Cập nhật

Nhật ký thay đổi

Cập nhật mới nhất, model mới và cải tiến cho Izzi API.

2026-03-30v1.2.0Feature

Phase 2 — Responses API & Extended Thinking

→Added POST /v1/responses endpoint — OpenAI Responses API format (compatible with n8n, LangChain)
→Added POST /v1/messages/count_tokens — pre-count tokens before requesting
→Extended Thinking support for Claude models (thinking.type + budget_tokens)
→Added /changelog page

2026-03-29v1.1.0Feature

→Concurrent request limiting (Redis semaphore) per plan
→Rate limit headers on ALL API responses (x-ratelimit-tier, x-ratelimit-limit-requests, etc.)
→Model aliases — 42 legacy model names auto-mapped to current equivalents
→IP brute force protection — 50 failed auth → 5-minute auto-block
→Standardized error response system — 11 error types with consistent format
→Endpoint aliases: /openai/v1/*, /api/v1/* now supported
→GET /v1/models/:id — single model lookup
→Added Privacy Policy & Terms of Service pages

2026-03-28v1.0.0Feature

→10 AI providers: Anthropic, OpenAI, Google, xAI, Groq, Cerebras, OpenRouter, StepFun, Z.ai, NVIDIA
→28+ models including Claude Sonnet 4, GPT-4.1, Gemini 2.5 Pro, Grok 3
→Smart routing with automatic fallback when providers are unhealthy
→Real-time billing with per-token pricing
→Admin dashboard with provider key management
→Response caching system with Redis
→Free tier with 10M tokens/month
→First-deposit bonus ($1 → $15)

2026-03-25Models

2026-03-20Improvement