- Add 5 native ingestors to cctv_pipeline.py: DGT (~1,917 cameras),
Madrid (~357), Málaga (~134), Vigo (~59), Vitoria-Gasteiz (~17)
- Fix DGT DATEX2 parser to match actual XML schema (device elements,
not CctvCameraRecord)
- Wire all new ingestors into the scheduler via data_fetcher.py
- Remove standalone spain_cctv.py by Alborz Nazari, replaced by native
pipeline ingestors that integrate with the existing scheduler pattern
- Fix CCTV image loading for servers with Referer-based hotlink
protection (referrerPolicy="no-referrer")
- Replace external via.placeholder.com fallbacks with inline SVG data
URIs to avoid dependency on unreachable third-party service
- Surface source_agency attribution in CCTV panel UI for open data
license compliance (CC BY / Spain Ley 37/2007)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Map ~35,000 power generation facilities from 164 countries using the
WRI Global Power Plant Database (CC BY 4.0). Follows the existing
datacenter layer pattern with clustered icon symbols, amber color
scheme, and click popups showing fuel type, capacity, and operator.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Merge both feature sets: keep JSDF bases (gsdf/msdf/asdf branches) from
PR #77 and East Asia adversary bases (missile/nuclear branches) from main.
Union all branch types in tests and MaplibreViewer labels.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add ASDF (8), MSDF (6), and GSDF (4) bases to military_bases.json.
Colocated bases (Misawa, Yokosuka, Sasebo) have offset coordinates
to avoid overlap with existing US entries. Add branchLabel entries
for GSDF/MSDF/ASDF in MaplibreViewer popup.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Add gsdf/msdf/asdf to known_branches in test_branch_values_are_known
- Add test_includes_jsdf_bases for Yonaguni, Naha, Kure
- Add test_colocated_bases_have_separate_entries for Misawa
- Add buildMilitaryBasesGeoJSON tests with ASDF branch validation
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Add 68 military bases (PLA, Russia, DPRK, ROC, Philippines, Australia)
with data-driven color coding (red/blue/green) on the map
- Add 6 news RSS feeds (Yonhap, Nikkei Asia, Taipei Times, Asia Times,
Defense News, Japan Times) and 15 geocoding keywords for islands,
straits, and disputed areas
- Extend ICAO country ranges for Russia, Australia, Philippines,
Singapore, DPRK and add Russian aircraft classification (fighters,
bombers, cargo, recon)
- Create PLAN/CCG vessel enrichment module (90+ ships) following
yacht_alert pattern for automatic MMSI-based identification
- Update frontend types and popup styling for adversary/allied/ROC
color distinction
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Infer country and military force (PLA, JSDF, ROK, ROC) from ICAO hex
address blocks when the flag field is Unknown. Extract and extend aircraft
model classification to cover East Asian fighters, cargo, recon, and
tanker types with hyphen-normalized matching.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add 18 US military bases (Japan, Guam, South Korea, Hawaii, Diego Garcia)
as a toggleable map layer. Follows the existing data center layer pattern:
static JSON → backend fetcher → slow-tier API → frontend GeoJSON layer.
Includes red circle markers with labels, click popups showing operator
and branch info, and a toggle in the left panel.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add 5 East Asia-focused RSS feeds (FocusTaiwan, Kyodo, SCMP, The Diplomat,
Stars and Stripes) and 22 geographic keywords (Taiwan Strait, South/East
China Sea, Okinawa, Guam, military bases, etc.) to improve coverage of
Taiwan contingency scenarios.
Refactor keyword matching into a pure _resolve_coords() function with
longest-match-first sorting so specific locations like "Taiwan Strait"
are not absorbed by generic "Taiwan".
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Validates that every destination path stays within project_root
before writing. Prevents a malicious zip from writing outside
the project directory via ../traversal entries.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Former-commit-id: 3140416e80b1b56e4e6cccc930d11c2d5f9b1611
In Docker, main.py lives at /app/main.py so Path.parent.parent
resolves to filesystem root /, causing PermissionError on .github
and other dirs. Now detects this case and falls back to cwd.
Also grants backenduser write access to /app for auto-update.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Former-commit-id: 12c8bb5816a70161d5ab5d79f9240e7eab6e6e15
os.makedirs was outside try/except so permission-denied on .github
directory creation crashed the entire update. Now both makedirs and
copy are caught. Also prunes protected dirs from os.walk so the
updater never even enters .github, .git, .claude, etc.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Former-commit-id: d4bdef4604095a82860a4bc91bec3435a878f899
The auto-updater tried to extract .github/ from the release zip,
causing Permission denied errors. Added .github and .claude to the
protected directories list so they are skipped during extraction.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Former-commit-id: 8916fa08e005820ddbfc3b195c387dbf6187587e
Auto-update POST goes through Next.js proxy which dies when extracted
files trigger hot-reload. Network drops now transition to restart polling
instead of showing failure. Also adds admin key header and FastAPI error
field fallback. Gitignore updated to protect internal docs.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Former-commit-id: 03162f8a4b7ad8a0f2983f81361df7dba42a8689
selectedEntity.id was stored as a numeric array index into data.gdelt[]
and data.news[]. After any data refresh those arrays rebuild, so the
stored index pointed to a different item — showing wrong popup content.
GDELT features now use g.properties?.name || String(g.geometry.coordinates)
as a stable id; popups resolve via find(). News popups resolve via find()
matching alertKey. ThreatMarkers emits alertKey string instead of originalIdx.
ThreatMarkerProps updated: id: number → id: string | number.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Former-commit-id: c2bfd0897a9ebd27e7c905ea3ac848a89883f140
Shadowbroker