System Requirements

Honeyframe is packaged as Python services with static frontends, fronted by an nginx reverse proxy. The default deployment runs all services on a single Linux host.

Operating system

• Linux — Ubuntu 22.04 LTS or Debian 12 are tested. Any glibc-based distribution from 2023+ should work. Ubuntu 24.04 LTS is what the v0.0.27 release was validated against.
• macOS — supported for development only.
• Windows — not supported as a host. WSL2 works for development.

Architecture: x86_64 / amd64 only. The release tarball is linux-x86_64; ARM64 is not currently shipped as a compiled tarball. Build from source if you need ARM.

✦First-boot pause is normal

On a freshly-provisioned cloud VM, setup-customer.sh waits up to 10 minutes for unattended-upgrades to release the dpkg lock before running apt-get install. This was added in v0.0.27 so the install no longer fails with Could not get lock /var/lib/dpkg/lock-frontend. If you see no output for several minutes during stage 1 of the install, it is waiting, not hung. Tail /var/log/honeyframe-install/*.log to confirm.

Tiers

Honeyframe is split into three tiers, configured in install.conf under tiers:. Pick what you need — paas is always required.

Tier	What it is	Adds
paas	Hub Data Platform — connectors, flow builder, dashboards, APIs, chat	Always required
saas	Vertical App — a domain-specific UI built on top of PaaS (e.g. Hub Data Intel for healthcare, or any other industry vertical)	+1 systemd unit, +1 nginx vhost
iaas	Hub Cloud — server mgmt, DB provisioning, LLM, billing	+1 systemd unit, +1 nginx vhost, +PostgreSQL admin role

The hardware sizing below assumes a paas-only install. Add ~30% RAM and ~50% disk per additional tier.

Hardware

For a single-tenant install (one organization, one project, dozens of users):

Resource	Minimum	Recommended
CPU	4 cores	8 cores
RAM	8 GB	16 GB
Disk	50 GB	200 GB SSD
Network	100 Mbps	1 Gbps

The Python services collectively cap at ~1.5 GB resident under load (MemoryMax=1500M per systemd unit). The remaining headroom is for the database, dbt runs, and the Lakehouse (Parquet + DuckDB) cache.

Runtime dependencies

• Python 3.10+ (3.10.12 tested in production; 3.11+ recommended for forward-compat — google-api-core deprecates 3.10 on 2026-10-04).
• Node.js 20.x (the v0.0.27 installer pulls nodejs 20.20.2 from deb.nodesource.com/node_20.x. The build server uses Node 22 for Vite, but the runtime host only needs Node 20.)
• PostgreSQL 14+ (16 recommended). The platform stores all data — operational and observability — in a single honeyframe schema. (v0.0.27 used two schemas, hubstudio and dataintel; v0.0.29 folded both into honeyframe.)
• nginx 1.18+ (reverse proxy, TLS termination, rate limiting).
• systemd (services run as systemd units — see Custom systemd Units).
• Certbot (Let's Encrypt issuer; see TLS / Certificates).

Optional dependencies

• dbt 1.7+ — required if you intend to run transformation models. Honeyframe ships dbt templates under paas/dbt/.
• DuckDB 0.10+ — bundled via duckdb Python package; required for Lakehouse-backed datasets.
• Ollama or another LLM provider — required for Agent Builder, Knowledge Bases, and the SQL chat surface. See Connectors for the supported LLM connector list.

Network ports

Port	Service	Exposed externally
80, 443	nginx	Yes (HTTPS only after Certbot)
8001	Platform API	No (proxied)
8002	Cloud API	No (proxied)
5432	PostgreSQL	No (loopback or VPC-internal)

The reverse proxy is the only public surface. All API ports listen on 0.0.0.0 for ease of inter-service communication; in a multi-host install, restrict them to the private network with a firewall.

Database

A dedicated PostgreSQL instance is recommended. You choose the database name (database.name in install.conf); the platform creates a single schema inside it on first run:

• honeyframe — all platform data: tenants, users, projects, dashboards, audit logs, query history, lineage.

The connecting role needs CREATE on the database (for migrations) and USAGE + CREATE on the honeyframe schema after creation.

✦Migrating from v0.0.27

v0.0.27 created two schemas (hubstudio for operational data, dataintel for observability). v0.0.29 folds both into a single honeyframe schema. Greenfield installs see only honeyframe. Upgrades from v0.0.27 are handled by a migration that moves tables from hubstudio and dataintel into honeyframe and drops the empty schemas — see the v0.0.29 release notes.

For multi-tenant installs, allocate one Postgres database per tenant. Tenants do not share schemas across the same database.

Browser support

The frontend targets evergreen Chromium-based browsers (Chrome, Edge, Brave) and Firefox 110+. Safari is supported on macOS / iOS 16+. IE and legacy Edge are not supported.