless, bypassing the platform check.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
kimandClaude Sonnet 4.6