mktask

A work task prioritizer built on mkio (config-driven microservice backend) and mkui (config-driven Web Components workspace with dockable panes).

Tasks live in a local SQLite database and show up in a live-updating blotter you can sort, filter, and arrange however you like. Each task carries an importance and an urgency (1–5); the blotter derives a score from them so the most pressing work floats to the top.

A task can be split into child tasks, and a child can be split again, to any depth. The blotter nests children under their parent, each level sorted by score, with carets to fold a subtree away. Completing a task completes everything split from it; reopening a child reopens its ancestors; deleting a task deletes its whole subtree — permanently, and it is the one thing here that cannot be undone.

Move re-files a task: pick a new parent, or — Top level — to make it a top-level task again. Everything split from it comes along. The picker offers every open task except the ones under the task being moved, and a move that would land a linked task in its own tree is refused, naming the links to remove first.

A task can be assigned to someone. Assigned To is a dropdown: pick a name, pick — New name — and type one — a name typed there joins the list and is offered from then on — or pick the blank first entry, which means nobody. The dropdown opens on whoever the task is assigned to, so leaving it alone changes nothing, and splitting a task opens on the parent's name, so a child usually needs no picking at all.

The list itself is yours to keep, under Configure › Assignees: add a name before anyone is assigned to it, fix a spelling, or take a name off it when someone leaves. Taking a name off the list changes the list and nothing else — every task already assigned to that name keeps it, and the name still shows in the dropdown of a task that carries it, so editing that task cannot lose it by accident. Renaming works the same way: it changes what the picker offers next time, not what any task says.

Every task gets a Task ID like TKMA00000042: TK, two letters from the username the server runs as (--user), and a sequence number that starts at 00000001, never repeats, and grows past eight digits rather than wrapping (TKMA100000000 follows TKMA99999999).

A task carries references: the things to refer back to while working on it. A reference is a URL (including mailto: and mail-client links), a pasted text snippet (an email, a chat exchange), a file (a screenshot, a PDF), or a link to another task with a relation. The References pane's Add URL, Add Text, and Add Link buttons add one each, and the Detail pane has the same three. Each dialog opens on a Task picker already set to the task selected in the blotter — change it to file the reference on another task, or pick one when nothing is selected — with the label suggested from the URL or the first line of the text; once added, the blotter selects the task it went to, so the new reference is on show whichever task that was. Files are added from the Detail pane instead: select a task and drop, paste, or pick a file there; an image pastes straight from the clipboard, a pasted URL becomes a URL reference, pasted text a snippet. The Detail pane opens with the task itself — its status, importance, urgency, score, due date, who it is assigned to, and notes — and lists every reference of the selected task the moment it is selected, and every reference of the tasks split from it — a section per relation ("Blocks", "Blocked by"), then the snippets, images, files, and URLs, each section counted, and a reference a child owns tagged with its Task ID. Images show as a grid of thumbnails; clicking one opens it full size under the grid, and clicking a snippet or a URL opens it where it stands. The subtree comes from the tasks themselves, not from what the blotter is showing, so a completed child's references stay with its ancestor.

The Detail pane edits what it shows. It has its own toolbar — Add URL, Add Text, and Add Link are the References pane's add buttons — and one cursor: click the task block or click a reference. Edit opens the dialog that matches — the same Edit dialog the blotter opens for a task, the same one the References pane opens for a reference — and Delete removes a reference (a task is deleted from the blotter, where the row being deleted is the row that was picked). An edit shows up here the moment the server announces it.

The References pane follows the selected tasks (mkui table linking; clear the selection to see every reference), each reference beside its task's Task ID and title (the title is joined in live: renaming the task updates the column), with URLs and files as links; selecting one there opens and marks the same line in the Detail pane, and a reference the Detail pane does not list leaves it alone. A task link shows on both tasks, worded from each side, and Go to selects the linked task in the blotter, so the Detail and References panes follow it (a linked task the current filter hides is revealed first). Only tasks from different trees can be linked; tasks that share a root are already related by splitting.

Relations are yours to define under Configure › Relations: each is a pair of wordings, one from this task to that one ("blocks") and one back ("blocked by"), or a single wording that reads the same both ways ("relates to"). A new database starts with those two. Renaming a relation rewrites every link that uses it; a relation in use cannot be deleted. Uploaded files live beside the database in <db>.files/, named by content hash; deleting the last reference to a file, or its task, removes the file.

Each task also shows its Last Event — the most recent thing to have happened to it, in words: Created, Split into Wire the pane, Attached a file: shot.png, Linked: blocks TKMA00000002, Moved under TKMA00000007.

Every change is kept. Tasks and references carry a version number, shown as a column, and every version of every row is recorded. Adding, editing or removing a reference makes a version of the task that owns it — both sides of a link — so a task's history covers its references as well as its own fields, and each version says what it was about. History on the Tasks or References pane — or in the Detail pane, for whatever its cursor is on — opens that record: the versions as a table you can sort and search, a Diff of what changed between any two, and Blame for which version last set each field.

Undo and Redo step a record back and forth along those versions, from the same panes. They step the whole action, not one row: undoing a completion reopens the subtree it completed, undoing a link removes both halves, undoing a move puts back the ancestors it reopened. A confirmation says what is about to change. Undoing the creation of a task removes it, and redo brings it back. Editing a task after undoing it discards what was undone, the way typing after an undo does in an editor. There is no keyboard shortcut on purpose — this writes to data everyone on the server shares.

Undo and redo are refused rather than half-applied when they would strand something: a task that has since been split, a parent or a linked task that has since been deleted, or a row another change has moved past.

