batch-render-typst

batch-render-typst

Render N PDFs in one pass: one Typst template + a directory of JSON payloads → one PDF per payload.

Inputs

1. Template path — .typ file. If the user just ran new-typst-template, default to that output.
2. Data directory — folder of *.json files, one per document.
3. Output directory — where PDFs land. Default <data-dir>/../out/.
4. Concurrency (optional) — how many typst compile processes to run in parallel. Default 4.

Pre-flight

• Verify typst is on PATH. If not, instruct install (cargo install typst-cli or distro package). Do not auto-install.
• Verify the template uses the #let data = json("data.json") pattern. If it hardcodes paths, fix or warn.
• Verify at least one JSON file exists in the data directory.

Approach

For each <data-dir>/*.json:

1. Stage the file as data.json next to the template (or use Typst's --input flag if the template supports sys.inputs).
2. Run typst compile <template.typ> <out-dir>/<basename>.pdf.
3. Capture stdout/stderr. On failure, record the bad payload and continue.

Run with bounded parallelism (e.g. xargs -P or GNU parallel). Don't shell-out one process per doc serially for big batches.

Reporting

After the run:

• Total processed, succeeded, failed.
• Per-failure: filename + first line of error.
• Output directory path.

Write a render-report.json next to the output directory with the same data so downstream automation can consume it.

Storage rules

Template, data, and output paths are user-owned — they live wherever the user keeps them. The plugin only persists pointers and run history under $CLAUDE_USER_DATA_ROOT/state/last-batch.json (resolve via ${CLAUDE_USER_DATA:-${XDG_DATA_HOME:-$HOME/.local/share}/claude-plugins}/programmatic-doc-generation/).