Bermuda Tuner

Bermuda Tuner is a separate HACS integration that helps beginners audit and tune Bermuda from inside Home Assistant. Its calculations are deterministic; optional AI only explains a redacted report through a Home Assistant Conversation agent selected by the user.

What it does

• Setup audit: flags stale scanner relationships, weak overlap, and unstable RSSI.
• One-metre calibration: uses the median and spread of Bermuda RSSI history to recommend ref_power.
• Distance calibration: calculates attenuation from a measured test beyond one metre; 4-6 m is a useful practical range.
• Scanner balancing: aligns per-scanner medians and recommends RSSI offsets.
• Walk test: reports nearest and second-nearest scanners, signal margin, and ambiguity.
• Plain-English settings: explains smoothing, velocity, radius, updates, and away timeout.
• Apply and rollback: previews supported option changes, snapshots Bermuda options, and restores the latest snapshot.
• Diagnostics: exports a redacted report without MAC addresses or IRKs.

Requirements

• Home Assistant 2024.6 or newer
• Bermuda installed and configured
• HACS for the recommended installation method

Installation

1. In HACS, add https://github.com/ComputerWhisperers/bermuda_tuner as a custom integration repository.
2. Download Bermuda Tuner and restart Home Assistant.
3. Open Settings > Devices & services > Add integration and select Bermuda Tuner.

Guided workflow

Open Developer tools > Actions, search for Bermuda Tuner, and run the actions in this order. Every analysis action returns structured response data that Home Assistant can display or capture in a script response variable.

1. Run Setup audit.
2. Put a beacon one metre from a scanner and run One-metre calibration.
3. Move it to a measured distance beyond one metre and run Distance calibration.
4. Keep the beacon still and run Balance scanners.
5. Walk between areas while repeatedly running Walk test.
6. Run Preview settings, inspect the changes, then run Apply settings.
7. Use Roll back to restore the previous snapshot if behavior gets worse.

Example script:

sequence:
  - action: bermuda_tuner.audit
    response_variable: audit
  - action: persistent_notification.create
    data:
      title: Bermuda audit
      message: "{{ audit.summary | join(' ') }}"

Supported option patch

preview and apply accept an object containing only Bermuda's documented tuning keys:

action: bermuda_tuner.preview
data:
  settings:
    ref_power: -59
    attenuation: 2.7
    smoothing_samples: 15
response_variable: preview

Supported keys are ref_power, attenuation, rssi_offsets, smoothing_samples, max_velocity, max_area_radius, update_interval, and devtracker_nothome_timeout.

AI and privacy

AI is off by default. The tuner performs all measurements and recommendations locally. When Ask AI is invoked, it creates an audit, removes MAC addresses and 32-character IRKs again as defense in depth, and sends only that report to the selected Home Assistant Conversation agent. The tuner stores no AI transcript and has no API key setting.

The selected Conversation provider may retain data under its own policy. Use a local Conversation agent when no cloud transmission is acceptable. Diagnostics never invoke AI.

Safety

Apply updates only documented Bermuda options and creates a local snapshot first. It does not write Bermuda internals or Bluetooth identifiers. Review every preview, change one variable at a time when possible, and use rollback if tracking quality declines.

License

MIT