Filename check:
- Rewritten to flexibly parse multiple H&M naming conventions
(Display, DOOH, OOH, SOME STATIC, Social, POS, DS)
- Extracts country code, language code, dimensions, campaign number
- Scores based on how much metadata was extracted (not rigid pattern)
- Tested against real filenames: BG_bg, ES_es, NO-no formats
Price/currency check (new):
- Detects prices in images via LLM vision API
- Validates currency against global pricing reference (deterministic)
- Falls back to LLM validation for unknown countries
- Optional campaign pricing sheet validation when has_pricing=True
- Added to profile with weight 30
Profile weights rebalanced: filename 30, quality 40, price 30
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Add /report/<id>/download route using send_file instead of broken
static file URL (fixes 404 on Download Report button)
- Add campaign label to HTML report header (Campaign: ID - Name)
- Store campaign_id in report metadata_json for traceability
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Fix background parsing thread: pass app reference explicitly instead of
trying to access current_app inside the thread (was silently failing)
- Add progress bar with animated stages during upload and parsing
- Add data-id/data-status attributes to table rows for auto-polling
- On page load, automatically poll any pending/parsing rows and update
their status badges in-place (fixes stale "Pending" on tab return)
- Immediately inject new row into table after upload so user sees it
without needing to refresh
- Remove broken _parse_pricing_background function
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Introduces a new Campaigns module for uploading campaign presentation PDFs
that QC checks reference to validate assets against campaign-specific
guidelines (typography, layout, copy, pricing format). Also adds a global
pricing reference system that maps country codes to currency symbols and
formats for deterministic price/currency validation.
- New CampaignPresentation model + campaigns blueprint with CRUD routes
- PDF parsing via LlamaParse (text + multimodal page images)
- Global pricing PDF parsed into structured JSON lookup
- Campaign context injected into both image and video QC executors
- Quality checks enhanced with campaign guidelines in LLM prompts
- Price/currency check uses global pricing lookup (saves an LLM call)
- Campaign dropdown added to HM QC and Video QC configure pages
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
AKAZE tier needs the actual video file to extract frames, but our
temp-download-and-delete approach means the file is gone by that point.
Perceptual hash (Tier 1) works fine with saved fingerprint data.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Complete README rewrite reflecting deployed production state
- Comprehensive CHANGELOG entry for v2.0.0 covering all changes:
Docker deployment, auth overhaul, HM QC improvements, Video QC
(built from scratch), Video Master campaign matching, Usage Dashboard,
Google Gemini support, reporting enhancements, Box API fixes
- Updated .env.example with all current config options
- Removed outdated references to Azure AD, local development setup
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Box SDK .get() on folder objects fails with "Item.get() takes 1
positional argument" in the deployed environment. Replaced all
folder.get() calls with a new _get_folder_items() helper that uses
get_items() with pagination, falling back to folder.get() only as
last resort. This fixes list_subfolders, list_video_files, and
search_subfolder.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Global Masters folder contains subfolders (DOOH, DS, OLV, etc.) with
videos inside them, not videos directly. Added recursive=True option
to search one level of subfolders for video files.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Box SDK v3 Item.get() doesn't accept fields as positional argument.
Remove fields parameter and let it return full folder info including
item_collection with inline entries.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The CAMPAIGNS folder is owned by a different user and shared via
collaboration. get_items() and search API fail with "not found" for
these folders, but folder.get() works and returns inline items.
- Rewrite search_subfolder() to use folder.get() first, with pagination
fallback for folders with >100 items
- Rewrite list_subfolders() and list_video_files() with same approach
- Add BOX_CAMPAIGNS_FOLDER_ID config (156182880490) separate from
the QC reports folder
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Add BOX_CAMPAIGNS_FOLDER_ID config (156182880490) separate from
BOX_REPORT_FOLDER_ID which is for QC reports
- Update search_subfolder() to use Box search API first (fast for large
folders with 1000+ campaigns), fall back to folder listing
- Increase folder listing limit from 200 to 500
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Full workflow:
- Enter campaign name → search Box for campaign folder
- Auto-discover Global Masters and Regional Masters subfolders
- Preview: shows master count, countries, adaptation count
- Phase 1: Download each master to temp, fingerprint, delete video
- Phase 2: Download each adaptation to temp, match against masters, delete
- Results: per-master adaptation mapping, unmatched items, match rate
- HTML report with detailed breakdown
- Previous Matching Jobs table with View/Delete
Box client additions:
- search_subfolder() - case-insensitive subfolder search
- list_subfolders() - enumerate child folders
- list_video_files() - list video files in folder
- download_file_to_disk() - streaming download for large files (ProRes)
Storage: only fingerprints (~50KB) + key frames stored permanently.
Videos deleted immediately after processing.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Censorship check now only runs if filename contains _CEN suffix
(matches legacy behavior). Non-CEN files get "skipped" with 100 score
- When censorship is skipped, visual quality score is 100% of overall
- Updated language consistency prompt to avoid false positives:
- Words like "Rock" (German for skirt), "Mode" (fashion), etc.
