<- Back to NOSIBLE WorldChangelog
NOSIBLE World

Changelog

Every significant change to NOSIBLE World in reverse chronological order. We ship continuously; this log captures the releases worth reading.

Tags:FeatureFixInfraDataDesign
  1. Feature

    Pricing simplified to Public + Enterprise — free signup deferred

    • Dropped the three-tier (anonymous / free / paid) model. v1 now ships two tiers: Public (free, no account, past 7 days, full UI and markdown download) and Enterprise (starting at $5,000/month, 30-year archive, 15-minute updates, API + webhook, SLA, dedicated CSM — book a demo).
    • AUTH_PROPOSAL.md marked as deferred. No public signup in v1; auth is scoped to enterprise SAML/SSO and will land when the first enterprise customer requests it. The provider and architecture analysis in the proposal remains as a reference.
    • lib/access.ts simplified to two tiers (public / enterprise). All markdown API endpoints updated: 403 responses now read "This date is part of the Enterprise archive. Book a demo at https://nosible.com/contact for full access."
    • "Get a free account" CTAs removed across the site. Replaced with "Book a demo for full archive access" where relevant. Pricing, Give Claude the World, API Docs, and Roadmap pages all updated to reflect the new model.
  2. Feature

    Auth & freemium tier design published

    • Published the full Auth & Tier Model Design Proposal (AUTH_PROPOSAL.md) — three tiers (anonymous / free / paid), Auth.js v5 recommendation, DB schema, session strategy, enforcement points, and 8-step migration plan.
    • Tier gates defined: 7-day anonymous window, 30-day free window, 20-year archive behind Pro. Saved searches, webhooks, and API access gated accordingly.
    • Pricing pages, about, changelog, roadmap, and press kit shipped as part of the enterprise positioning push.
  3. Design

    Frontend security audit and security.txt

    • Completed a full frontend security audit: Content-Security-Policy hardened, X-Frame-Options and Referrer-Policy set in Next.js headers config.
    • Added /.well-known/security.txt with responsible disclosure instructions and PGP contact details placeholder.
    • Reviewed all user-facing data paths for XSS vectors; no issues found in the current mock-data frontend.
  4. Fix

    Residual filter counts, full-width methodology body, advanced-filters spacing

    • Fixed a bug where the active-filter count badge in the world header showed stale counts after clearing individual filters via the filter sidebar.
    • Methodology page body copy now spans full content width on narrow viewports; the previous max-width cap was too tight for tablet-portrait.
    • Advanced-filter panel internal spacing regularised — gap between filter groups was inconsistent depending on which filters were visible.
  5. Data

    IPTC-keyed legend, uniform placeholder color, fixed methodology heading

    • World map bubble legend switched from IAB-keyed to IPTC Media Topics keyed, matching the filter sidebar and event detail classification rows.
    • Placeholder event cards and loading skeletons now use a single consistent bg-slate-2 color rather than the previous mix of slate-1 and slate-2.
    • Methodology page H1 heading was incorrectly weighted at font-normal; corrected to font-bold consistent with other content pages.
  6. Feature

    LLM-generated event titles via Gemini Flash Lite, country centroids

    • Pipeline Stage 7 now generates a clean editorial title for every event using a single Gemini 2.5 Flash Lite call constrained by the spaCy NER entity map — no hallucinated people or organisations that did not appear in the source documents.
    • Country centroid coordinates baked into the event record; the world map now positions bubbles at the correct country centroid rather than a fixed capital.
    • Long-tail IAB categories (those with fewer than 3 qualifying events per day) now surface in the filter sidebar rather than being collapsed into "Other".
  7. Infra

    Multi-ontology classifier rebuild with OpenAI title-description embeddings

    • Replaced the embedding ensemble with one OpenAI text-embedding-3-large vector over each event title and description, scored by cosine similarity against taxonomy caches.
    • Classification pipeline now covers all 13 v1 ontologies in a single pass rather than sequential per-taxonomy runs; wall-clock time per day reduced by ~60%.
  8. Data

    Data model lockdown — deterministic event IDs, 13-ontology schema

    • Published DATA_MODEL.md: the full canonical event schema specification with deterministic event_id encoding (date + IAB + geo + topic packed into a 256-bit buffer with ~48 bits entropy suffix).
    • Locked the 13 v1 ontologies: IPTC Media Topics, IPTC Genre, Wikidata QIDs (staging), GICS, NOSIBLE Corporate Events, PLOVER, EM-DAT, MITRE ATT&CK, ICD-11, SportsML, schema.org Event, Media Frames, Ekman 6, C2PA (staging), IAB.
    • Event shape frozen: date, title, description, continent, region, country, language, sentiment, materiality, time_horizon, key_entities, similar_events.
  9. Design

    Frontend redesign — world HUD, event detail, and methodology pages

    • Complete frontend redesign: dark obsidian theme, neo-green accent, Orbitron / Space Grotesk / Inter / Space Mono type stack, zero border-radius sharp aesthetic.
    • World HUD rebuilt: sticky header, interactive date picker, Leaflet world map with IAB-bucketed bubble layer, filter sidebar with live residual counts, spotlight search (Ctrl+K).
    • Event detail page redesigned: classifications row, entities row, coverage section, stats bar, deep-dive panel, back-strip breadcrumb.