April 19, 2026
Port Authority Governance dataset launches, AIS coverage expands, NYC zoning data populates, and APRS envelope rolls out across all Codex datasets.
Automatic detection of unresponsive data sources
The platform now detects when a data source stops returning records across multiple consecutive collection runs and flags it as failed — even if the source's API responds without errors. Previously, a data source could go completely silent (returning zero records indefinitely) without triggering any alert, because the system only flagged runs where records were fetched but failed to persist. This change is what caught the civic and environmental review sources described below. No action is required on your part — the detection runs automatically and feeds into existing data freshness monitoring.
Council decisions, zoning variances, and environmental reviews now populated
Three Civic Intelligence tables — council_decisions, zoning_variances, and environmental_reviews — were empty since launch because upstream data sources changed. Legistar-powered cities now use an HTML scraping fallback when the JSON API is unavailable, and California environmental reviews pull from the new CEQANet endpoint. If you query council decisions or zoning data through the Locus API or MCP server, you should now see results for cities that use Legistar (including San Francisco and New York City). EPA EIS environmental reviews are not yet available.
Port authority governance dataset
Codex now includes a comprehensive governance dataset covering five major US port authorities — Oakland, Long Beach, Los Angeles, JAXPORT, and the Port Authority of New York and New Jersey. The dataset contains structured board meeting records, agenda items, claims, facts, tariff documents, and resolutions going back over a year. Each port's records are normalized into a consistent schema so you can run cross-port queries out of the box. JAXPORT tariff data includes line-item extraction with bitemporal diffing, letting you track rate changes over time. The original release referenced a dedicated schema page; that page is not part of the current public catalog.
APRS envelope coverage across all datasets
All 111 Codex tables now carry the full APRS envelope — record_id, chunk_id, bitemporal fields, confidence_score, and provenance. A new compliance gate in CI ensures every table passes shape checks before deployment, so envelope coverage won't regress. You can track current coverage on the envelope coverage dashboard.
In-app support chat
Locus now has an in-app support chat. When you're signed in, the chat verifies your identity so conversations are automatically linked to your account — meaning faster responses with no need to re-identify yourself. Anonymous chat is available for unauthenticated visitors.
Expanded AIS coverage area
Overwatch now tracks vessels across a much wider area around each monitored port. The default coverage radius has been expanded from approximately 55 km to 220 km, catching in-transit vessels well before they reach anchorage so your map views populate earlier. If you need full ocean tracking for use cases like dark fleet detection or open-ocean STS monitoring, you can now enable global AIS coverage. See AIS coverage for configuration details.
Tiered AIS data retention
Overwatch now applies structured data retention and downsampling to AIS position data. Positions from the last 7 days are kept at full resolution. Data between 7 and 30 days old is thinned to one report per vessel per minute — enough for route replay and behavioral analysis — and data between 30 and 90 days is reduced to 5-minute intervals. Older positions leave the fast query tier entirely, but full-resolution archives are preserved in cold storage and can be rehydrated on request. This keeps queries fast on recent data while preserving deep history for investigations.
Landing page visual refresh
The Locus landing page has been polished with motion effects, improved card interactions, and a refreshed editorial layout including section breaks, a masthead dateline, and updated product marks across all four Axiom products.
Improved data quality for NFIP and HMDA datasets
NFIP flood claims with missing dates have been removed, cleaning up approximately 97,000 orphan records. HMDA mortgage disclosure data has been corrected to fix field-name mismatches and hash collisions that could cause duplicate records. These changes improve the reliability of queries against both datasets.
Faster APRS backfill for large datasets
HMDA loan records — previously missed during the initial backfill rollout — are now being populated with full APRS envelope metadata. Processing speed for the five largest legacy tables (EPA facilities, HMDA lending, CDC health metrics, building permits, and tax assessments) has also been increased 5x, bringing projected backfill completion from months to under 60 days. You can monitor progress on the envelope coverage dashboard.
Restored NTD transit ridership data
The National Transit Database ridership data source has been restored after the upstream API was retired by FTA. Codex now pulls directly from the official monthly data files, so transit ridership records are flowing into cell scoring again without interruption.
Code enforcement data collection restored
Fixed an issue where the code enforcement data collector was attempting a full refresh on every run instead of incremental updates, causing it to time out repeatedly. The collector now processes only the last 30 days of records on each run, keeping code enforcement data in cell scoring current without exceeding processing limits.
Platform security hardening
A comprehensive security audit addressed several critical issues: API authentication now correctly rejects malformed bearer tokens, webhook endpoints enforce idempotency to prevent replay attacks, admin endpoints require proper authorization, and rate limiting no longer fails open under load. Filter parameters are validated to prevent injection. No action is required on your part — all fixes apply automatically.
More reliable global vessel tracking
Fixed an issue where enabling global AIS tracking could cause vessel processing errors. Vessels with shared or recycled MMSIs — and vessels seen for the first time — would fail to process, resulting in dropped position updates. Both cases are now handled correctly, so global tracking processes all vessels without interruption.
Flag-hopping detection now tracks all flag changes
Fixed an issue where flag-state changes were not being recorded for some vessels due to an incorrect column reference in the tracking trigger. Flag-hopping history is now captured reliably for all vessels, so the risk scoring system has complete data when evaluating registry-change patterns.
Fixed a client-side crash on page load
Resolved an issue where Overwatch could display "Application error: a client-side exception has occurred" when loading a page. The crash was caused by an analytics initialization timing issue and has been fixed — pages now load reliably.
Performance and privacy improvements
Video playback components now load on demand instead of at page load, reducing initial bundle size. Scroll-heavy pages use frame-throttled listeners for smoother performance. Analytics tracking now respects consent preferences before initializing.
