@retrivora-ai/rag-engine 1.9.3 → 1.9.7

package/dist/server.d.mts

@@ -1,7 +1,8 @@
-import { n as VectorDBConfig, L as LLMConfig, e as EmbeddingConfig, k as RagConfig, I as ILLMProvider, g as IngestDocument, C as ChatMessage, d as ChatResponse, p as RetrievalResult, m as UpsertDocument, V as VectorMatch, G as GraphDBConfig, q as GraphNode, r as Edge, s as GraphSearchResult, o as VectorDBProvider, h as LLMProvider, f as EmbeddingProvider, j as RAGConfig, U as UIConfig, c as ChatOptions, E as EmbedOptions } from './ILLMProvider-Bw2A28nU.mjs';
-export { C as Chunk, a as ChunkOptions, D as DocumentChunker } from './DocumentChunker-Dh9TvmGG.mjs';
-import { H as HealthCheckResult, I as IProviderValidator, a as IProviderHealthChecker } from './index-B70ZLkfG.mjs';
-export { C as ConfigValidator, V as ValidationError, b as VectorPlugin, c as createChatHandler, d as createHealthHandler, e as createIngestHandler, f as createStreamHandler, g as createUploadHandler, s as sseErrorFrame, h as sseFrame, i as sseMetaFrame, j as sseTextFrame } from './index-B70ZLkfG.mjs';
+import { p as VectorDBConfig, L as LLMConfig, e as EmbeddingConfig, k as RagConfig, n as UniversalRagConfig, o as UpsertDocument, V as VectorMatch, G as GraphDBConfig, r as GraphNode, s as Edge, t as GraphSearchResult, I as ILLMProvider, q as VectorDBProvider, h as LLMProvider, f as EmbeddingProvider, j as RAGConfig, U as UIConfig, C as ChatMessage, c as ChatOptions, E as EmbedOptions } from './ILLMProvider-Bhk6zJOK.mjs';
+export { d as ChatResponse, g as IngestDocument, l as RetrievalConfig, W as WorkflowConfig } from './ILLMProvider-Bhk6zJOK.mjs';
+export { A as AuthenticationException, C as Chunk, a as ChunkOptions, b as ConfigurationException, D as DocumentChunker, E as EmbeddingFailedException, g as Pipeline, P as ProviderNotFoundException, R as RateLimitException, c as RetrievalException, d as Retrivora, e as RetrivoraError, f as RetrivoraErrorCode, w as wrapError } from './index-BJ4cd-t5.mjs';
+import { H as HealthCheckResult, I as IProviderValidator, a as IProviderHealthChecker } from './index-C3SVtPYg.mjs';
+export { C as ConfigValidator, V as ValidationError, b as VectorPlugin, c as createChatHandler, d as createHealthHandler, e as createIngestHandler, f as createRagHandler, g as createStreamHandler, h as createUploadHandler, s as sseErrorFrame, i as sseFrame, j as sseMetaFrame, k as sseTextFrame } from './index-C3SVtPYg.mjs';
 import 'next/server';
 
