Silvasonic - Advanced Bioacoustic Monitoring

Architecture Outline

Built for Ecology & Conservation

Silvasonic closes the gap between fragile DIY scripts and highly expensive proprietary eco-acoustics equipment. By transforming standard hardware (Raspberry Pi 5 + NVMe) into a robust appliance, we empower researchers to gather vast amounts of continuous audio data flawlessly.

We believe that Data Capture Integrity is paramount. You should never miss a rare species vocalization just because a system crashed overnight or an SD card failed.

Active Core Capabilities

Resilient Capture Engine

Highly optimized FFmpeg pipeline capturing dual streams (Raw Lossless + Processed). Audio is buffered natively in RAM (ALSA) to eliminate SD-card thrashing. Backed by a strict software watchdog for zero-downtime auto-recovery.

Autonomous Orchestration

Built on an immutable, rootless Podman architecture. A central Python Controller manages ephemeral recording containers via strict state reconciliation. Redis Pub/Sub drives system-wide heartbeats.

Smart Storage Retention

Zero-trust data architecture. The Processor service instantly indexes every segment into TimescaleDB, while an autonomous Janitor enforces NVMe capacity restrictions to actively prevent "disk full" panics.

Hardware Auto-Enrollment

Plug in a USB microphone and the Controller does the rest. It extracts ALSA configurations, assigns matched audio profiles natively, and dynamically spawns isolated containers per device.

The Expansion Roadmap

v0.6.0 – v1.0.0

Phase I: Edge AI & Cloud Sync

Immutable Uploader via Rclone Automated ingestion pipeline that recompresses heavy raw WAVs into FLAC and pushes them concurrently to targets like Nextcloud, Amazon S3, or SFTP. Designed to handle patchy forest connections automatically with robust retry-logic.

BirdNET Avian Classification Instead of uploading terabytes over LTE, Silvasonic shifts the processing to the edge. The system pulls audio chunks to run BirdNET ML models locally. High-confidence detections are indexed as tiny CSV results instantly synced to the cloud.

Interactive Dashboard (HTMX) A blazing fast, API-driven Web Interface leveraging FastAPI and HTMX. Provides real-time container controls, configuration seeding, visual detection timelines, and instant database querying directly on the station.

v1.1.0+

Phase II: System Ecosystem

Icecast Live-Streaming Broadcast real-time Opus audio streams directly from the forest to researchers' headphones around the globe, allowing live listening capabilities right in the browser.

BatDetect & Climate Sensors Unlocking ultrasonic classification capabilities for bat activity correlation. Integrates BME680 (Climate) and SPS30 (Particulate) environmental hardware sensors directly into the timeseries database.

Big-Data Operations (Tailscale) Establishing native Tailscale VPN-mesh support for secure remote fleet access anywhere in the world, alongside scheduled daily Parquet snapshot exports for huge data lakes.

:: Spec Sheet

INFRASTRUCTURE	Rootless Podman Containers [Zero-Trust]
DB / PUB-SUB	TimescaleDB (PostgreSQL 16) / Redis
CORE ENGINE	Python 3.13, Asyncpg, FFmpeg, FLAC
HARDWARE	Raspberry Pi 5 + NVMe Solid State Drive

Technical Docs Source Code

Licensed under CC BY-NC-SA 4.0 // Built for the wild.