Twelve copy-paste JSON examples for the managed graph database. Direction-tagged relation keys mean forward and reverse adjacency are both single prefix scans — same WAL, same recovery, same backup as SQL and vector. See /docs/graph for the full reference.
Walk forward edges from a source node along a single relation. Direction-tagged keys make this a single prefix scan.
curl "$ENGINE/v1/tenants/$T/graph/social.users/neighbors?rel=follows&pk=alice" \
-H "Authorization: Bearer $OC_TOKEN"{
"src": "alice",
"rel": "follows",
"neighbors": ["bob", "carol", "dan"],
"count": 3,
"elapsed_ms": 2
}Cap the fan-out. Useful for hot influencer-style nodes where the full neighbor set is large.
curl "$ENGINE/v1/tenants/$T/graph/social.users/neighbors?rel=follows&pk=alice&limit=10" \
-H "Authorization: Bearer $OC_TOKEN"{
"src": "alice",
"rel": "follows",
"neighbors": ["bob", "carol", "dan", "ed", "fay", "gita", "hari", "ira", "jay", "kim"],
"count": 10,
"elapsed_ms": 3
}Walk reverse edges. Direction-tagged keys (REL_DIR_FWD / REV) make reverse a single prefix scan, not a full table walk.
curl "$ENGINE/v1/tenants/$T/graph/social.users/reverse?rel=follows&pk=bob" \
-H "Authorization: Bearer $OC_TOKEN"{
"dst": "bob",
"rel": "follows",
"neighbors": ["alice", "ed", "fay"],
"count": 3,
"elapsed_ms": 2
}BFS at depth 1 returns the same set as neighbors but tagged with depth. Useful when the consumer expects level annotation.
curl "$ENGINE/v1/tenants/$T/graph/social.users/bfs?rel=follows&pk=alice&max_depth=1" \
-H "Authorization: Bearer $OC_TOKEN"{
"src": "alice",
"rel": "follows",
"max_depth": 1,
"frontier": [
{ "node": "bob", "depth": 1 },
{ "node": "carol", "depth": 1 },
{ "node": "dan", "depth": 1 }
],
"elapsed_ms": 3
}Bounded BFS. Visits each node at most once. Honour max_depth to keep wide graphs from blowing up.
curl "$ENGINE/v1/tenants/$T/graph/social.users/bfs?rel=follows&pk=alice&max_depth=3" \
-H "Authorization: Bearer $OC_TOKEN"{
"src": "alice",
"rel": "follows",
"max_depth": 3,
"frontier": [
{ "node": "bob", "depth": 1 },
{ "node": "carol", "depth": 1 },
{ "node": "dan", "depth": 1 },
{ "node": "ed", "depth": 2 },
{ "node": "fay", "depth": 2 },
{ "node": "gita", "depth": 3 }
],
"elapsed_ms": 7
}Returns whether a directed path exists within the depth bound, plus the actual path if so.
curl "$ENGINE/v1/tenants/$T/graph/social.users/path?rel=follows&src=alice&dst=ed&max_depth=4" \
-H "Authorization: Bearer $OC_TOKEN"{
"src": "alice",
"dst": "ed",
"rel": "follows",
"found": true,
"depth": 2,
"path": ["alice", "bob", "ed"],
"elapsed_ms": 3
}Bounded reachability. found=false means no path within max_depth — not necessarily that no path exists at all.
curl "$ENGINE/v1/tenants/$T/graph/social.users/path?rel=follows&src=alice&dst=zara&max_depth=2" \
-H "Authorization: Bearer $OC_TOKEN"{
"src": "alice",
"dst": "zara",
"rel": "follows",
"found": false,
"elapsed_ms": 4
}Caller supplies edge weights as a JSON object. The engine runs Dijkstra over (src, dst, rel) and returns total cost plus path.
curl "$ENGINE/v1/tenants/$T/graph/transit.stations/dijkstra?rel=line&src=central&dst=harbour&weights_json=%7B%22central-park%22%3A2%2C%22park-pier%22%3A4%2C%22pier-harbour%22%3A1%7D" \
-H "Authorization: Bearer $OC_TOKEN"{
"src": "central",
"dst": "harbour",
"rel": "line",
"found": true,
"cost": 7,
"path": ["central", "park", "pier", "harbour"],
"elapsed_ms": 5
}When edges are real rows with a numeric weight column, point Dijkstra at that field and skip the inline weights_json.
curl "$ENGINE/v1/tenants/$T/graph/transit.edges/dijkstra?rel=connects&src=central&dst=harbour&weight_field=minutes" \
-H "Authorization: Bearer $OC_TOKEN"{
"src": "central",
"dst": "harbour",
"rel": "connects",
"weight_field": "minutes",
"found": true,
"cost": 14,
"path": ["central", "park", "pier", "harbour"],
"elapsed_ms": 6
}A relation can be defined from a table to itself, for example a follower graph. Source and destination both live in social.users.
curl "$ENGINE/v1/tenants/$T/graph/social.users/neighbors?rel=blocks&pk=alice" \
-H "Authorization: Bearer $OC_TOKEN"{
"src": "alice",
"rel": "blocks",
"neighbors": ["spammer-1", "spammer-2"],
"count": 2,
"elapsed_ms": 2
}BFS at depth 2 surfaces second-degree neighbours. Combine with a follow-up SQL fetch on the resulting ids to materialise rows.
curl "$ENGINE/v1/tenants/$T/graph/social.users/bfs?rel=follows&pk=alice&max_depth=2" \
-H "Authorization: Bearer $OC_TOKEN"{
"src": "alice",
"rel": "follows",
"max_depth": 2,
"frontier": [
{ "node": "bob", "depth": 1 },
{ "node": "carol", "depth": 1 },
{ "node": "ed", "depth": 2 },
{ "node": "fay", "depth": 2 }
],
"elapsed_ms": 5
}Relations cross tables when both endpoints are declared in the manifest. Reverse adjacency answers "which customer placed this order?".
curl "$ENGINE/v1/tenants/$T/graph/shop.customers/neighbors?rel=placed&pk=c_42" \
-H "Authorization: Bearer $OC_TOKEN"{
"src": "c_42",
"rel": "placed",
"neighbors": ["o_201", "o_205", "o_318"],
"count": 3,
"elapsed_ms": 3
}