Overview

MySTRA lets you write publications on top of ASTRA analyses — in plain MyST Markdown. You write a normal MyST document and reference ASTRA components — decisions, outputs, findings, prior insights, data tables, live numbers — by reference. The components stay single-sourced in your astra.yaml; MySTRA reads it at build time and emits standard MyST AST.

It runs on the stock myst CLI and themes — no custom server, no copy-pasted numbers, no figures that drift out of sync with the analysis.

We adopt the {astra}`decisions.algorithm.multigrid` and measure
$\alpha =$ {astra:value col=alpha where="tracer=elg1" pm=true}`outputs.alpha_table`,
consistent with {astra:cite}`prior_insights.recon_sharpens_bao`.

:::{astra} outputs.bao_fit_plot
:::

→ the decision links to its full record, the value is interpolated live from the result table (uncertainty included), the prior insight renders as a citation, and the figure is pulled from the resolved artifact binding. Edit astra.yaml and rerun the analysis; the report updates itself.

In the Lightcone StackCopy Anchor Link

Projects created with lc init are already set up for MySTRA: their myst.yml registers the plugin and uses the ASTRA article theme, and index.md is the report, ready to write. The Lightcone CLI commits each output to results/<universe>/<output>.<format>, which is where MySTRA finds the figures, tables and values it places in the report. For an agent writing one up, see the last step of Your first analysis.

MySTRA is intentionally a publication adapter, not a project viewer. It does not emit a whole-project inventory, graph application, or host-specific UI, so reports remain normal, portable MyST documents. To explore a project interactively, use Lightcone Lab.

Why MySTRA?Copy Anchor Link

ASTRA already holds the truth of an analysis — every decision, the inputs and outputs of each step, the findings, and the materialised result products. A write-up usually re-types all of that into prose, where it immediately starts to rot: a number gets stale, a figure is from an old run, a stated assumption no longer matches the spec.

MySTRA removes the duplication. The report references the analysis instead of restating it, so there is one source of truth for the data and the prose can focus on the argument.

The plugin is a pure projector between data and document: it renders the elements you place, fills in the numbers, and makes no authoring or styling decisions of its own.

At a glanceCopy Anchor Link

One name — astra — drives every surface, the MyST way (just as {math} is both a role and a directive). You address any part of the analysis with a path that mirrors astra.yaml:

# BAO from the ELG sample

We adopt the {astra}`decisions.algorithm.multigrid` and measure
$\alpha =$ {astra:value col=alpha where="tracer=elg1" pm=true}`outputs.alpha_table`,
consistent with {astra:cite}`prior_insights.recon_sharpens_bao`.

:::{astra} outputs.bao_fit_plot
:caption: The post-reconstruction BAO fit.
:::

:::{astra} findings.bao_detected
:compact:
:::

What MyST handles for youCopy Anchor Link

MySTRA writes only the ASTRA→AST bridge and leans on the stock myst engine for everything else: building, serving, asset hashing and copying, live reload of Markdown, numbering, cross-references, and search. Citations are delegated to MyST too — DOI evidence renders as a doi.org link, and a linked reference list comes for free once a project bibliography is wired.

Where to go nextCopy Anchor Link