> ## Documentation Index
> Fetch the complete documentation index at: https://docs.elumenta.ru/llms.txt
> Use this file to discover all available pages before exploring further.

# Changelog

```mdx theme={null}
---
title: "Changelog"
description: "API updates, new models, and breaking changes"
---

## March 2026

### 2026-03-17 — Tele2 Carrier Billing Token Pack Limit

**Breaking changes:**
- Tele2 builds now hide the Ultra token pack (10,000 tokens / 24,990 ₽) due to Tele2's 14,990 ₽ per-transaction carrier billing ceiling
- Premium pack (5,000 tokens / 13,990 ₽) is now the highest available pack on Tele2
- `/api/v2/payments/t2/quick-checkout` and `/api/v2/payments/t2/initiate` now return HTTP 400 for unknown article IDs instead of silently falling back to Basic subscription

**Improvements:**
- Defense-in-depth: Ultra pack filtered client-side on T2 builds and rejected server-side with clear error message
- YooKassa and Telegram Stars payment methods unaffected — all packs remain available

---

### 2026-03-16 — Nano Banana Pro Migration & Error UX

**Breaking changes:**
- Nano Banana Pro migrated from Replicate to Google Imagen API
- Model provider changed: `replicate` → `google`
- Provider model updated: `google/nano-banana-pro` → `imagen-3.0-generate-002`

**Improvements:**
- Enhanced error messaging for image generation with specific error types:
  - Model overload (503 errors) — suggests waiting or switching models
  - Generation timeout — retry suggestion
  - Content policy violations — prompt modification guidance
- Google Imagen integration with 4K resolution support and reference images (up to 14)

---

### 2026-03-15 — Phase 18/19: N8N Automation & Error Handling

**New features:**
- N8N workflow auto-release system — new product phases automatically create "In Dev" release pages in БД Релизы
- Automated deployment notifications to Notion

**Bug fixes:**
- ImageGenPage now properly handles synchronous generation failures with 6-second error messages instead of silent resets

---

### 2026-03-14 — Phase 17: Academy & Content Engine

**New features:**
- AI Academy launched at [elumenta.ru/academy](https://elumenta.ru/academy) — 97 learning articles across 8 categories (basics, prompting, models, use cases, and more)
- Blog launched at [elumenta.ru/blog](https://elumenta.ru/blog) — trend articles auto-generated weekly
- Web admin panel `/admin/content` and `/admin/academy` for content management
- Scheduled Generations — create recurring generation jobs via `/api/v2/schedules`
- Marketing automation: weekly Telegram post plan generated via Celery

---

### 2026-03-12 — Phase 16: Code Review & Stability

**Improvements:**
- All 118 code review issues resolved (16 critical, 20 major, 29 minor, 26 suggestions)
- Generation audit table (`generation_status_changes`) for tracking all status transitions
- SSRF protection via `url_validator.py` for all user-supplied URLs
- PII hashing in logs
- Idempotency keys on `/generate` to prevent duplicate charges
- Rate limit remaining header (`X-RateLimit-Remaining`) on all responses
- SSE heartbeat to prevent connection timeouts on slow generations
- Single-flight auth deduplication

---

### 2026-03-08 — Phase 15.1: Economy Patch

**Breaking changes:**
- All models now cost a minimum of **1 token** — no zero-cost models exist
- `is_free` flag removed; use `min_plan_slug` to check tier access
- `min_plan_slug = NULL` → Starter-tier (19 models); `min_plan_slug = "basic"` → Basic+ required

**Model access:**
- Starter-tier models (19): all text models at 1 tkn, `dall-e-3`, `photon-flash`, `luma-ray2-flash`, `musicgen`, `whisper`, `gpt-4o-transcribe`, `elevenlabs-scribe`, all 7 tools
- Basic+ (remaining 33 models): require paid subscription

**Rate limits updated:**
- Free plan: 20 req/hour, 150 req/day (previously 5/20)

---

### 2026-03-08 — Phase 15: Dynamic Pricing

**New features:**
- Dynamic pricing engine for media models — costs vary by parameters (size, quality, duration)
- New endpoint: `POST /api/v2/models/{slug}/calculate-price` — check token cost before generating
- Price ranges in model list endpoint (`min_token_cost`, `max_token_cost`)
- Output format preference per model stored in user settings (`webp` default)

**Model updates:**
- DALL-E 3 HD and GPT Image 1.5 HD merged back into base slugs — use `quality: "hd"` param instead
- MidJourney consolidated into single `midjourney` slug (previously split into v6/turbo variants)
- Kling consolidated into single `kling-v2.1` slug with dynamic pricing by quality/duration
- Luma Ray2 Flash consolidated into `luma-ray2-flash` slug (previously had resolution variants)

---

## February 2026

### 2026-02-15 — Phase 14: Web Platform & Features

**New features:**
- Web platform live at [elumenta.ru](https://elumenta.ru)
- Multi-auth: Telegram Login Widget, Google OAuth, email/password
- JWT authentication (access token 15min + refresh token 30 days)
- Public REST API access for VIP and Elite plan users
- Generation history with filtering and pagination
- MidJourney Upscale (U1–U4 buttons) via API
- ElevenLabs Sound FX (`elevenlabs-sfx`)
- Assistant Marketplace with moderation
- Support RAG chat via `@ElumentaSupportBot`
- API key management (`/api/v2/keys`)
- User and admin statistics endpoints
- PWA support with iOS viewport fix

**New endpoints:**
- `POST /api/v2/auth/telegram`
- `POST /api/v2/auth/register`
- `POST /api/v2/auth/login`
- `POST /api/v2/auth/google`
- `POST /api/v2/auth/refresh`
- `POST /api/v2/auth/logout`
- `GET /api/v2/user/me`
- `GET /api/v2/user/balance`
- `GET /api/v2/user/history`
- `GET /api/v2/user/stats`
- `POST /api/v2/keys` / `GET /api/v2/keys` / `DELETE /api/v2/keys/{id}`
- `POST /api/v2/schedules` / `GET /api/v2/schedules`

---

## January 2026

### 2026-01-20 — Model Catalog Expansion

**New models added:**
- `sora-2` (video, OpenAI)
- `veo-3.1` (video, Google)
- `midjourney-video` (video)
- `nano-banana-pro` (image, Google)
- `gpt-4o-transcribe` (STT, OpenAI)
- `elevenlabs-scribe` (STT)
- `grok-4-fast` (text, xAI, Starter-tier)
- `el-voice-changer`, `el-voice-isolator`, `el-dubbing`, `el-dialogue` (ElevenLabs audio tools)

---

## December 2025

### 2025-12-01 — Rate Limits & Streaming

- Elite plan: no rate limits
- VIP: 100/hour, 1,000/day + public API access
- Pro: 60/hour, 500/day
- Basic: 30/hour, 200/day
- SSE streaming for all LLM models: `POST /api/v2/generate/stream`
- Streaming events: `generation_start`, `content_delta`, `generation_end`

---
```