Which output preset should I use?
Every conversion detects one structure; the preset control decides how it's written out. The 30-second rule: pick by destination, not by file.
1 · The decision in one table
| You're about to… | Preset | What changes |
|---|---|---|
| Paste into a chat window | Chat | Outputs truncated hard, token estimate on line one — fits first, asks questions later |
| Upload to a Project / Space / knowledge base | Standard | Plain GFM — the workspace does its own chunking; don't decorate |
| Index into your own retrieval pipeline | RAG | <!-- chunk: slug --> boundaries + stable anchors for citations |
| Drop into your Obsidian vault | Obsidian | YAML front matter, callouts, ![[wikilink]] figure embeds |
| Keep a copy of record | Archive | Full front matter (source, date), faithful body, nothing extra truncated |
2 · Three details people miss
- Switching is free and instant. Presets re-format the already-parsed structure — click through them on one conversion and diff the Markdown pane. Nothing is re-parsed, nothing re-uploaded (nothing was uploaded).
- Chat ≠ smaller everywhere. Chat truncates outputs (logs, dataframe dumps); your prose and tables are untouched. If the document is still too big, that's a budgeting problem, not a preset problem.
- Standard is the deliberate default, not the boring one: for workspaces that chunk internally, added anchors and comments are noise. Decoration belongs where you control the pipeline.
3 · Try each in ten seconds
These links load a sample with the preset pre-selected — watch the same document change discipline:
- Chat — token estimate appears at the top, outputs cut at 12 lines
- RAG — chunk comments and anchors appear at every section
- Obsidian — front matter and callouts
- Archive — full provenance block
One structure, five disciplines — flip through them live.