Open-FDD

Open-FDD is a local-first Rust edge platform for building telemetry, semantic modeling, supervisory fault detection, and HVAC analytics.

Open-source building analytics for operators, integrators, and engineers who need vendor-neutral fault detection on premises — without sending BACnet traffic or historian data to the public internet.

Field to cloud

Secure HVAC data collection, fault detection, and portfolio access for mechanical systems integrators.

Open-FDD field-to-cloud architecture — BACnet/Modbus at the building, MQTTS to the hub, HTTPS/JWT for office and AI agents
Path Protocol
Building OT → field edge BACnet / Modbus (+ Haystack model)
Field → hub MQTTS
Office browser HTTPS → React SPA
AI / automation agents JWT REST (+ optional MCP)
Weather assist HTTPS Open-Meteo fetch at the edge
Offline import CSV / JSON package

What it does

  • Collects live data from BACnet, Modbus, Haystack, JSON API, and CSV imports
  • Stores telemetry in an Apache Parquet historian (Arrow/DataFusion)
  • Models sites with Project Haystack semantics
  • Runs DataFusion SQL rules (sql_rules/ registry)
  • Serves a React SPA — Overview, FDD, RCx, findings against central /api only
  • Exposes JWT REST and optional MCP for engineering agents

Pandas recipes stay in the Pandas cookbook for notebooks / PyPI. Production FDD is DataFusion SQL only.

Who it is for

Role Typical use
OT / BAS integrators Live edge on LAN/VPN — BACnet/Modbus, historian, FDD
Energy / RCx engineers CSV analytics on a workstation — import, SQL rules, reports
Developers & agents API + MCP for commissioning and validation

Deployment models

Live OT edge

Linux edge host on the building LAN. Pull GHCR images, bind-mount workspace/, keep API on loopback or behind a reverse proxy.

Offline engineering

Same stack in Docker on a laptop. Import CSVs, map Haystack roles, run SQL FDD — no live field bus required.

Intended for LAN, VPN, or OT-network deployment. Do not expose the API on the public internet without an independent security review.

Get started

  1. Quick Start — Compose + Railway + GHCR
  2. Rule Cookbook — DataFusion SQL + Pandas patterns (+ anomaly screening)
  3. Haystack modeling — packages and SQL roles
  4. Architecture — services and data flow
  5. API & Security — REST + vulnerability reporting
  6. MCP & Agents — Cursor / OpenClaw
  7. PyPI agent tools — open-fdd-anomaly report from mapped roles (local CSV, Railway or self-hosted Open-FDD, or a future vendor API) plus ECM workbooks (pip install open-fdd)
  8. Web App — SPA routes and RCx examples

Stack images 

ghcr.io/bbartling/openfdd-central:${OPENFDD_IMAGE_TAG:-nightly}
ghcr.io/bbartling/openfdd-web:${OPENFDD_IMAGE_TAG:-nightly}
ghcr.io/bbartling/openfdd-fieldbus:${OPENFDD_IMAGE_TAG:-nightly}
ghcr.io/bbartling/openfdd-mqtt:${OPENFDD_IMAGE_TAG:-nightly}
ghcr.io/bbartling/openfdd-mcp:${OPENFDD_IMAGE_TAG:-nightly}

See Build recipes and Release channels.