Traverse connections graph
traverse_graphMulti-hop graph walk from a start node, with filters. This is the EXACT-enumeration tool: 'every recipe using brown sugar' (start='ingredient:brown-sugar', node_types=['entry']), 'all of Grandma's recipes' (start='author:...', node_types=['entry']), 'the ingredient fingerprint of mom's collection' (start='user:...', depth 2, node_types=['ingredient']). Results are complete up to limit — if truncated, the response says so. Node ids are ':' — e.g. 'entry:' (prefix a search_recipes hit id with 'entry:'), 'ingredient:brown-sugar', 'author:mother-in-law', 'appliance:oven', 'user:'. User-supplied metadata is traversable too: 'tag:' (a shared free-form label) and 'property::' (a shared key=value, e.g. 'property:region:alaska') — every entry carrying that tag/value links to the same node. Recipes with named sub-preparations carry 'component::' nodes (HAS_COMPONENT from the entry; the component links to its own ingredients/steps) — e.g. "what else uses this sauce?" is one hop.
Input Schema
| Name | Required | Description | Default |
|---|---|---|---|
| depth | No | Max hops from start (default 1). | |
| limit | No | Max nodes visited (default 100). | |
| start | Yes | Start node id (bare recipe ids auto-resolve). | |
| edge_types | No | Only follow these relationship types. | |
| node_types | No | Only RETURN nodes of these types (traversal still passes through others). |