~/changelog
Changelog
What shipped, when. New entries land as features do.
-
Default spend caps and hosting grace
- New accounts start with a $50 spend cap and pause-on-cap.
- A spend cap or a zero balance pauses the AI Gateway and agents.
- Hosting keeps running while you have balance, and gets a 7-day grace period at zero — with nothing deleted.
-
Agents in the catalog
- The Agents product is now visible in the catalog under AI Cloud, with pooled durable state, schedules, tasks, and runs.
- Functions (serverless) rates are published on the pricing page.
-
Reply to tickets by email
- Reply to a support ticket straight from the notification email and it threads back into your ticket.
- Staff replies arrive as branded email plus an in-app notification.
-
WHOIS privacy included
- WHOIS privacy is now included free with every domain registration and transfer, on by default.
-
AI SDK and per-epoch fine-tuning
- The @setcloud/ai package provides typed chat, streaming, and model helpers.
- A maintained LangGraph starter connects serverless agents to SetCloud AI Gateway and pooled PostgreSQL through SET_AGENT_POSTGRES_URL.
- Fine-tuning added a one-time estimated training charge, with no hourly hosting fee for deployed custom models. That preview estimate was superseded by accepted job quotes and custom-model storage pricing in October; see the current pricing page.
-
Serverless agents
- Pick the AI Agent runtime at project creation to deploy an agent that scales to zero.
- Durable sessions survive restarts and resume where they left off.
- The Agents workspace includes Builder, Tools, Knowledge, Triggers, Schedules, Tasks, and Runs.
- Schedule recurring agent prompts without standing infrastructure.
- An always-warm toggle keeps latency-sensitive agents hot.
- Ship from the command line with setcloud projects deploy --agent.
- Per-request metering draws from your existing credit balance.
-
Sign-In Protection and Connected Credentials
- Sign-In Protection gates any deployment behind your own identity provider.
- Connected Credentials vends scoped, short-lived third-party credentials into deployments without hardcoding secrets.
- Hardened outbound-fetch validation on client-supplied URLs.
-
Spend management
- Spend caps with pause-on-cap for agents, tasks, schedules, and AI Gateway calls.
- The deploy pipeline checks your spend state before starting a build.
- Always-warm capacity is billed and respects the cap.
-
Deploys that resume themselves
- An interrupted deploy now resumes from the last completed phase instead of failing outright.
- Resumed deploys never re-run finished build steps or bill you twice.
-
Scoped platform tokens
- Platform tokens can now be scoped to specific capabilities instead of granting full access.
- Every deployment automatically receives its own scoped token, injected at deploy and revoked when the deployment is deleted.
-
Tool calling and embeddings
- Chat completions now support tool calling — tools, tool_choice, and streamed tool calls, so OpenAI SDKs and the agent frameworks built on them work unchanged.
- New POST /v1/embeddings endpoint: OpenAI-compatible text embeddings at 256, 512, or 1024 dimensions, float or base64, $0.0226 per 1M tokens.
- Truthful finish_reason on every response (stop, length, tool_calls, content_filter), invalid requests now return 400, and capacity limits return 429 with Retry-After.
-
Usage-based databases
- One pooled PostgreSQL product with automatic scaling and a SetCloud-funded always-ready floor.
- Per-database metering of compute-hours, storage, and I/O at the published rates.
- Shows up as the Managed Databases line on your monthly invoice.
-
Deploy templates
- Clone licensed, SetCloud-maintained project and agent starters into your own GitHub account.
- Generated repositories use the normal project pipeline after you grant the SetCloud GitHub App access.
-
Pay with credits at checkout
- Domain checkout can charge your prepaid credit balance directly when it covers the total.
- Manual credit top-ups from the dashboard, $50 minimum.
- Auto-reload can be switched off without losing your configured threshold and amount.
-
Premium domain orders
- Premium and aftermarket domains now require identity verification before checkout.
- Premium orders are placed by our team after payment; you're notified at each step, and failed orders are refunded to your original payment method — card or credits.
-
Usage and invoices, grouped by product
- The Usage page groups consumption by product instead of raw line items.
- Invoice line items carry product names and final prices.
- Billing history labels monthly roll-ups, domain renewals, and refunds.
-
Databases at deploy time
- Add a serverless database when you deploy.
- The connection string is generated, stored encrypted, and injected into your app as DATABASE_URL — no copy-paste.
-
Fine-tune your own models
- Upload a CSV of prompt/response pairs and fine-tune a model on it from the dashboard.
- Finished models get a custom- model id callable from the playground and the API, like any built-in model.
- Job status is tracked from upload to ready.
-
Push to deploy
- Connect a GitHub repository and every push builds and deploys automatically.
- Build progress streams live to the dashboard.
- Roll back to a previous deployment in one step.
-
Secrets from the CLI
- Manage Secrets Vault values, versions, and tags from the setcloud CLI.
- Current availability: scheduled rotation, replica changes, and custom policy grants are unavailable; use project links for runtime access.
- setcloud secrets generate-password creates strong random passwords without leaving the terminal.
-
A terminal in your dashboard
- Open a shell in a running container straight from the dashboard.
- Every container operation is recorded in an audit log.
-
Domain search and WHOIS for everyone
- Search domain availability and run WHOIS lookups without an account.