- Must verify a word is NOT valid in the primary language before flagging
- Brand names and international terms are excluded from checks
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Full video QC workflow:
- Upload → Configure (LLM provider + job number) → Execute → Results
- Extracts 1 frame per second using existing FFmpeg/extract_thumbnails()
- Stitches frames into labeled grid image for efficient AI analysis
- Two separate AI checks:
1. Visual Quality (50%): language consistency, text legibility, logo clarity
2. Censorship (50%): body coverage and content appropriateness
- Progress tracking via SSE/polling
- HTML report generation with per-check scores
- Previous Video QC Reports table with View/Delete on index page
- Usage dashboard integration (logs tokens + cost per API call)
- Supports OpenAI GPT-4o and Google Gemini provider choice
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- New UsageLog model tracking every LLM API call (provider, model,
tokens, estimated cost, user, module, check name)
- Instrument LLMConfig.call_vision_api() to auto-log each call
- New /usage tab in nav bar with dashboard showing:
- Summary cards (total calls, tokens, estimated cost)
- Breakdowns by provider, model, tool, and user
- Recent API calls table
- Time filters (All Time, 30 Days, 7 Days, Today)
- Cost estimates based on per-model token pricing
- Pass logged-in user through executor context for tracking
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- HM QC: trash icon per report row, DELETE /hm-qc/report/<id> removes
DB record and file from disk
- Reporting: trash icon per Box job row, DELETE /reporting/history/delete/<job>
removes all saved Box reports for that job number
- Confirmation prompts before deletion
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Add /hm-qc/report/<id> route to serve saved reports by database ID
- Create view_report.html template with score summary and embedded report iframe
- Add "View" button column to Previous QC Reports table
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Replace dimension_check with filename_parse in H&M Image Check profile
- Rewrite quality check prompt to be much stricter on text legibility:
- Text legibility is now the #1 priority (CRITICAL check)
- Any illegible text forces score below 70 (FAILED)
- Explicit instructions to check ALL text including small overlays
- Low contrast text on dark/busy backgrounds flagged as common failure
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Remove "Previous QC Reports" table from reporting index
- Add "Previous QC Reports" table to HM QC index page
- Update HM QC index route to pass recent reports
- Update feature list to reflect current checks
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Add /reporting/history/<job_number> route that loads saved reports from
disk/database instead of re-fetching from Box
- Split "Previous Searches" into "Previous Box Reports" and "Previous QC
Reports" sections with separate tables
- "View" buttons link to history_dashboard (reads from saved files)
- Box reports show job-grouped view, QC reports show individual files
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- After a successful Box search, save downloaded HTML reports to disk
and record them in qc_reports table (report_type='box_import')
- Skip duplicates by checking box_id in metadata
- Update reporting index to show "Previous Searches" with source badges
- Rename "Recent Reports" to "Previous Searches" for clarity
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Box search generator auto-paginates through ALL results (35k+ for broad
queries). Added iteration caps to prevent runaway API calls:
- _search_folder_by_name: cap at 50 results
- _search_files_by_job_number: cap at 100 results
- _scan_folder_by_name: cap at 1000 items
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Update image quality prompt to evaluate text/title legibility
- Add Google Gemini (generativeai) as LLM provider in LLMConfig
- Add AI Provider dropdown on configure page (OpenAI GPT-4o / Google Gemini)
- Pass selected provider through execute routes to override profile defaults
- Add google-generativeai to requirements.txt
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Replace 3 profiles with single "H&M Image Check" (dimension_check + image_quality)
- Remove filename_parse check (pattern didn't match actual filenames)
- Create DimensionCheck class for image dimension validation
- Fix configure page to route multi-file uploads to batch endpoint
- Auto-select single profile, show file list on configure page
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Fix back navigation on reporting dashboards (linked to / instead of /reporting/index)
- Add "Run Another QC" button on HM QC results page
- Add Recent Reports table on reporting search page (grouped by job number)
- Add Recent QC Reports table on HM QC upload page
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The require_auth decorator was never applied to routes, leaving
the entire app publicly accessible. Added a before_request hook
that redirects unauthenticated users to the login page.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Add Dockerfile, docker-compose.yml, .dockerignore for containerised deployment
- Add deploy/ scripts (deploy.sh, nginx/apache configs, password generator)
- Replace MSAL/Azure AD auth with local username/password authentication
- Add login.html template
- Simplify app.py, middleware, and auth routes for production use
- Update gunicorn_config.py and wsgi.py for Docker/production
- Update templates to work with new auth and URL prefix handling
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Rewrite CHANGELOG.md to cover platform v1.0.0 and auth fix,
with reporting module history preserved as subsection
- Replace stale DOCUMENTATION_SUMMARY.txt with current project
structure and key decisions
- Rewrite MIGRATION_GUIDE.md to document legacy tool consolidation
with complete file mappings for hm_qc and video_qc
- Add legacy context headers to module docs (legacy_README,
legacy_DEV_SETUP, legacy_CLAUDE) pointing to main README
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The popup login flow was broken because the Flask 302 redirect from
/ to /reporting/index caused MSAL in the popup to consume the auth
code hash before the parent window could detect it, leaving the
parent stuck on "Authenticating..." while the popup rendered the
full app.
- Switch signIn() from loginPopup() to loginRedirect()
- Add handleRedirectPromise() at start of initAuth() to process
the auth code on page load after returning from Microsoft
- Change root route from 302 redirect to direct template render
so the #code=... hash fragment is preserved for MSAL
- Switch signOut() from logoutPopup() to clearCache()
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Merge original CLI check implementations from hm_qc/ and
hm_qc_video/ repos into modules/*/checks/legacy/ directories.
Includes profiles, launchers, utils, orchestrators, and the
standalone video Flask web app. Reference files (test data,
results, cheat sheets) copied to gitignored reference/ directory.
Censorship trainset copied to gitignored data/supporting/.
The legacy/ naming convention separates original run_check()
function-based implementations from the new BaseCheck class
architecture.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
New blueprint-based module system (hm_qc, video_qc, video_master,
reporting), core framework (database, config, templates), and
unified web interface with progress tracking and tab navigation.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
