API Changelog & Versioning

Backfilled entry — this change shipped earlier but was not logged at the time.

• Billing unit is now analyzer runs (1 run = 1 analyzer applied to 1 text), replacing per-request accounting
• Plans consolidated to a single all-analyzers family — every plan includes all production analyzers
• Automatic overage on paid plans; optional spend caps; no hard quota stop by default

Backfilled entry — this change shipped earlier but was not logged at the time.

• Added the Growth plan: $349/mo, 150K analyzer runs, 25 RPS, $2.00/1K overage, onboarding call + quarterly office hours
• Pricing page reworked: monthly/annual toggle, honest break-even map, embedded workload calculator, worked examples, Success Package on Production+
• /verify is public again — verify a receipt without an account (it had been shadowed behind the portal login)
• Cost calculator now shows the full-plan comparison, honest savings/additional-investment framing, and an LLM full-pipeline-cost toggle

Backfilled entry — this change shipped earlier but was not logged at the time.

• Plan lineup repositioned: Build $149/mo, Production $799/mo, Scale $2,999/mo, Enterprise custom
• Annual billing at 15% off on all paid plans
• Admin commercial, operations, and content infrastructure (audit, promotions, dunning, emergency pages)

Backfilled entry — this change shipped earlier but was not logged at the time.

• Billing unit is now analyzer runs (1 run = 1 analyzer applied to 1 text), replacing per-request accounting
• Plans consolidated to a single all-analyzers family — every plan includes all production analyzers
• Automatic overage on paid plans; optional spend caps; no hard quota stop by default

Backfilled entry — this change shipped earlier but was not logged at the time.

• Added MCP server integration — manage API keys, check usage, and verify receipts from LLM agents
• Added ticket-problem pipeline, security lifecycle, and data governance infrastructure
• Expanded production analyzer coverage beyond the original Trust & Safety domain

• Added public determinism verification page (/verify) — run a detection and verify output hash reproducibility
• Added data handling transparency page (/zero-retention) — API design proof with request lifecycle diagram
• Added no-training transparency page (/no-training) — architecture diagram showing inference/training isolation
• Narrowed homepage focus to Trust & Safety content moderation
• Standardized language counts across all pages (25 unique languages across 5 domains)
• Removed VPC/on-prem deployment language — submitted-content-not-retained SaaS only
• Removed cost-comparison claims from hero and docs sections
• Updated homepage CTA to link directly to live playground demo

• Published SDK interface preview (Python sync/async + JavaScript/TypeScript client design, retries, backoff) - packages in development, not yet on PyPI/npm
• Added OpenAPI 3.0.3 specification as source of truth
• Added Trust Center page with GDPR/CCPA/SOC 2 roadmap
• Added benchmark methodology page with per-language results
• Added LLM moderation comparison page
• Added status page with live API health check
• Added administrative access controls to the admin panel
• Reframed case study as Reference Architecture with illustrative disclaimer
• Playground v3: language auto-detect, client-side rate limiting, server 429 handling
• Analytics funnel: 24h/7d/30d trends + drop-off percentages

• Added Cookie consent banner with Accept/Decline
• Added analytics opt-out via a browser-level opt-out flag
• Added AI-generated content disclosure on blog posts
• Added Privacy page
• Added Signup, API-key, checkout, and plan activation analytics events
• Added admin analytics funnel widget
• Added homepage interactive playground
• Added Quickstart page with curl/Python/JS/Go/PHP examples
• Added API reference page with error codes and rate limits
• Added sitemap generation (36 static pages + blog posts)

• Added Market Intelligence product namespace
• Added Crisis Intelligence product namespace
• Added Cybersecurity product namespace
• Added product-specific pricing tiers and bundles
• Added volume calculator and pricing quiz
• Added competitor comparison tables

• Improved cross-language detection quality
• Improved Polish precision to 86.4%
• Improved Ukrainian precision to 84.1%
• Improved Russian precision to 81.8%

• Initial public release
• Trust & Safety detection across 14 languages
• 200 indicators across 20 groups
• Deterministic detection with confidence scores
• Published performance target: p95 under 500ms
• Stripe and PayPal payment integration
• Administrative panel
• API key management with usage tracking

• Published benchmark catalog bcat_Q7K2N — per-language Top-3 results for analyzers with published evidence; analyzers without published results are marked pending
• Methodology and honest-spread reporting documented at /benchmark-methodology

• Multi-domain expansion - configuration IDs extended to all production analyzers; improved multilingual robustness and production detection quality. (Reconstructed entry - original ship window, not exact date.)

• Improved multilingual robustness - expanded per-language evaluation across the first wave of supported languages. (Reconstructed entry - original ship window, not exact date.)

• Initial production configuration - Trust & Safety signal catalog v1 (200 indicators, 20 groups), initial language coverage.

• Published benchmark catalog bcat_Q7K2N — per-language Top-3 results for analyzers with published evidence; analyzers without published results are marked pending
• Methodology and honest-spread reporting documented at /benchmark-methodology