Changelog

What shipped, when.

2026-05 - Markup Billing pivot (BYOK)

New pricing model: token packs and monthly subscriptions retired - moved to degressive BYOK markup. FREE (basic) / SCALE (9-12.5% markup, automatically decreasing) / ENTERPRISE (negotiated). All features included from Scale.
CORTEX: autonomous intelligence layer (5 agents: Guardian Triage, Workspace Intelligence, Strategic Advisor, Self-Tuning Engine, Network Intelligence). CORTEX Inbox now available in the dashboard.
Auto-reload: wallet automatically tops up when balance drops below a configurable threshold.

Initial V2 pricing: flat monthly subscription + BYOK (Free / Build $15 / Scale $39 / Business $249 / Enterprise). *Replaced by degressive BYOK markup model in May 2026.*
New API key format: hw_live_ + 32 hex, minted from Dashboard → Keys.
New base URL: https://app.hiway2llm.com/v1.
Response envelope: _hiway metadata object with routed_model, routed_tier, score, fallback, cache_hit, upstream_cost, savings.
Provider fallback: auto-retry against cheapest same-tier model on upstream errors (max 2 retries).
Semantic cache: Qdrant-backed, cosine ≥ 0.92.
PII masking: opt-in regex on email/phone/card/IBAN/API-keys before cache hashing.
A/B Experiments: parallel model benchmarking.
Open-source SDK & CLI: pip install hiway2llm, npm install @hiway2llm/client, curl ... | sh CLI. MIT-licensed at Hiway2llm/client.

Guardian: per-workspace editable rules (dedup, cost spike, context bloat, zombie, rate limit)
Budget Control: monthly cap on upstream BYOK cost with three verdicts (DOWNGRADE / LIGHT_ONLY / BLOCK)
Alert webhooks on Guardian fires and Budget Control transitions