open-fdd 3.1.5

Arrow-native HVAC fault detection runtime — lint, test, and run apply_faults_arrow rules on columnar telemetry (PyPI). Use GHCR Docker images for the full edge operator stack.

Open-FDD

Open-source supervisory fault detection for buildings — Arrow-native rules, optional DataFusion SQL rules for Rust-ready migration, optional PyPI embeddable runtime, and a full Docker/GHCR edge operator stack (BACnet, bridge API, dashboard, MCP).

---

Install / run

Full Open-FDD edge stack (Docker / GHCR)

Use GHCR (GitHub Container Registry — GitHub’s OCI image registry at ghcr.io) for BACnet polling, Operator Bridge API, React dashboard, historian, and MCP sidecar in a docker pull workflow.

Image	Role
ghcr.io/bbartling/openfdd-bridge	API, dashboard, historian
ghcr.io/bbartling/openfdd-commission	BACnet discover, read, poll
ghcr.io/bbartling/openfdd-mcp-rag	MCP + doc-search

New edge host (no git clone — pulls latest from GHCR):

curl -fsSL -o /tmp/openfdd_edge_bootstrap.sh \
  https://github.com/bbartling/open-fdd/raw/refs/heads/master/scripts/openfdd_edge_bootstrap.sh
bash /tmp/openfdd_edge_bootstrap.sh --start

Update an existing site (backup workspace, pull latest containers from GHCR, recreate stack):

cd ~/open-fdd
./scripts/openfdd_site_backup.sh
./scripts/openfdd_site_update.sh

Python package (PyPI)

Use PyPI when you only need the embeddable Arrow-native FDD runtime — lint, test, and run rules in your own pipelines (cloud, IoT, notebooks) without Docker.

pip install open-fdd

import pyarrow as pa
import pyarrow.compute as pc

from open_fdd.arrow_runtime import run_arrow_rule


def high_sat(table, cfg, context=None):
    return pc.greater(table["SAT"], float(cfg["high"]))


table = pa.table({"SAT": [70.0, 90.0]})

result = run_arrow_rule(high_sat, table, {"high": 85})

print(result.true_count)  # 1

DataFusion SQL (same telemetry table, optional pip install 'open-fdd[datafusion]'):

from open_fdd.arrow_runtime import run_datafusion_sql_rule

SQL = """
SELECT
  *,
  "SAT" > 85.0 AS fault
FROM telemetry
"""

result = run_datafusion_sql_rule(SQL, table, {"min_true_rows": 5, "poll_interval_s": 60})

print(result.true_count)  # 1 — same confirmed count as PyArrow when cfg matches

Rule config fields such as min_elapsed_minutes and min_true_rows apply to both backends (fault confirmation / minimum duration). See fault confirmation.

---

Develop

git clone https://github.com/bbartling/open-fdd.git && cd open-fdd
python -m venv .venv && source .venv/bin/activate
pip install -e ".[test,dev,analytics]"
pytest open_fdd/tests -q

Contributor layout: AGENTS.md and developer docs.

---

License

MIT — see LICENSE.