BirdNet-Behavior v0.1.0

First public release. BirdNet-Behavior is a ground-up Rust rewrite of
BirdNET-Pi that ships as a single static binary for Raspberry Pi and
x86_64 Linux.

Added

Core detection pipeline

• Pure-Rust audio pipeline with symphonia (decode), rubato (resampling),
  and realfft (mel spectrogram) — zero C dependencies in the audio path.
• ONNX Runtime inference through the ort crate, statically linked into
  release binaries. BirdNET+ V3.0 is the default model; BirdNET V2.4 FP16
  and V1 remain compatible.
• File-watcher detection daemon with configurable chunking, overlap,
  sensitivity, per-species confidence thresholds, and privacy filtering.
• Audio quality pre-filtering: SNR estimation, spectral flatness,
  adaptive noise-floor tracking, and rain / wind detection.
• Species occurrence frequency filter driven by the BirdNET metadata
  model, with whitelist, include, and exclude lists.
• Rare-bird quarantine workflow: detections that fall below per-species
  thresholds are quarantined for manual review rather than dropped.

Audio capture

• ALSA, PulseAudio, PipeWire, and RTSP capture sources, each supervised
  as a restart-aware subprocess with gap detection and disk monitoring.
• Multiple simultaneous RTSP streams via --rtsp-urls.
• Solar-aware recording scheduler with sunrise / sunset computation,
  twilight offsets, fixed-window schedules, and a night-inhibit mode.
• tmpfs support for transient audio storage to reduce SD card wear on
  Raspberry Pi deployments.
• Automatic disk management: per-species retention caps, auto-purge, and
  configurable disk-usage thresholds.

Storage and resilience

• SQLite operational database with WAL mode, ten idempotent schema
  migrations, integrity checks, hot backup, restore, and auto-recovery.
• Per-IP rate limiter on API and admin routes (token-bucket with
  Retry-After header).
• HTTP Basic Auth with constant-time comparison.

Web server and dashboard

• axum HTTP server with REST API, WebSocket, Server-Sent Events, and
  server-rendered HTMX pages. No client-side JavaScript framework.
• HTMX pages: dashboard, today, history, species list, species detail,
  species gallery, life list, activity heatmap, correlation, charts,
  weekly report, recordings browser, audio player, livestream, kiosk,
  notification center, quarantine, system health, and weekly report.
• Admin panel: settings editor, species thresholds, species filter
  tester, BirdNET-Pi migration wizard, system info, backup management,
  live log viewer (SSE), notification history, alert rules, data
  quality dashboard, and binary update check.
• Full dark / light theme support with OS preference detection.

Analytics (optional analytics feature)

• DuckDB behavioral analytics: sessionize, retention, funnel, sequence,
  and next-species prediction, implemented via the duckdb-behavioral
  extension.
• Phenology analytics: migration timing percentiles, weekly abundance
  index, peak weeks, monthly totals, species richness, and
  effort-corrected abundance.
• Time-series analytics: activity, diversity (Shannon), trend, peak,
  gap, and session windows (tumbling, sliding, hopping, session).

Integrations

• BirdWeather detection and soundscape uploads with retry and backoff.
• Apprise notifications across 80+ channels with per-species cooldown,
  watchlist, and template rendering.
• SMTP email alerts via lettre with rustls TLS (no OpenSSL).
• Wikipedia species image cache with on-disk and in-memory indexing.
• Pure-Rust MQTT 3.1.1 publisher (no external broker library) with
  Home Assistant auto-discovery.
• GitHub Releases auto-update with atomic binary replacement.
• Heartbeat URL pinging for uptime monitors.

Migration

• Non-destructive BirdNET-Pi import wizard. Source database is opened
  read-only. Transactional, idempotent, with pre- and post-migration
  species reports and a data quality report.
• Supports both BirdNET-Pi SQLite databases and BirdDB.txt CSV flat
  files.

Observability and deployment

• Prometheus metrics endpoint (/api/v2/metrics).
• tracing-based structured logging with SSE log streaming.
• Multi-architecture Docker images published to GHCR (linux/amd64,
  linux/arm64), with and without the analytics feature.
• Cross-compiled release binaries for aarch64-unknown-linux-gnu and
  x86_64-unknown-linux-gnu. The ort crate does not ship prebuilt
  ONNX Runtime binaries for armv7-unknown-linux-gnueabihf, so 32-bit
  ARM is not supported — Pi 3 / Pi Zero 2W users should install the
  64-bit Raspberry Pi OS, or build from source.
• Release binaries are built on Ubuntu 24.04 (GCC 13, glibc 2.39) to
  match the libstdc++ and glibc baselines that pyke's prebuilt ONNX
  Runtime archives require. Runtime requirement: glibc >= 2.39
  (Raspberry Pi OS Trixie, Debian 13, Ubuntu 24.04, or newer).
• systemd installer script with ALSA microphone auto-detection and
  automatic BirdNET+ model download from Zenodo.

---

Installation

One-line installer (Raspberry Pi / bare-metal Linux)

VERSION=0.1.0 bash <(curl -fsSL https://raw.githubusercontent.com/tomtom215/BirdNet-Behavior/main/install.sh)

Download a binary directly

Target	Platform
aarch64-unknown-linux-gnu	Raspberry Pi 4 / 5, ARM64 Linux
x86_64-unknown-linux-gnu	Standard 64-bit Linux

Archives are gzipped tarballs containing the stripped binary,
README.md, CHANGELOG.md, LICENSE, LICENSE-UPSTREAM, and
install.sh. Binaries require glibc 2.39 or newer
(Raspberry Pi OS Trixie / Debian 13 / Ubuntu 24.04).

Docker

docker pull ghcr.io/tomtom215/birdnet-behavior:0.1.0
docker pull ghcr.io/tomtom215/birdnet-behavior:0.1.0-analytics

Multi-arch manifests cover linux/amd64 and linux/arm64.

Verification

Checksums

sha256sum -c SHA256SUMS --ignore-missing

SLSA build provenance

Every release archive carries a signed SLSA build provenance
attestation produced by this workflow run. Verify it with the GitHub CLI:

gh attestation verify \
  --repo tomtom215/BirdNet-Behavior \
  birdnet-behavior-0.1.0-aarch64-unknown-linux-gnu.tar.gz