All Classes and Interfaces
Class
Description
Intelligence-backed adapter from one canonical source declaration to the
existing chat-level
KnowledgeBase contract.Default intelligence implementation of the unified retrieval-engine SPI.
Action enforcement boundary backed by the unified retrieval engine.
One heading-bounded section of a long document.
Binding-time FILE chunks available to the in-memory graph provider.
One ingested FILE chunk.
Complete source-configuration and lifecycle identity for one binding.
Exact-owner lifecycle handle for one registered corpus.
Builds graph nodes and edges from ingested FILE chunks.
Canonical filesystem ingestion module shared by declarative and Server RAG.
One normalized retrieval chunk emitted by the canonical file pipeline.
Synchronous source-ingestion progress snapshot.
Immutable ingestion output and filesystem snapshot completeness.
Lifecycle event emitted while a source is discovered and normalized.
One admission-cap policy for declarative and Server file ingestion.
What one FILE-source load observed, per file, so the next load can reuse
the files that did not change.
Reloads a FILE knowledge source against the manifest of its previous load
and reports which files and chunks changed.
Outcome of one reconcile: the load result, the manifest for the next
reconcile, and the diff against the prior manifest.
Signals that graph retrieval was requested but no unambiguous graph
capability is available for the Role.
Deterministic, bounded graph traversal retrieval strategy.
Vendor-neutral graph data access contract used by graph retrieval.
A directed graph edge.
An outgoing edge paired with its destination node.
A graph node that can become a retrieved document.
A scored query seed.
Service-provider contract for graph database and graph corpus adapters.
Graph binding request.
Expands ranked child hits through a deterministic parent hierarchy.
Signals malformed or inconsistent hierarchical retrieval data.
Immutable hierarchy document index.
Canonical metadata keys for hierarchical retrieval.
Loads bounded HTTP/HTTPS documents through the ordinary knowledge-source SPI.
JDK
HttpClient transport for Qdrant REST.Outcome of attaching the optional knowledge-graph stream to
HybridRAGStrategy.RAG strategy that fuses BM25, vector similarity, and — when a
GraphRAGStrategy is supplied — knowledge-graph hits via
Reciprocal Rank Fusion (RRF).Zero-config in-process
GraphStore for development and FILE ingest.Bundled FILE-corpus graph adapter.
JDK JDBC transport for a pgvector table.
RAG strategy that uses BM25-style keyword matching.
Setup-time checks that a role's knowledge loads, before the first action
runs.
Raised when file content contradicts a declared knowledge-source format.
Built-in graph adapter for live
KnowledgeTools triples.Declarative FILE-source adapter over the canonical
FileIngestionService.Expands one query, retrieves every unique variation through the same
source-scoped base strategy, and fuses the results deterministically.
GraphStore over a Neo4j property graph.Minimal Cypher execution seam.
Optional Neo4j
GraphStoreProvider.Optional binary-document extractor for PDF, Office Open XML and EPUB.
JDBC failure from
JdbcPgvectorTransport.JDBC seam for a pgvector table.
Connection settings for one pgvector table.
pgvector-backed
VectorIndex.Optional pgvector adapter with the same CRUD surface as
VectorMemoryStore.Builder for an explicit pgvector connection.
HTTP failure from
HttpQdrantTransport.HTTP seam for the Qdrant REST API.
Connection settings for a Qdrant collection.
Qdrant-backed
VectorIndex.Optional Qdrant adapter with the same CRUD surface as
VectorMemoryStore.Builder for an explicit Qdrant connection.
Service provider interface for model-backed retrieval query expansion.
Immutable input passed to a
QueryExpander provider.Drops the role corpora, graph corpora and checks built from class
configuration files when
ClassConfig.useBaseDirectory(java.nio.file.Path)
starts a new run, so the next dispatch reads its sources again.Configuration context for a RAG retrieval request.
Read-only snapshot of what a Role's
@Knowledge and
@Search declarations resolved to.Installed SPI providers for one unbundled seam.
One
@Search declaration on the Role.Ingest snapshot for one declared knowledge source.
Orchestrates the full RAG flow: preprocess query, retrieve, rerank, format context.
Builder for
RAGPipeline.Service provider interface for pluggable RAG retrieval strategies.
Descends a document tree with a model choosing the branch, without scoring.
Service provider interface for query-aware document reranking.
Immutable input passed to a
Reranker provider.A document retrieved by a
RAGStrategy.Per-Role RAG state, lazily initialized on first
RoleRagBinding.forRole(java.lang.Class<?>) call and
cached for the lifetime of the JVM.What a
RoleRagBinding.sync(java.lang.Class<?>) found, by source name, each list in
name order.Local vector index that re-embeds only chunks whose content hash changed.
Counts from one
SelectiveReembedIndex.syncSource(String, List) call.One search hit.
Thread-safe cursor over one source's ingested units in exact ingest order.
Heading-aware splitter for long-document analysis.
Built-in bounded SOURCE_CODE extractor.
Loads documents for a single
KnowledgeSourceConfig declaration into
the per-Role RAG memory store at binding-init time.One document loaded by a
SourceLoader — the unit ingested
into the vector store.Built-in bounded extraction for UTF-8 textual knowledge-source formats.
Signals invalid temporal metadata or an inconsistent temporal index.
Immutable temporal document index.
Canonical input and provenance metadata keys for temporal retrieval.
Re-ranks content-relevant child hits with bounded time decay.
Service provider interface for model-driven traversal of a document tree.
What the navigator decided to do at one node.
One node as the navigator sees it.
Immutable input passed to a
TreeNavigator provider.Signals that vectorless reasoning retrieval was requested but no
unambiguous
TreeNavigator is available for the declared model.Picks at most one optional vector backend for Role bindings.
RAG strategy that uses embedding-based semantic similarity search.