Skip to main content

The filesystem is the schema: a folder-as-type typed-knowledge layer for Markdown.

Project description

tropo

A typed-knowledge layer for any folder of Markdown. The filesystem is the schema: a document's type is the folder it lives in, and its metadata is only what can't be derived from where it sits and what it says.

tropo is the elegant successor to a frontmatter typechecker. Where the old model made every file pay a ceremony tax — type:, created:, updated:, slug: hand-declared on all of them — tropo derives all of that and asks you to write down only the irreducible signal. A clean note can have zero frontmatter and still be fully typed and valid.

Status: working engine (v0.2). tropo.py implements spec v1 end-to-end — folder-as-type resolution, derivation, validation, packs, overlays, the signal report, fix (de-noise), init, the graph layer (graph/blast/view/plan), and the data layer (query plus file → embedded migration). Cloud adapters are future 0.3.x work. An agent can drive the whole thing via .claude/skills/tropo/SKILL.md. See SPEC.md.

Quickstart

python tropo.py init my-vault                   # scaffold a tropo.toml (--packs dev-project)
python tropo.py types  --root examples/vault    # the resolved type registry
python tropo.py check  --root examples/vault    # validate — opinionated: warnings fail too (--lenient to relax)
python tropo.py signal --root examples/vault    # print ONLY the irreducible metadata
python tropo.py graph  --root examples/vault    # emit typed nodes + edges
python tropo.py view   --root examples/vault --out examples/vault/graph.html
python tropo.py fix    --dry-run                 # preview redundant-frontmatter removal
python tropo.py query "meeting notes" --root examples/vault   # search the graph
python tropo.py migrate --from file --to embedded --root my-vault --yes  # switch to LanceDB
python tests/test_tropo.py                       # run the test suite

Requires Python 3.11+ (stdlib tomllib), zero third-party dependencies for the core. Optional extras: pip install vivary-tropo[embedded] for LanceDB full-text search. Cloud extras are reserved for the 0.3.x adapter work.

Built-in packs are embedded in the single-file engine, so installed wheels can resolve starter packs without a repo-local packs/ directory. Workspace-local .tropo/packs/<name>.toml files still take precedence.

TOML config and frontmatter parsing tolerate a single leading UTF-8 BOM, which keeps Windows-created files from failing to load or being misread as body-only documents. tropo view --out keeps generated HTML under the tropo root, rejects symlink targets, and replaces output files without mutating hard-linked files outside the workspace.

Overlays — tighten a subtree

Drop a tropo.toml in any subdirectory to add stricter rules for that subtree only (a new required field, a narrowed enum, a nested type). It may only add constraints, never remove them — so you can always reason about a document top-down. See examples/vault/projects/tropo/tropo.toml, which requires every decision under that project to record its deciders.

The one idea

people/jeff.md          →  type = person   (the folder says so)
projects/tropo/README.md →  type = project  (nearest registered ancestor)
meetings/2026-06-12.md  →  type = meeting

No type: field. No hand-written dates. The path carries the type; git and the filesystem carry the dates; the first # H1 carries the title. What's left in frontmatter is the handful of fields that are genuinely irreducible — a person's relationship, a meeting's attendees, a decision's status.

Frontmatter is the exception, not the rule.

Before / after

A person note, the old way:

---
type: person
created: 2026-06-12
updated: 2026-06-12
slug: jeff
title: Jeff
relationship: self
---
# Jeff

The tropo way — same information, no noise:

---
relationship: self
---
# Jeff

type comes from people/. created/updated come from git. slug comes from the filename. title comes from the H1. Only relationship is irreducible, so only relationship is written down.

Why it's not just a second-brain tool

The config resolves by walking up the tree — like git, tsconfig, or pyproject.toml. Drop one tropo.toml at a repo root and that repo gains a typed-knowledge layer: decisions/, runbooks/, specs/, adr/ become enforceable document types with derived metadata and CI-checkable rules. Same engine, same grammar, whether it's a personal vault or a codebase's docs/.

tropo is a portable convention plus a tiny engine, not anyone's particular vault. Types ship as composable packs; a subfolder can overlay tighter rules without redefining anything.

Graphify-friendly repo maps

Use the repo-graph pack when a repository or control vault wants Markdown notes that double as Graphify-readable nodes:

packs = ["repo-graph"]

It defines folder-backed types for modules/, changes/, decisions/, verification/, and gates/, while allowing explicit graph fields such as id, related_modules, related_changes, and verification. This is a deliberate bridge: Tropo validates the node shape, Markdown/wiki links remain human-readable, and Graphify can index the relationships.

For multi-agent coordination, add the opt-in coordination pack:

packs = ["repo-graph", "coordination"]

It declares top-level assignee = "string" without changing the default workspace schema. exo claim <id> --agent <handle> uses that field for graph-native work ownership.

Design tenets

  • Signal over noise. If a value can be derived, never make a human write it.
  • Location is type. The directory tree is the type hierarchy.
  • Tighten, never loosen. Overlays and packs may add constraints, not remove them.
  • Zero-dependency, CI-clean. The engine stays a single file with an honest exit code, the one virtue worth keeping from its predecessor.

Layout

tropo/
├─ README.md            you are here
├─ SPEC.md              the normative model + config format reference
├─ tropo.toml            the project's own (dogfooded) config
└─ examples/
    └─ vault/           a tiny tree showing zero-frontmatter notes

License

MIT — see LICENSE.


Website & docs: https://vivary.vercel.app/

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

vivary_tropo-0.2.3.tar.gz (35.6 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

vivary_tropo-0.2.3-py3-none-any.whl (25.5 kB view details)

Uploaded Python 3

File details

Details for the file vivary_tropo-0.2.3.tar.gz.

File metadata

  • Download URL: vivary_tropo-0.2.3.tar.gz
  • Upload date:
  • Size: 35.6 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.14.3

File hashes

Hashes for vivary_tropo-0.2.3.tar.gz
Algorithm Hash digest
SHA256 a96760a4d0f5fe279b72856a48e4ed671521ed077f767836924e653f6641055d
MD5 971e82b6347c5330bd9491f33c980793
BLAKE2b-256 a6146a5619b987872ea81850247da8f24a6607d0278ad3dd6de9c64681ab4ef7

See more details on using hashes here.

File details

Details for the file vivary_tropo-0.2.3-py3-none-any.whl.

File metadata

  • Download URL: vivary_tropo-0.2.3-py3-none-any.whl
  • Upload date:
  • Size: 25.5 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.14.3

File hashes

Hashes for vivary_tropo-0.2.3-py3-none-any.whl
Algorithm Hash digest
SHA256 2df5710590405508949bd45ea6c1173ee43a4894e086bdeb766b3cd1bd5c58c6
MD5 74af3ba5071e79a5865810287574ad1f
BLAKE2b-256 d9da5d1e0e6545b5449c7d9ffca5bba670d713124f7e36b886d76d27e60b45e5

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page