paraiso

An open, dependency-free Python foundation for the PARAISO organization framework: one calm home for everything on your mind.

PARAISO extends Tiago Forte's PARA method (Projects, Areas, Resources, Archive) with three additions — an Inbox for raw capture, Seeds for not-yet ideas, and Objectives for direction:

P  Projects    Things with a finish line.
A  Areas       Ongoing parts of life you maintain.
R  Resources   Reference you might want later.
A  Archive     Done or dormant, and still findable.
I  Inbox       Where every raw capture lands first.
S  Seeds       Ideas you let grow, with no pressure.
O  Objectives  The direction you're moving toward.

This package is not tied to any app. It's a shared, framework-first base other tools can build on. Two principles are baked in:

1. The user decides. Capture is separate from sorting; nothing is filed until a person says so.
2. AI is bring-your-own. Classification is an optional seam you implement yourself — the core needs no model, no network, and no dependencies.

Install

pip install paraiso

Requires Python 3.9+.

Quickstart

paraiso is both an importable library and a command-line tool — the same core, two front doors.

As a library:

import paraiso

p = paraiso.Paraiso("Personal")
c = p.capture("call the dentist tomorrow")               # lands in the Inbox
p.file(c, "project", title="Book dentist appointment")   # you decide where it goes

As a CLI (state persists under ~/.paraiso):

paraiso new personal
paraiso capture "call the dentist tomorrow"
paraiso inbox
paraiso file <cap_id> --bucket project --title "Book dentist appointment"
paraiso show

The library is in-memory — you choose when to persist, with Store. The CLI saves automatically and remembers the active workspace, so you can keep several paraisos and switch between them. The rest of this README goes deeper on each.

Library

import paraiso

p = paraiso.Paraiso("Personal")

# Capture first — no decision required. Everything lands in the Inbox.
c = p.capture("someday: write an essay about calm systems")

# Sort later, when you're calm. You decide the bucket, Area, etc.
writing = p.create_area("Writing", tags=["essay"])
item = p.file(c, "seed", title="Essay: calm systems", area=writing)

p.summary()
# {'name': 'Personal', 'inbox': 0, 'items': 1,
#  'buckets': {'Projects': 0, 'Resources': 0, 'Seeds': 1, 'Archive': 0}, ...}

The four fileable buckets are project, resource, seed, archive. Areas and Objectives are cross-cutting: an item may belong to an Area and advance an Objective, but you never file into either.

Multiple workspaces

Keep several paraisos (say work and personal) and switch between them. They persist as plain JSON under ~/.paraiso (override with $PARAISO_HOME).

from paraiso import Store

store = Store()
store.create("work")            # becomes active
p = store.current()             # load the active workspace
p.capture("ship the release")
store.save(p)

store.spaces()                  # ['work']
store.use("personal")           # switch active workspace

Bring your own AI

Classifier is the extension point. The default ManualClassifier suggests nothing (the user decides). Implement the protocol around any model to get suggestions, then let a person accept them:

from paraiso import Bucket, Suggestion

class MyLLM:
    def classify(self, capture, *, areas, objectives):
        # ...call your own model, referencing (never inventing) areas/objectives
        return Suggestion(bucket=Bucket.PROJECT, area_id=areas[0].id,
                          rationale="looked like a deliverable")

p = paraiso.Paraiso("Personal", classifier=MyLLM())
suggestion = p.suggest(capture)     # advisory only
if suggestion:
    item = p.accept(capture, suggestion)   # the user confirms

A tiny offline KeywordClassifier is included as a worked example.

Command line

paraiso framework                 # what PARAISO stands for
paraiso new personal              # create + switch to a workspace
paraiso spaces                    # list workspaces (active one starred)
paraiso rename work               # rename the active workspace (or `rename <old> <new>`)
paraiso capture "buy mulch"       # drop a thought in the Inbox
paraiso inbox                     # see what's waiting to be sorted
paraiso area add Health --tags fitness,sleep
paraiso file cap_ab12 --bucket project --title "Book dentist"

# Browse your PARAISO (color-coded in a real terminal)
paraiso projects                  # also: resources, seeds, archive
paraiso tree                      # a colorful overview of everything (clears the screen)
paraiso show                      # quick counts

# Areas & objectives (bare noun lists; add/show/edit)
paraiso area                      # list Areas (numbered, with counts)
paraiso area add Health --tags fitness,sleep
paraiso area show                 # explore one Area (its items by bucket + objectives)
paraiso area edit                 # rename / recolor (from the palette) / retag
paraiso objective                 # list Objectives (`objective add` creates)

# Sort, reclassify, delete
paraiso sort                      # guided: press one key per capture to file it
paraiso move <item_id> project    # reclassify a filed item (add --area to re-home it)
paraiso delete <item_id>          # delete an item (deleting an Area id keeps its items)

# Data portability
paraiso export backup.json        # export the active workspace as JSON
paraiso import export.json --name mine   # import a JSON export (paraiso or a PARA-style app)

Guided sorting. paraiso sort (or sort in the shell) walks the Inbox one capture at a time: press p / r / s / a to choose a bucket (a single keypress in a real terminal), pick an Area from a numbered list (or create one on the spot), and press Enter to accept the title. move and delete with no id drop you into a numbered picker, so you never have to type an id — guided move lets you change an item's bucket and/or its Area, choosing "keep" for whichever you want to leave alone. Deleting is the only destructive action and asks first (pass --yes to skip); deleting an Area keeps its items and just detaches them. tree prints the workspace name in an ===== banner above the buckets.

Color. Areas are assigned distinct colors from a calm 7-hue palette (cycling as you add them); recolor any Area with area edit, which offers only those predefined colors, never a free-form hex. Buckets are color-coded too. Output uses 24-bit ANSI and turns itself off automatically for pipes, NO_COLOR, or a dumb terminal — so scripts stay clean.

Explore & port. area show opens a single Area's page (its items grouped by bucket, plus objectives); tree and the Area page clear the screen for a clean view. import reads either paraiso's own export or a PARA-style app export (mapping module → bucket, e.g. seeds → seed) — Area and Objective links survive when the export includes their ids.

Interactive shell

Run paraiso with no arguments (or paraiso shell) to drop into a REPL with Tab-completion, command history, and arrow-key editing — all from the standard library, no dependencies:

It opens on a calm welcome screen (the PARAISO wordmark in its colors, your active workspace, and a hint), then you're at the prompt:

paraiso › new personal
paraiso (personal) › capture buy mulch
paraiso (personal) › file <Tab>                    # completes capture ids
paraiso (personal) › file cap_03dc --bucket <Tab>  # → project resource seed archive
paraiso (personal) › sort                          # guided: one keypress per capture

The prompt is colored per workspace (each name gets a stable palette color), and help lists commands grouped by category (Workspaces, Capture, Sort, Browse, Organize, Info); help <command> explains one. It runs the exact same commands as the one-shot CLI — anything you can script, you can also do interactively.

Status

0.7.0 — early but usable. The public API (models, Paraiso, Store, Classifier) is settling; expect additions before 1.0.

Related

paraiso is framework-first and app-agnostic. If you'd like PARAISO as a calm, full app — with AI that suggests where each capture belongs — see Chiefly, a calm AI "Chief of Mind" built on the same framework, by the same team (EJ Labs).