 /**
@@ -95,94 +96,6 @@ declare class ProviderHealthCheck {
     }>;
 }
 
-/**
- * Pipeline — orchestrates the RAG flow: Embed → Search → Augment → Generate.
- *
- * Features:
- * - Lazy initialization of providers
- * - Batch processing with retry logic
- * - Smart embedding strategy (integrated / separate / external)
- * - Error recovery for transient failures
- * - Multi-tenancy support via namespacing
- * - LRU-bounded embedding cache (max 500 entries, prevents memory leaks)
- * - Full observability tracing (latency, tokens, hallucination scoring)
- */
-declare class Pipeline {
-    private config;
-    private vectorDB;
-    private graphDB?;
-    private llmProvider;
-    private embeddingProvider;
-    private llmRouter;
-    private chunker;
-    private llamaIngestor?;
-    private entityExtractor?;
-    private reranker;
-    private agent?;
-    /** LRU-bounded cache: avoids re-embedding identical queries within the same process. */
-    private embeddingCache;
-    private initialised;
-    constructor(config: RagConfig);
-    /**
-     * Expose the underlying LLM provider (set after initialize()).
-     * Used by the stream handler to pass to UITransformer.analyzeAndDecide().
-     */
-    getLLMProvider(): ILLMProvider | undefined;
-    initialize(): Promise<void>;
-    /**
-     * Ingest documents with automatic chunking, embedding, and batch upsert.
-     */
-    ingest(documents: IngestDocument[], namespace?: string): Promise<Array<{
-        docId: string | number;
-        chunksIngested: number;
-    }>>;
-    /** Step 1: Chunk the document content. */
-    private prepareChunks;
-    /**
-     * Step 2: Generate embeddings for chunks with retry logic.
-     * Uses batchEmbed when available for efficiency; falls back to sequential embedding.
-     */
-    private processEmbeddings;
-    /** Step 3: Upsert chunks to vector database with retry logic. */
-    private processUpserts;
-    /** Step 4: Optional graph-based entity extraction and ingestion. */
-    private processGraphIngestion;
-    ask(question: string, history?: ChatMessage[], namespace?: string): Promise<ChatResponse>;
-    /**
-     * High-performance streaming RAG flow.
-     * Yields text chunks first, then the retrieval metadata + observability trace at the end.
-     *
-     * Latency optimizations:
-     * - Strategy classification runs in parallel with query embedding (saves ~400ms)
-     * - Hallucination scoring is fire-and-forget (doesn't block metadata yield)
-     * - UITransformation is computed after text streaming and emitted with metadata
-     * - SchemaMapper.train runs while answer generation streams
-     */
-    askStream(question: string, history?: ChatMessage[], namespace?: string): AsyncIterable<string | ChatResponse>;
-    /**
-     * Universal retrieval method combining all enabled providers.
-     * Uses an LRU-bounded embedding cache to avoid re-embedding the same query.
-     */
-    private generateUiTransformation;
-    private applyStructuredFilters;
-    private resolveNumericPredicateValue;
-    private extractNumericValueFromContent;
-    private matchesNumericPredicate;
-    private normalizeComparableField;
-    private fieldSimilarityScore;
-    private fieldTokens;
-    private toFiniteNumber;
-    retrieve(query: string, options: {
-        namespace?: string;
-        topK?: number;
-        filter?: Record<string, unknown>;
-    }): Promise<RetrievalResult>;
-    /** Rewrite the user query for better retrieval performance. */
-    private rewriteQuery;
-    /** Generate 3-5 short, relevant questions based on the vector database content. */
-    getSuggestions(query: string, namespace?: string): Promise<string[]>;
-}
-
 /**
  * ConfigResolver — validates and normalizes host configuration.
  * It merges host-provided config with environment defaults.
@@ -198,10 +111,17 @@ declare class ConfigResolver {
      * @param hostConfig - Partial configuration passed from the host application.
      */
     static resolve(hostConfig?: Partial<RagConfig>, env?: Record<string, string | undefined>): RagConfig;
+    /**
+     * Resolves the public SDK config shape used by `new Retrivora({...})`.
+     * Supports aliases from the product prompt while preserving existing env
+     * fallback behavior.
+     */
+    static resolveUniversal(hostConfig?: UniversalRagConfig, env?: Record<string, string | undefined>): RagConfig;
     /**
      * Validates the configuration for required fields.
      */
     static validate(config: RagConfig): void;
+    private static mergeRetrievalWorkflow;
 }
 
 /**
@@ -313,6 +233,8 @@ declare class ProviderRegistry {
     static createVectorProvider(config: VectorDBConfig): Promise<BaseVectorProvider>;
     static createGraphProvider(config: GraphDBConfig): Promise<BaseGraphProvider>;
     static createLLMProvider(llmConfig: LLMConfig, embeddingConfig?: EmbeddingConfig): ILLMProvider;
+    static registerLLMProvider(name: string, factory: (config: LLMConfig) => ILLMProvider): void;
+    static createEmbeddingProvider(embeddingConfig: EmbeddingConfig): ILLMProvider;
 }
 
 /**
@@ -974,7 +896,6 @@ declare class OllamaProvider implements ILLMProvider {
     static getHealthChecker(): IProviderHealthChecker;
     chat(messages: ChatMessage[], context: string, options?: ChatOptions): Promise<string>;
     chatStream(messages: ChatMessage[], context: string, options?: ChatOptions): AsyncIterable<string>;
-    private buildSystemPrompt;
     embed(text: string, options?: EmbedOptions): Promise<number[]>;
     batchEmbed(texts: string[], options?: EmbedOptions): Promise<number[][]>;
     ping(): Promise<boolean>;
@@ -1003,4 +924,4 @@ declare class UniversalLLMAdapter implements ILLMProvider {
     ping(): Promise<boolean>;
 }
 
-export { AnthropicProvider, BaseVectorProvider, type BatchOptions, BatchProcessor, type BatchResult, ChatMessage, ChatOptions, ChatResponse, ChromaDBProvider, ConfigBuilder, ConfigResolver, DocumentParser, EmbedOptions, EmbeddingConfig, EmbeddingProvider, EmbeddingStrategy, EmbeddingStrategyResolver, HealthCheckResult, ILLMProvider, IProviderHealthChecker, IProviderValidator, IngestDocument, LLMConfig, LLMFactory, LLMProvider, LLM_PROFILES, MilvusProvider, MongoDBProvider, MultiTablePostgresProvider, OllamaProvider, OpenAIProvider, PRESETS, PineconeProvider, Pipeline, PostgreSQLProvider, ProviderHealthCheck, ProviderRegistry, QdrantProvider, RAGConfig, RagConfig, RedisProvider, UIConfig, UniversalLLMAdapter, UniversalVectorProvider, UpsertDocument, VECTOR_PROFILES, VectorDBConfig, VectorDBProvider, VectorMatch, WeaviateProvider, createFromPreset, getRagConfig };
+export { AnthropicProvider, BaseVectorProvider, type BatchOptions, BatchProcessor, type BatchResult, ChatMessage, ChatOptions, ChromaDBProvider, ConfigBuilder, ConfigResolver, DocumentParser, EmbedOptions, EmbeddingConfig, EmbeddingProvider, EmbeddingStrategy, EmbeddingStrategyResolver, HealthCheckResult, ILLMProvider, IProviderHealthChecker, IProviderValidator, LLMConfig, LLMFactory, LLMProvider, LLM_PROFILES, MilvusProvider, MongoDBProvider, MultiTablePostgresProvider, OllamaProvider, OpenAIProvider, PRESETS, PineconeProvider, PostgreSQLProvider, ProviderHealthCheck, ProviderRegistry, QdrantProvider, RAGConfig, RagConfig, RedisProvider, UIConfig, UniversalLLMAdapter, UniversalRagConfig, UniversalVectorProvider, UpsertDocument, VECTOR_PROFILES, VectorDBConfig, VectorDBProvider, VectorMatch, WeaviateProvider, createFromPreset, getRagConfig };

