Sources & Methodology
Primary source: the game files
The vast majority of this wiki is not copied from community pages. It is parsed directly from the No Man's Sky game files via nms.source.gs, a community tool that re-extracts the game's internal database on every update and renders it as a browsable catalogue.
What this means in practice:
- Item values, rarities, recipe inputs, technology costs, expedition rewards — all of these are the exact numbers the game ships with, not a community editor's last edit.
- Coverage is complete for finite content: if an item, recipe, technology, fish, frigate trait, settlement perk, upgrade module, expedition, quicksilver item or Twitch drop exists in the game files, it is in this wiki.
- Procedural content is excluded by design: planet names, system names, base names, creature names and outposts are generated per-player and per-system. Listing them would be infinite, so only the finite, game-crafted content is seeded.
Secondary source: Fandom
The No Man's Sky Fandom wiki was used as a secondary reference for:
- cross-checking numbers where the game-file parse was ambiguous,
- providing lore and context (expedition themes, faction backgrounds, expedition story beats) that the raw game-file rows don't carry,
- and resolving naming conflicts where a community page and the game file disagreed.
When the two disagreed, the game-file value won (it is the authoritative number), and the conflict was noted in the build log.
Rewording
All prose in this wiki — descriptions, summaries, guide pages — is reworded in our own words. Numbers, names, tiers, prices, percentages and dates are preserved exactly. No block of text is copied verbatim from any source.
Data pipeline
- Fetch — each
nms.source.gssection (items × 35 categories, technology, refining, cooking, fishing, frigates, settlements, upgrades, expeditions, quicksilver, twitch-drops) is fetched as server-rendered HTML. - Parse — the HTML rows carry the data as attributes (
data-nname,data-ssubtitle,data-ffeature/rarity,data-value, etc.). A Python parser extracts every row into structured JSON. - Build — the JSON is grouped into 10 entity types and written to
seed.jsonwith unique slugs. - Seed — the engine's
saveEntityupserts every row into Postgres. Zero failures required. - Verify — DB counts are checked against the seed exactly; 10+ random facts are spot-checked against the source HTML; the production build is smoke-tested.