Infrastructure, Not Organizer
File organizers solve for the present. They sort files into folders that make sense right now, according to categories that feel right today. When your needs change, the organizational scheme becomes a liability. You reorganize again.
fialr does not organize your files. It builds infrastructure around them.
The distinction
Section titled “The distinction”An organizer optimizes for immediate retrieval. It asks: “Where should this file go so I can find it?” The answer depends on how you think about your files today, which is a moving target.
Infrastructure optimizes for compounding value. It asks: “What properties does this file have, and how do I make those properties durable, queryable, and machine-readable?” The answer does not depend on your current workflow.
| Organizer | Infrastructure |
|---|---|
| Folder hierarchy is the system | Folder hierarchy is one possible view |
| Moving a file changes its identity | Moving a file changes its path metadata |
| Rename to find | Query to find |
| Manual maintenance | Self-describing corpus |
| Value decays as needs change | Value compounds as tools improve |
The compounding corpus
Section titled “The compounding corpus”When fialr processes a directory, the result is a corpus with four properties that compound over time:
Consistent structure. Every file follows the same naming convention, with semantic tokens that encode date, entity, descriptor, and version. This consistency is what makes the corpus queryable without a search index.
Cryptographic identity. Every file is identified by its BLAKE3 content hash. Names and paths change. The hash does not. Deduplication, integrity verification, and provenance tracking all derive from this single stable identifier.
Open formats. Metadata lives in SQLite and extended attributes. Configuration is TOML and YAML. Audit logs are JSON. Nothing requires a proprietary reader. Nothing requires fialr to be installed to access.
Documented logic. Every operation is logged to an append-only ledger. Every job produces a manifest, a plan, and a report. The corpus is not just structured — it is self-documenting.
These properties are not useful on their own. They are useful because they make the corpus a reliable input to systems that do not exist yet.
What this enables
Section titled “What this enables”A well-structured, hash-identified, semantically-named corpus is infrastructure for:
AI agents. An agent that can trust filenames to be semantic, metadata to be queryable, and content hashes to be stable can operate on a corpus without needing to understand its history. The naming convention is effectively a machine-readable API.
Local RAG and search. Structured metadata in SQLite is a retrieval layer without a vector database. Sensitivity tiers tell the search system which files it can index and which it cannot.
Automation pipelines. Shell scripts, cron jobs, and CI workflows can operate on a corpus where filenames follow a predictable pattern and integrity can be verified with a single command.
Audit and compliance. The append-only operation ledger, pre- and post-hash verification, and job-level provenance chain provide a complete audit trail without additional tooling.
Future applications. The value of infrastructure is that it serves applications you have not built yet. A corpus with stable identity, semantic naming, and documented provenance is ready for whatever comes next. This is the point.
Open design commitments
Section titled “Open design commitments”fialr makes specific commitments to prevent the corpus from becoming dependent on fialr itself:
| Commitment | Implementation |
|---|---|
| Open file formats | SQLite, TOML, YAML, JSON, CSV, Markdown |
| Documented schemas | schema.yaml is versioned, migrations are committed before files move |
| No vendor lock-in | Every metadata artifact is readable without fialr installed |
| Stable interfaces | Content hash as identity means external tools can reference files by hash |
| Plaintext configuration | fialr.toml and sensitivity.yaml are human-readable and editable |
| Platform portability | SQLite is the source of truth on every platform; XATTRs are a derived cache |
The corpus should outlive the tool that built it.