More reliable kinematic fingerprint computation
Fixed an issue where the kinematic fingerprint computation could silently fail without logging an error. If the function timed out before reaching its success or error handler, no ingestion log was written, making it appear as though the run never happened — even though the scheduler reported it as complete. Kinematic fingerprint profiles now update reliably, so vessel behavioral data stays current. No action is required on your part.
More reliable Equasis vessel enrichment
Fixed an issue where vessel metadata enrichment from Equasis failed approximately 96% of the time after a session renewal. When a session expired mid-run, the immediate retry did not fully establish the new session before requesting vessel data, causing the request to be rejected. Enrichment retries now mirror the full session setup flow, so vessel profiles — including registered owner, flag state, and deadweight tonnage — are populated more consistently.
More reliable data ingestion
All Codex data loaders now use a safer batch upsert strategy that logs the first distinct error per batch and provides a summary of attempted, succeeded, and failed records. Previously, a single malformed record could silently stall a batch. This means data coverage stays current even when upstream sources occasionally deliver incomplete records.
Improved scoring model accuracy
The Locus scoring model has been recalibrated with updated decay parameters for commercial license and foot traffic signals. These refinements mean that cell-level scores now more precisely reflect how quickly each signal loses relevance over time, giving you sharper risk and opportunity assessments without any changes to your workflow.
Cross-port board meeting backfill
Overwatch now includes a full year of structured board meeting data across all five supported port authorities — Oakland, Long Beach, Los Angeles, JAXPORT, and the Port Authority of New York and New Jersey. Over 90 meetings and 1,400+ agenda claims have been backfilled and normalized, so you can run cross-port queries against a meaningful historical window from day one. Oakland meetings also include extracted vote outcomes and resolution facts via Legistar. See the data catalog for the full list of port authority tables.
Configurable AIS vessel tracking coverage
Operators can configure how far around a monitored port Overwatch tracks vessels through deployment environment settings. The worker's AISSTREAM_BBOX_BUFFER_DEG and AISSTREAM_GLOBAL settings let operators choose between the default expanded radius (~220 km) and a maximum global range, depending on their coverage needs. This gives operators control over the tradeoff between broader vessel visibility and data volume.
Stronger data access controls
Access controls have been tightened on civic intelligence and flood exposure datasets. These tables are now subject to the same row-level security policies as the rest of the platform, ensuring that data access is properly scoped to your account. No action is required on your part.
Additional platform security hardening
A comprehensive security audit resulted in tighter access controls across the platform. Key improvements include enforced tier-based access checks on data export and event endpoints, so free-tier API keys can no longer access paid datasets. Audit log CSV exports are now sanitized against formula injection. Row-level security policies have been extended to additional tables that were previously unprotected. No action is required on your part — these changes apply automatically.
Corrected Port Authority of New York and New Jersey query matching
Fixed a data inconsistency where the Port Authority of New York and New Jersey was stored under a mismatched identifier in the port authority lookup table. Cross-port queries that join against the port authority governance schema now correctly include NY-NJ records. No action is required on your part — the fix applies automatically to all existing and future queries.
Tighter access controls on 311 complaint data
Row-level security policies have been extended to 311 service request and complaint signal tables, matching the access controls already in place across the rest of the platform. Direct API access to these tables is now properly scoped to your account. No action is required on your part.
Stricter virtual card issuance checks
Fixed a security issue where a virtual card could theoretically be issued against a cardholder belonging to a different organization. Layer now verifies that the cardholder is part of your organization before creating the card, preventing any possibility of cross-tenant spend. No action is required on your part — the fix applies automatically to all new card requests.
More reliable AIS data ingestion
Fixed an issue where AIS position updates from AISHub were intermittently dropped due to rate-limit responses being misclassified as failures. Approximately 26% of ingestion runs were incorrectly logged as errors, which could cause brief gaps in vessel position data. Rate-limited responses are now handled gracefully, so AIS coverage is more consistent. No action is required on your part. See AIS coverage for details on how Overwatch sources vessel positions.
USPS vacancy data now requires HUD registration
HUD has restricted access to USPS vacancy data to registered governmental and nonprofit entities. If you use the Locus CLI to load vacancy data, you now need to either download the file manually from HUD after registering, or provide an authenticated session cookie for automated downloads. Existing vacancy data already in your account is unaffected. See cell scoring for how vacancy data feeds into territory assessments.
Encrypted Slack integration tokens
Slack integration tokens are now encrypted at rest using AES-256-GCM. Previously, tokens were stored in plaintext. This change applies automatically to all new and existing Slack connections — no action is required on your part. See the Slack integration page for setup details.
Unified credential handling across integrations
Fixed an inconsistency where different integrations read stored credentials through separate code paths, which could cause token refresh failures for some connectors. All integrations now use a single credential-read path with consistent decryption and error handling, improving sync reliability across the board.
More resilient zoning variance data collection
Zoning variance data now has the same HTML scraping fallback that council decisions use (see "Council decisions, zoning variances, and environmental reviews now populated" above). When a Legistar-powered city's JSON API is unavailable or restricted, Locus automatically falls back to scraping the public meeting calendar, so zoning variance records continue flowing into your cell scores and the Civic Intelligence dataset without interruption. No action is required on your part.
April 20, 2026
Slack credential encryption, NYC council decisions and zoning variances populated, faster kinematic processing, and improved loitering detection accuracy.
April 18, 2026
Layer GA, Codex APRS envelope rollout, Overwatch sanctions traversal and flag-hopping detection, Locus FEMA flood scoring, and 12 new public data sources.