Capabilities & limits reference

Apps

Sources

container image OR git repo (Dockerfile or nixpacks buildpacks)

Build

sandboxed (rootless BuildKit / kaniko in gVisor); ~280s timeout. Heavy builds (e.g. next build ~4GB) may need NODE_OPTIONS=--max-old-space-size; OOM reports build_oom

Deploys

zero-downtime (new container promoted only if healthy) + rollback to exact image

Custom domains

CNAME/ALIAS + automatic TLS (Let’s Encrypt)

Env vars

encrypted at rest (envelope, AES-256-GCM), injected on redeploy

Volumes

persistent block storage per app (survives redeploys)

Isolation

gVisor syscall sandbox (untrusted code cannot escape)

Runtime

RAM/CPU limits, health checks, per-app logs; multi-runner with auto-failover

Databases

PostgreSQL versions

15, 16, 17 (default 17)

pgvector

pre-installed (embeddings / vector search, AI-native)

Extensions

pg_trgm, unaccent, uuid-ossp, pgcrypto, citext, hstore, pg_stat_statements — CREATE EXTENSION, or POST /databases/:id/extensions {name, schema?} to install into a specific schema (e.g. "extensions" for ex-Supabase apps)

High availability

leader + standby, automatic failover (Patroni + etcd)

Connection pooling

PgBouncer (incluido)

Logical replication / CDC

enabled — CREATE SUBSCRIPTION (in) + publications (out)

Roles

manage Postgres roles per database (least-privilege create/list/drop/reset-password) — POST /databases/:id/roles. Never superuser; passwords never hit argv/logs. Use for GRANT + SET ROLE (scoped access) + object ownership.

Data API

auto REST over your tables (PostgREST/Supabase-style) — GET/POST /databases/:id/data/:table with select/filter/order/limit. Read/write data over HTTP without a Postgres driver. Identifiers validated, values parameterized (injection-safe). AI-first.

Backups

automatic daily + on-demand, point-in-time restore (wal-g)

Read replicas

hot-scaling, separate read endpoint (readUri) on HA

Branching

branch the production database (full-copy clone at "now" or a PITR point) to test migrations safely — POST /databases/:id/branches. Reset a branch back to production’s current state (POST .../reset, connection unchanged), ephemeral TTL + delete protection. A branch is a real cluster (counts vs plan).

Password rotation

automatic scheduled password rotation (AWS Secrets Manager rotation substitute), ZERO-DOWNTIME. Alternating dual-role over a reloadable pooler → the old credential stays valid until the next rotation (overlap), so live apps never break. Enable: POST /databases/:id/rotation/migrate ONCE (switch the pooler to reloadable file mode), then PUT /databases/:id/rotation {intervalDays} (or POST .../rotation/rotate to rotate now). Apps wiring ${{db.x.DATABASE_URL}} auto-redeploy with the fresh credential.

TLS

verificado (sslmode=verify-ca con CA propia — sin rejectUnauthorized:false)

Cron jobs

Schedule

5-field cron with timezone/DST (cron-parser)

Types

container (docker run per job in gVisor) OR http (koigrid hits a URL on schedule — headers encrypted)

Runtime

RAM/CPU limits + timeout; captured logs + exit code

Output

captured logs + exit code; manual trigger available

Env vars

encrypted at rest, injected per run

Cache / Redis

Redis version

7

High availability

master + replica(s) + 3 Sentinels + master-aware HAProxy, auto-failover

Eviction policies

noeviction, allkeys-lru/lfu/random, volatile-lru

Persistence

RDB + AOF

Backups

RDB snapshots → S3, daily + on-demand, restore to a new cluster

Security

TLS (rediss://) + password (requirepass/masterauth)

Storage

S3-compatible

MinIO — PutObject, GetObject, ListObjectsV2, HeadObject, DeleteObject, presigned URLs

Clients (zero code change)

the standard AWS SDK works as-is (virtual-hosted bucket.s3.koigrid.com supported) — set AWS_ENDPOINT_URL_S3 + keys. Also aws-cli, rclone; path-style works too

Endpoint

https://s3.koigrid.com

Access keys

per-org scoped keys (secret shown once)

Bandwidth

included per plan, flat rate (no egress surprise)

Observability

Events

timeline of every action, per org (GET /events)

Metrics

daily consumption series + cost (GET /metrics)

Alarms

threshold alarms → emit alarm.triggered (GET/POST /alarms)

Anomaly detection

statistical spike detection (>Nσ from baseline, no fixed threshold) on cost/usage (GET/POST /anomalies)

Scope

observability:read / observability:write

Email

Model

bring-your-own SMTP (Resend/SES/Postmark/Postfix…) — koigrid relays only

Config

SMTP per org, encrypted at rest, verified on save

Sending

POST /email/send with rate limit per plan + delivery ledger

Cost

zero email cost/reputation for koigrid; no per-domain limits

Checks

Check types

URL / heartbeat + API check with assertions (v2: browser via Playwright)

Assertions

HTTP status (200/2xx/list), latency < X ms, body contains/regex/JSON path, header equals/contains

Scheduling

interval per plan (Free ≥5 min, paid ≥1 min) + run-once (test now)

Reliability

anti-flaky retries + per-check timeout; status up/down/degraded

Auth

encrypted custom headers (Authorization, x-api-key) for protected endpoints

Alerts

email + webhook (Slack/PagerDuty/Discord) on down and recovery — only on state change

History

per-run status + latency; uptime %, average + p95 latency

Agent-native

full CRUD + run-once by API/CLI — monitoring-as-code

Share

API: /api/v1/ · docs /docs · /share · full endpoints + scopes in /llms.txt and OpenAPI.

