Installation

Requirements

macOS 12+ or Linux (glibc 2.31+)
No runtime dependencies required

Download

Purchase a license at fialr.com/licensing to receive a download link and license key.

After purchasing, download the binary for your platform:

Platform	File	BLAKE3
macOS (Apple Silicon)	`fialr-1.0.0-macos-arm64`	`a1b2c3d4...`
macOS (Intel)	`fialr-1.0.0-macos-x64`	`e5f6a7b8...`
Linux (x64)	`fialr-1.0.0-linux-x64`	`c9d0e1f2...`

Full checksums and detached signatures are published with each release.

Verify checksum

b3sum fialr-1.0.0-macos-arm64

Compare the output against the BLAKE3 hash listed above.

Verify signature

Each binary ships with a detached Ed25519 signature (.sig file). Verify with the fialr public key:

minisign -Vm fialr-1.0.0-macos-arm64 -P RWSGmBqXfRPhKPRMN4E+JGuhMGZ6NlEiXPNnA8tY6VDz

The public key is also published at fialr.com/signing-key.

Install

macOS / Linux

chmod +x fialr-1.0.0-macos-arm64
sudo mv fialr-1.0.0-macos-arm64 /usr/local/bin/fialr

Activate

On first run, fialr requires a one-time activation with your license key:

fialr activate YOUR-LICENSE-KEY

This contacts api.fialr.com once to validate the key and bind it to the machine. After activation, fialr operates fully offline. No further network calls are made.

Each license allows activation on up to 3 machines.

Verify

Confirm the installation:

fialr --version

fialr 1.0.0 (build a1b2c3d4)

Optional: AI dependencies

For AI-assisted enrichment, semantic search, and vector embeddings, install Ollama and pull the required models:

# Install Ollama
brew install ollama    # macOS
# See ollama.com for Linux

# LLM for enrichment inference (~2 GB)
ollama pull llama3.2

# Embedding model for semantic search and similarity (~275 MB)
ollama pull nomic-embed-text

Both models require approximately 2.3 GB of disk space total. Ollama downloads them once and caches locally.

fialr runs without Ollama, but the following features require it: enrichment (AI metadata generation), vector embeddings, semantic search, and embedding-based near-duplicate detection. Without local models, these features are unavailable unless you configure a cloud provider.

Cloud-only alternative

If disk space is a constraint, you can skip the local models and use a cloud provider (Claude API) for enrichment instead:

fialr config ai --provider claude --key sk-ant-...

Cloud enrichment works for Tier 2–3 files. Tier 1 files require local AI by default (cloud access needs explicit two-step confirmation). Vector embeddings and semantic search currently require local Ollama — there is no cloud embedding provider.

OCR for scanned documents

To extract text from scanned PDFs and images, install Tesseract:

brew install tesseract    # macOS
sudo apt install tesseract-ocr    # Debian/Ubuntu

The ocrmypdf Python package (included in fialr[enrichment]) handles the OCR pipeline. Without Tesseract, scanned documents produce no extracted text and enrichment falls back to filename-only metadata. Native PDFs, Office documents, photos, and audio files are unaffected.