Crate issundb 

IssunDB is an embedded graph database in Rust.

Visit the project repository for documentation and examples

This is the main crate of IssunDB, and exposes the Graph type, Cypher query execution, vector search, full-text search, and hybrid retrieval APIs. Application code, bindings, and tools depend on this crate only; the internal crates (issundb-core, issundb-vector, issundb-text, issundb-retrieval, and issundb-cypher) are not part of the stable API.

Entry Points

• Graph is the central handle. Open it with Graph::open, then use its methods for node and edge CRUD, adjacency, and graph algorithms.
• GraphQueryExt adds Cypher execution (query, query_with_params, and explain) to Graph.
• VectorGraphExt adds vector indexing and search to Graph.
• TextGraphExt and TextIndexExt add full-text indexing and search.
• retrieve, retrieve_with, and retrieve_hybrid run hybrid retrieval over vector hits, text hits, and graph expansion.

Working with Query Results

Query parameters and Record values use serde_json::Value. The serde_json crate is re-exported as issundb::serde_json so callers do not need to track a separate, version-compatible dependency.

Re-exports

pub use serde_json;

Structs

Bm25Scorer

BM25 (Okapi BM25) relevance scorer with default parameters k1 = 1.2, b = 0.75.

DirectedNeighborEntry

A neighbor entry with a direction flag, returned by crate::Graph::all_neighbors.

EdgeRecord

Stored in the edges LMDB sub-database as msgpack bytes.

Graph

The graph database handle. It is cheap to clone, since all state is behind Arc.

GroupedDegreeSpec

Describes the pattern Graph::grouped_edge_counts counts, typed edges grouped by one endpoint. With group_is_dst, edges are grouped by their destination and the source is the counted endpoint (in-degree per destination); otherwise edges are grouped by their source and the destination is counted (out-degree per source). group_label and counted_label optionally constrain each endpoint (None is unconstrained). counted_nonnull_prop counts an edge only when the counted endpoint’s property is non-null (the semantics of count(v.prop) over the expansion); None counts every qualifying edge (the semantics of count(*) or count(v), where a bound node variable is never null).

Hit

A single result from vector search.

HybridRetrieveOptions

Options for retrieve_hybrid.

IdGroupCodes

Dense group codes of one property, indexed by node id rather than by a request’s position. codes[node_id] is the node’s group code under exact value identity, ID_GROUP_ABSENT where no such node exists, and reps holds one representative value per code. Built once per write generation by crate::Graph::node_prop_group_codes_by_id and shared, so a grouped aggregation over a bulk row set reads one array cell per row instead of interning one value per row per query.

NeighborCountSpec

Describes the pattern Graph::typed_neighbor_counts counts, the typed neighbors of each source across one hop. incoming follows incoming edges instead of outgoing ones. A neighbor qualifies when it carries every label in neighbor_labels (an empty slice is unconstrained) and, when neighbor_allow is present, is a member of that set; it adds to the counted total only when neighbor_nonnull_prop is absent or non-null on it (the semantics of count(v.prop) over the expansion, against count(*)).

NeighborEntry

The result of a single adjacency lookup entry returned by crate::Graph::out_neighbors and crate::Graph::in_neighbors.

NodeRecord

Stored in the nodes LMDB sub-database as msgpack bytes.

PathCountSpec

Describes the pattern Graph::count_linear_paths counts, an open directed path of one or two hops, (v0)-[t1]->(v1) or (v0)-[t1]->(v1)-[t2]->(v2), with an optional relationship type per hop and an optional label per node variable. None means unconstrained.

Procedure

A table-backed procedure. rows each contain inputs.len() + outputs.len() cells: the first segment matches the declared inputs, the second the outputs.

ProcedureRegistry

A runtime registry of procedures available to CALL.

QueryResult

The tabular result of a Cypher query execution.

ReadTxn

A read-only transaction on the graph.

Record

An individual row in the query result table.

RetrieveOptions

Options for retrieve_with.

Subgraph

A subgraph extracted by a retrieval call.

TextHit

A single ranked full-text search result.

TextSearchOptions

Options for full-text search.

TfIdfScorer

TF-IDF relevance scorer. Lighter than BM25; no length normalization.

TriangleCountSpec

Describes the pattern Graph::count_triangle_cycles counts, the directed cycle (a)-[t1]->(b)-[t2]->(c)-[t3]->(a) with an optional relationship type per hop and an optional label per node variable. None means unconstrained.

VectorIndexOptions

Construction options for VectorIndex.

VectorSearchOptions

Options for vector_search_with.

WeightedPath

A path with an associated total weight, returned by weighted path algorithms.

WriteTxn

A read-write transaction on the graph.

Enums

BooleanMode

Boolean candidate-set filtering for multi-term queries.

CypherError

Structured query engine errors representing all parsing, optimization, planning, and runtime execution faults.

CypherType

The declared Cypher type of a procedure argument or output field.

DegreeDirection

The direction of edges to count for degree centrality.

Error

FusionStrategy

Strategy for fusing vector and text relevance scores.

Language

Supported languages for Full-text Search indexing and stemming.

LinkPredictionMetric

Which score Graph::link_prediction_score computes for a pair of nodes.

PropCmp

A comparison operator for [PropColumns::cmp_mask], the typed in-column predicate evaluation behind Graph::nodes_prop_cmp_mask.

PropValue

A typed property value used in index lookups and range queries.

RetrievalError

Structured errors representing all hybrid retrieval faults.

TextError

VectorError

Structured errors representing all vector search and indexing faults.

VectorMetric

Distance metric for the vector index.

VectorQuantization

Quantization format for in-memory vector storage.

Constants

ID_GROUP_ABSENT

The slot value in IdGroupCodes::codes for a node id that does not exist (an allocation hole, or an id past the array).

Traits

GraphQueryExt

Extension trait to execute Cypher queries on the Graph handle.

Scorer

Relevance scoring strategy for full-text search.

TextGraphExt

Full-text search operations for Graph.

TextIndexExt

Text index lifecycle operations for Graph.

VectorGraphExt

Vector search operations for Graph.

Functions

retrieve

Wraps a vector search to k seeds and a hops-hop undirected BFS expansion into one subgraph materialization.

retrieve_hybrid

Merges vector search seeds with full-text search seeds, fuses their scores using opts.fusion, then expands via undirected BFS.

retrieve_with

Full retrieve with configurable options.

Type Aliases

EdgeId

LabelId

NodeId

TypeId