Links

API: /api/v1/ · docs /docs · /links · full endpoints + scopes in /llms.txt and OpenAPI.

Feature Flags

API: /api/v1/flags · docs /docs · /flags · full endpoints + scopes in /llms.txt and OpenAPI.

Status Pages

API: /api/v1/status-pages · docs /docs · /status-pages · full endpoints + scopes in /llms.txt and OpenAPI.

Heartbeats

API: /api/v1/heartbeats · docs /docs · /heartbeats · full endpoints + scopes in /llms.txt and OpenAPI.

Config & Secrets

API: /api/v1/config · docs /docs · /config · full endpoints + scopes in /llms.txt and OpenAPI.

Web Analytics

API: /api/v1/analytics/sites · docs /docs · /analytics · full endpoints + scopes in /llms.txt and OpenAPI.

Error Tracking

API: /api/v1/errors · docs /docs · /errors · full endpoints + scopes in /llms.txt and OpenAPI.

Speed Insights

API: /api/v1/speed-insights · docs /docs · /speed-insights · full endpoints + scopes in /llms.txt and OpenAPI.

Sandboxes

API: /api/v1/sandboxes · docs /docs · /sandboxes · full endpoints + scopes in /llms.txt and OpenAPI.

Image Optimization

API: /api/v1/images/config · docs /docs · /images · full endpoints + scopes in /llms.txt and OpenAPI.

Firewall

API: /api/v1/waf/rules · docs /docs · /firewall · full endpoints + scopes in /llms.txt and OpenAPI.

Webhooks

API: /api/v1/webhooks · docs /docs · /webhooks · full endpoints + scopes in /llms.txt and OpenAPI.

Log Drains

API: /api/v1/log-drains · docs /docs · /log-drains · full endpoints + scopes in /llms.txt and OpenAPI.

Spend Management

API: /api/v1/spend · docs /docs · /spend-management · full endpoints + scopes in /llms.txt and OpenAPI.

Bot Management

API: /api/v1/bots/policy · docs /docs · /bot-management · full endpoints + scopes in /llms.txt and OpenAPI.

OpenTelemetry

API: /api/v1/otel · docs /docs · /opentelemetry · full endpoints + scopes in /llms.txt and OpenAPI.

DNS

API: /api/v1/dns/zones · docs /docs · /dns · full endpoints + scopes in /llms.txt and OpenAPI.

Queues

Model

durable message queue on Postgres (SQS-like) — send / receive / ack

Delivery

at-least-once with a visibility timeout (received messages hidden until ack or timeout)

No double delivery

concurrent consumers claim different messages (FOR UPDATE SKIP LOCKED)

Dead-letter

built in — a message past maxReceives is dead-lettered (no separate DLQ/redrive policy)

Send options

delaySeconds (up to 15 min) + dedupeId (idempotent sends)

Limits

body up to 256 KB, receive up to 10 at a time (SQS-parity)

Lock-in

none — portable Postgres model, not the AWS-specific SQS API

Workflows

Model

durable execution — ordered steps with journaled outputs (Step-Functions-like activities / Inngest steps)

Durability

a run survives crashes and redeploys — completed steps are never re-run (journal in Postgres)

Durable sleep

pause a run from seconds to 90 days; it wakes and continues

Retries

per-step exponential backoff up to maxAttempts, then the run fails

At-least-once steps

a claimed step has a lease — if the worker dies, another re-claims it

Worker model

any worker/agent drives it: POST /next → run → complete/fail/sleep. No SDK lock-in

Lock-in

none — open API + Postgres, not Amazon States Language

Container Registry

Protocol

standard OCI Distribution API (/v2) — docker, crane, skopeo, buildkit all work

Auth

HTTP Basic — password = a koigrid token with registry:read / registry:write

Login

docker login koigrid.com -u token -p <koigrid-token>

Storage

blobs + manifests on koigrid Storage (S3/MinIO), content-addressed by digest

Isolation

images scoped per org (no cross-tenant access)

Deploy from it

push an image, then deploy it as a koigrid App — one platform

Lock-in

none — standard protocol + portable object storage

Audit Log

What

immutable record of who did what — every security/config change, per org (CloudTrail-like)

Actor

user (web/API) or agent (API token) — resolved to email/token name, with source IP

Coverage

curated from the platform event log — members, tokens, resources, deploys, DNS, firewall, billing, integrations

Filters

by category, action type, actor, resource type and date range

Export

CSV download (RFC-4180, formula-injection-safe) for compliance/SIEM

Access

owners and admins only (scope audit:read); reads never expose secrets

Lock-in

none — open API + CSV; portable, not a proprietary trail

Managed WordPress

API: /api/v1/wordpress/accounts · docs /docs · /wordpress · full endpoints + scopes in /llms.txt and OpenAPI.

Plan limits