NevTan Cloud
Changelog

Connect your knowledge bases directly to Vector RAG collections. Sync is incremental — only changed documents are re-embedded.

• NewGoogle Drive connector — sync folders to a RAG collection with OAuth2 auth
• NewNotion connector — sync databases and pages, with automatic re-sync on change
• NewConnector sync scheduler — set intervals from 15 min to daily
• ImprovedRAG query latency reduced by 18% via index pre-warming on collection open
• FixFixed: large PDF uploads (>50 MB) occasionally stalling at 99%

A focused patch release addressing billing display issues reported after the June 10 infrastructure update.

• FixFixed: invoice totals showing $0.00 for some Pro accounts after plan change
• FixFixed: Add Balance modal closing on backdrop click when typing an amount
• FixFixed: GPU instance list not refreshing after a new instance is launched
• ImprovedBilling period selector now shows exact start/end dates on hover

AI Agents can now route to any OpenAI-compatible endpoint — your own fine-tuned model, a private model server, or a third-party provider.

• NewCustom model endpoint field in Agent configuration — any OpenAI-compatible URL
• NewAgent execution history: full trace of tool calls, token counts, and latency per step
• NewTeam-scoped agents — share agent configs with teammates without sharing credentials
• ImprovedAgent list now shows last-run time and avg token cost per run
• ImprovedCode sandbox for testing agent prompts before deploying to production
• FixFixed: streaming responses cutting off at 4096 tokens when tool_choice was set

The NevTan monitoring agent now captures per-route HTTP metrics automatically when the Spring actuator is active.

• NewTop endpoints table — requests/sec and avg latency per route
• NewHTTP 4xx / 5xx error rate charts with per-route breakdown
• ImprovedMemory chart now shows heap used vs committed vs max in a single stacked view
• FixFixed: GC pause metric occasionally showing negative values on JDK 21

Full billing infrastructure: Stripe checkout, email alerts at 70/90/100% usage, and automatic plan upgrades at 80% memory.

• NewStripe Checkout integration — upgrade plans directly from the dashboard
• NewUsage alerts via email at 70%, 90%, and 100% of plan limits
• NewAuto-scale trigger at 80% memory — upgrades plan and prorates Stripe subscription
• NewInvoice history page — download PDF invoices for each billing period
• ImprovedResource usage bars now show daily trend sparkline

Full audit log with Merkle hash chain for tamper evidence. Exportable to CSV and drainable to external SIEM via webhook.

• NewAudit log with Merkle hash chain — every entry cryptographically linked to previous
• NewAudit log notarization — weekly SHA-256 root published to /audit/notarization endpoint
• NewTier-based retention: Hobby 30d, Pro 90d, Scale 365d
• NewCSV export filtered by date range, actor, or resource type
• NewSIEM drain via webhook — push audit events to Splunk, Datadog, or any HTTPS endpoint

Full platform relaunch: new console UI, team workspaces, fine-tuning, and the Vector RAG service.

• NewRedesigned console — Next.js 14 migration, 3× faster navigation
• NewTeam workspaces — invite collaborators, assign roles (Owner, Admin, Member, Viewer)
• NewFine-tuning jobs — LoRA adapters on Llama 3 and Mistral families
• NewVector RAG service — collections, documents, semantic + hybrid search
• NewAI Agents — define tools, connect to inference, run autonomously
• NewManaged Postgres v16 with pgvector 0.7, PostGIS, and pg_partman pre-installed