The Stargate Graph — Investigation Deck
Every line of dialogue from Stargate Atlantis, in our
own Neo4j — plus an extraction layer that reads the episodes and builds
the web of people, places, and facts. No typing queries, no
copy-paste: click a bubble and it becomes the centre of the universe —
its solar system explodes out around it. Hover a bubble to light its
relationships; hover a line itself to read the fact written on it.
Text first, speakers later — nothing here claims who said a line
unless a receipt says so.
Conversation — coming
This rail is reserved. Phase B puts the crew here: talk to
Rodney, Zelenka, whomever — they answer from this same graph, and
you watch the universe they are reading light up beside the
conversation.
Not switched on yet: this page is public, and
the crew only answers behind your WordPress login.
Episode lines
The full script opens here, in timestamp order.
Watch — the talks behind this build
Build reliable AI with Aura Agents (Neo4j)
NODES AI 2026 — Temporal Substrate Architecture: persistent identity for agents
Playlist — How to Build Better GenAI Apps with Knowledge Graphs
How to investigate
Travel: click any bubble and it becomes the centre
of the universe — the web explodes and redraws around it, neighbours
taking their orbits by how strongly they are bound to it. Click
another to travel there. The dropdowns and the search box fly you the
same way. Hover: rest the mouse on a bubble and only
its world stays lit, its lines labelled with their facts; rest it on
a line itself and the fact on that line lights up gold.
Categories: the chips hide everything outside the family —
the status line counts what is showing — and clicking the chip again
brings the whole web back. In a bubble’s panel, Mentions
ranks every episode whose script names it; if a name has no
bubble yet, the search goes into the scripts themselves and finds the
lines anyway. Bacon Path finds the shortest chain of
facts between any two bubbles. Show everything always brings
the whole web home.
Rather drive the raw graph yourself? Open the Neo4j Browser → — paste a query, press Ctrl+Enter, double-click bubbles to walk the web.
Starter queries (for the Browser)
Zelenka’s star — every episode he appears in
MATCH p = (c:SG_FictionalCharacter {name: 'Radek Zelenka'})-[:SG_APPEARS_IN]->(e:SG_Episode) RETURN p
Explode one episode into its lines (first 100)
MATCH (e:SG_Episode {episode_key: 'SGA S01E10'})-[:SG_HAS_BLOCK]->(b:SG_TextBlock) RETURN e, b ORDER BY b.block_index LIMIT 100
Search any line of dialogue
MATCH (b:SG_TextBlock) WHERE b.text CONTAINS 'Rodney' RETURN b.episode_key, b.t_start, b.text LIMIT 25
The extracted web — Zelenka’s facts (Graphiti layer)
MATCH p = (n:Entity)-[:RELATES_TO]-(m:Entity) WHERE n.name CONTAINS 'Zelenka' RETURN p
Who is named alongside whom (same line, whole corpus)
MATCH (b:SG_TextBlock) WHERE b.text CONTAINS 'McKay' RETURN b.episode_key AS episode, count(*) AS mentions ORDER BY mentions DESC
Built by Lisa Muse for Paul Prime · Project Free Life · the deck reads a read-only snapshot baked from the living graph; refresh re-bakes it.