RetrievalManager class

Enhanced Retrieval manager with improved RAG capabilities

Constructors

RetrievalManager({required LlmInterface llmProvider, DocumentStore? documentStore, VectorStore? vectorStore, String? defaultNamespace})

Constructor alias for convenience

factory

RetrievalManager.withDocumentStore({required LlmInterface llmProvider, required DocumentStore documentStore})

Create a retrieval manager with a document store (legacy mode)

RetrievalManager.withVectorStore({required LlmInterface llmProvider, required VectorStore vectorStore, String? defaultNamespace})

Create a retrieval manager with an external vector store

Properties

hashCode → int

The hash code for this object.

no setter, inherited

llmProvider → LlmInterface

final

runtimeType → Type

A representation of the runtime type of the object.

no setter, inherited

usesVectorStore → bool

Check if the manager is using an external vector store

no setter

Methods

addDocument(Document document) → Future<String>

Add a document to the store with embeddings

addDocuments(List<Document> documents) → Future<List<String>>

Add multiple documents in batch with parallel processing

clearCache() → void

Clear the retrieval cache

close() → Future<void>

Close and clean up resources

contextAwareSearch(String currentQuery, List<String> previousQueries, {int topK = 5, double? minimumScore, String? namespace, Map<String, dynamic> filters = const {}}) → Future<List<Document>>

Context-aware search utilizing conversation history

deleteDocument(String id, {String? namespace}) → Future<bool>

Delete a document

deleteDocuments(List<String> ids, {String? namespace}) → Future<int>

Delete multiple documents

hybridSearch(String query, {int semanticResults = 5, int keywordResults = 5, int finalResults = 5, double? minimumScore, double keywordBoost = 0.3, String? namespace, Map<String, dynamic> filters = const {}}) → Future<List<Document>>

Hybrid search combining semantic and keyword search

multiChunkAnswer(String query, List<Document> chunks, {Map<String, dynamic> generationParams = const {}, bool useDynamicChunkSelection = true}) → Future<String>

Create focused answer from multiple chunks

multiCollectionSearch(String query, List<String> collectionIds, {int resultsPerCollection = 3, int finalResults = 5, double? minimumScore, bool rerankResults = true}) → Future<List<Document>>

Search across multiple collections

noSuchMethod(Invocation invocation) → dynamic

Invoked when a nonexistent method or property is accessed.

inherited

rerankResults(String query, List<Document> candidates, {int topK = 5, bool useLightweightRanker = false}) → Future<List<Document>>

Rerank documents based on relevance to query

retrieveAndGenerate(String query, {int topK = 5, double? minimumScore, String? namespace, Map<String, dynamic> filters = const {}, Map<String, dynamic> generationParams = const {}, List<String>? previousQueries, bool useHybridSearch = true}) → Future<String>

Retrieve and generate a response in one call with improved context handling

retrieveAndRerank(String query, {int retrievalTopK = 10, int rerankTopK = 5, double? minimumScore, String? namespace, Map<String, dynamic> filters = const {}, bool useLightweightRanker = false}) → Future<List<Document>>

Retrieve and rerank in one operation

retrieveRelevant(String query, {int topK = 5, double? minimumScore, String? namespace, Map<String, dynamic> filters = const {}, bool useCache = true}) → Future<List<Document>>

Retrieve relevant documents for a query with caching

timeWeightedRetrieval(String query, {int topK = 5, double? minimumScore, String? namespace, Map<String, dynamic> filters = const {}, double recencyWeight = 0.3, Duration freshnessWindow = const Duration(days: 30)}) → Future<List<Document>>

Retrieve documents with time-based weighting

toString() → String

A string representation of this object.

inherited

Operators

operator ==(Object other) → bool

The equality operator.

inherited