anytimeline

Interactive timeline and event charts for Jupyter and marimo notebooks, drawn on the GPU (WebGPU, WebGL2, or Canvas 2D on machines without a usable GPU).

pip install anytimeline

Documentation: https://mluttikh.github.io/anytimeline/ · Changelog: CHANGELOG.md · Licence: MIT.

Status: version 0.1. It draws event charts: one mark per event at the moment it happened, on lanes, so the order of what happened is easy to see. Time can be datetimes (any time zone, any century), durations from nanoseconds up, or plain numbers; a Time | Order switch spaces events by rank instead of time. Up to about a million events. It also draws Gantt charts and schedules: bars from a start to an end, with dependencies, progress and milestones, and overlapping bars stacked per lane; and traces: spans under spans.

from anytimeline import Timeline
from anytimeline.synthetic import service_log

log = service_log()  # a dict of columns; any Polars, pandas or PyArrow table works too
tl = Timeline()
tl.events(
    log,
    time="time",
    lane="service",
    color="level",
    symbol="event",
    route="request",
    order="seq",
    tooltip=["request", "user", "bytes"],
)
tl  # click it, then scroll to zoom time; drag to pan; the toolbar has Zoom and Select

A Gantt chart from a table with a row per task:

from anytimeline import gantt

gantt(plan, task="task", start="start", end="end", depends="after", progress="done",
      color="phase", today="2026-10-12")

Bars on other lanes (machines, people, services) come from Timeline.intervals(...), and mix with events on the same lanes.

Traces (OpenTelemetry, or any table of spans with a parent) plot as a waterfall that folds, with each span's duration and the critical path:

from anytimeline import Timeline
from anytimeline.otel import read_otlp

tr = read_otlp("traces.jsonl")  # or anytimeline.synthetic.traces()
Timeline().spans(tr.spans, color="service").events(tr.events, time="time", lane="span_id")

The toolbar has three tools: Pan (drag moves through time and lanes), Zoom (drag a time range to show it) and Select (drag a box, or click an event; the selection is rows of your table in tl.selection). Time and Order switch the scale; Aa sets the text size (also Timeline(font_size=14)). Press ? in the plot for every mouse gesture and key; f or the button shows it full screen, where lanes fill the screen and overlapping events spread out.

States (Timeline.states: up, degraded, down per lane) and values (Timeline.values: a number over time, as a line, area or steps) go on the same lanes, behind the events.

group="host" on a layer puts each host's lanes under a header of their own, in one plot (Timeline(align="start") compares groups from their start); separate plots link with a.link(b) and zoom and pan together.

plot.save_html("plot.html") writes the plot to one file that opens in any browser, still interactive; the toolbar saves it as a PNG, and so does plot.save_png("plot.png").

Documentation online: https://mluttikh.github.io/anytimeline/; or build it with uv run python scripts/build_docs.py and open site/index.html. Example notebook: examples/event_charts.ipynb. How the code is built: docs/architecture.md.

Development

Needs Python (via uv) and Node.js 20+ for the JavaScript bundle.

(cd js && npm ci && npm run build)    # build src/anytimeline/static/widget.js
uv sync                               # create .venv with dev tools
uv run pytest                         # Python tests
(cd js && npm run typecheck && npm test)
uv run ruff check . && uv run ruff format --check . && uv run ty check
uv run python scripts/make_scenes.py  # scenes for the visual tests
(cd js && npx playwright install chromium && npm run test:visual)  # every scene × backend, marimo
uv run nbstripout --install           # once per clone: commit notebooks without outputs
uv run python scripts/build_docs.py   # user documentation into site/ (examples run; warnings fail)
uv run python scripts/build_docs.py serve   # preview it at http://localhost:8000

The user documentation is in docs/user/, built with Zensical and organised as tutorials, how-to guides, reference and explanation. Its examples are ```python exec="on" blocks that run while building, and its plots are live: each example prints plot._embed_html(), which standalone.js draws in the page. The API reference comes from the docstrings (Google style). The browser tests check the built site too (docs.spec.ts, skipped until it is built).

Visual tests draw the scenes in js/test/visual/scenes/ with each backend and compare them with stored screenshots (macOS); npm run test:visual:update stores new ones after an intended change. PERF=1 npx playwright test perf measures a million events (after uv run python scripts/make_scenes.py --million).