Activity (under the Tasks menu) is the other half: a plain list of what happened to the selected task — created, split from, split to, edited, moved, completed, reopened, references added, edited and deleted, and every undo and redo. It follows the blotter's selection like the References pane, and unlike the version record it is never rewritten.

Quick start

pip install mktask
mktask                    # http://127.0.0.1:8080/

Everything installs via pip; nothing is fetched at runtime.

CLI

mktask [config] [-p PORT] [--host HOST] [-d PATH] [-u USER] [--files DIR] [--version]

• config — path to a mktask.toml. Defaults to ./mktask.toml if present, otherwise the one bundled with the package.
• -p, --port — override the listening port (default 8080).
• --host — override the listening host (default 127.0.0.1).
• -d, --db — database file; .db is appended when there is no extension. :memory: runs without persistence.
• -u, --user — the username whose first two alphanumeric characters, uppercased, prefix new Task IDs (mark → TKMA…; a one-character name is padded with X). Defaults to the OS login name.
• --files — directory for uploaded reference files, served at /files. Defaults to <db>.files beside the database (mktask.db.files/), a temporary directory for :memory:.

The server prints the URL to open once it is listening. If the port is already taken it exits immediately with an error instead of starting.

Customizing

Copy the bundled config out and edit it:

python -c "import mktask, pathlib; print(pathlib.Path(mktask.__file__).parent / 'mktask.toml')"

mktask.toml declares the SQLite tables, the mkio services, and the static routes (keep relations.json, the seed for the relations table, beside a copied config); static/app.json next to it declares the UI (menus, panes, frames, dialogs). Both are plain config — see the mkio and mkui READMEs for the formats. The one piece of code is mktask/services.py, which the tasks service points at: it assigns Task IDs, runs the complete, reopen, and delete cascades, re-parents a task on a move (refusing a cycle or a move that would swallow a task link), and manages references and relations (task links are written as a mirrored pair, labels follow the linked task's title, a renamed relation rewrites its links, orphaned files are removed). static/refs.js is the one custom widget and the whole Detail body: the toolbar, the task block, the drop box, the reference list of the selected task and its descendants (its own live queries), and Go to. Its Edit and Delete open the dialogs the Tasks and References panes already declare, borrowed from app.json by pane and button name rather than copied.

Upgrading

Until 1.0, a release may change the database schema without migrating an older database. 0.2.0 did (Task IDs, splitting, and complete replacing done): delete a 0.1.0 mktask.db before starting a newer version. 0.3.0 only adds the task_refs table, which auto_migrate creates in an existing database, so a 0.2.0 database carries over as is. 0.3.1 changes no schema. 0.4.0 adds the relations table (created and seeded on first start) and stores a link's relation as its wording ("blocked by") rather than a key (blocked_by): links made before 0.4.0 must be removed and re-added. 0.5.0 changes no schema — the Detail pane's new reference list reads the tables 0.4.0 already had, and neither does 0.6.0: moving a task rewrites one existing column. 0.7.0 changes no schema either: the Detail pane's toolbar sends the ops that were already there. 0.8.0 adds a last_event column, the task_events table, and a recorded history of tasks and task_refs — all of which an existing database picks up on first start: auto_migrate adds the column and the tables, and every row already there is recorded as its own version 1 so it can be stepped back to. Nothing has to be deleted. 0.9.0 adds an assigned_to column and the assignees table behind the Assigned To dropdown, both of which auto_migrate adds to an existing database on first start; every task starts unassigned and the list starts empty. 0.10.0 changes no schema: the Add URL / Add Text / Add Link dialogs read a new lookup service over the tables that were already there. 0.11.0 changes no schema either: the References pane's Task Title is joined in by the task_refs query, which mkio 0.8.0 keeps live, so nothing new is stored. 0.12.0 changes no schema: it only takes the Show Open Only / Show All / Expand All / Collapse All entries off the Tasks menu. 0.13.0 changes no schema: the statusbar now shows the server's version beside the name, read from the same reply that checks the server is mktask. It is also the first release to require mkio and mkui 1.x (pip install --upgrade mktask brings them in); a 0.12.0 database carries over as is. 0.13.1 changes no schema: it is the Windows release, and raises the mkio floor to 1.0.1. 0.13.2 changes no schema: it raises the floor to mkio 1.1.0, whose server runs on asyncio's selector loop on Windows, so the connections a browser opens ahead of a page load and drops no longer print a ConnectionResetError traceback each. 0.13.3 raises it to mkio 1.1.1, whose Windows loop lets Ctrl+C stop the server at once instead of waiting for the next timer or packet.

Development

pip install -e '.[test]'
mktask -d :memory:
python -m pytest

The tests cover the CLI and config loading, Task ID formatting, a real server over HTTP and WebSocket (every task op, splitting and the cascades, moving a task to another parent or the top level, the Task ID sequence across restarts and past eight digits, references and task links with their validation, the same-tree rule, and cascades, user-defined relations (seeding, uniqueness, rename rewriting links in both directions, a swap flipping them, delete refused in use), file upload, dedupe, and cleanup, live delete announcements, the query filter, saved layouts, the port, host, user, and files flags), the port probe on a fake Windows, what only breaks on Windows (every text read names its encoding, LF pinned), and the static integrity of app.json against mktask.toml and refs.js (every column, dialog field, service, and state path a pane names, the Detail pane's task block and the dialogs its toolbar borrows, and the reference sections against the kinds the server accepts).

Runs on Linux, macOS and Windows with the standard CPython 3.11+ interpreter; Windows needs mkio 1.1.1 or later, which pip install brings in. On Windows, Ctrl+C stops the server the same way as elsewhere. The Task ID prefix comes from the account name's first two ASCII letters or digits, so a name starting with a space or a non-ASCII letter still yields a valid one.

License

Apache License 2.0. See LICENSE.