package/dist/server.d.ts

@@ -1,7 +1,8 @@
-import { n as VectorDBConfig, L as LLMConfig, e as EmbeddingConfig, k as RagConfig, I as ILLMProvider, g as IngestDocument, C as ChatMessage, d as ChatResponse, p as RetrievalResult, m as UpsertDocument, V as VectorMatch, G as GraphDBConfig, q as GraphNode, r as Edge, s as GraphSearchResult, o as VectorDBProvider, h as LLMProvider, f as EmbeddingProvider, j as RAGConfig, U as UIConfig, c as ChatOptions, E as EmbedOptions } from './ILLMProvider-Bw2A28nU.js';
-export { C as Chunk, a as ChunkOptions, D as DocumentChunker } from './DocumentChunker-Dh9TvmGG.js';
-import { H as HealthCheckResult, I as IProviderValidator, a as IProviderHealthChecker } from './index-DVu-mkAM.js';
-export { C as ConfigValidator, V as ValidationError, b as VectorPlugin, c as createChatHandler, d as createHealthHandler, e as createIngestHandler, f as createStreamHandler, g as createUploadHandler, s as sseErrorFrame, h as sseFrame, i as sseMetaFrame, j as sseTextFrame } from './index-DVu-mkAM.js';
+import { p as VectorDBConfig, L as LLMConfig, e as EmbeddingConfig, k as RagConfig, n as UniversalRagConfig, o as UpsertDocument, V as VectorMatch, G as GraphDBConfig, r as GraphNode, s as Edge, t as GraphSearchResult, I as ILLMProvider, q as VectorDBProvider, h as LLMProvider, f as EmbeddingProvider, j as RAGConfig, U as UIConfig, C as ChatMessage, c as ChatOptions, E as EmbedOptions } from './ILLMProvider-Bhk6zJOK.js';
+export { d as ChatResponse, g as IngestDocument, l as RetrievalConfig, W as WorkflowConfig } from './ILLMProvider-Bhk6zJOK.js';
+export { A as AuthenticationException, C as Chunk, a as ChunkOptions, b as ConfigurationException, D as DocumentChunker, E as EmbeddingFailedException, g as Pipeline, P as ProviderNotFoundException, R as RateLimitException, c as RetrievalException, d as Retrivora, e as RetrivoraError, f as RetrivoraErrorCode, w as wrapError } from './index-B9J_XEh0.js';
+import { H as HealthCheckResult, I as IProviderValidator, a as IProviderHealthChecker } from './index-Bu7T6xgr.js';
+export { C as ConfigValidator, V as ValidationError, b as VectorPlugin, c as createChatHandler, d as createHealthHandler, e as createIngestHandler, f as createRagHandler, g as createStreamHandler, h as createUploadHandler, s as sseErrorFrame, i as sseFrame, j as sseMetaFrame, k as sseTextFrame } from './index-Bu7T6xgr.js';
 import 'next/server';
 
 /**
@@ -95,94 +96,6 @@ declare class ProviderHealthCheck {
     }>;
 }
 
-/**
- * Pipeline — orchestrates the RAG flow: Embed → Search → Augment → Generate.
- *
- * Features:
- * - Lazy initialization of providers
- * - Batch processing with retry logic
- * - Smart embedding strategy (integrated / separate / external)
- * - Error recovery for transient failures
- * - Multi-tenancy support via namespacing
- * - LRU-bounded embedding cache (max 500 entries, prevents memory leaks)
- * - Full observability tracing (latency, tokens, hallucination scoring)
- */
-declare class Pipeline {
-    private config;
-    private vectorDB;
-    private graphDB?;
-    private llmProvider;
-    private embeddingProvider;
-    private llmRouter;
-    private chunker;
-    private llamaIngestor?;
-    private entityExtractor?;
-    private reranker;
-    private agent?;
-    /** LRU-bounded cache: avoids re-embedding identical queries within the same process. */
-    private embeddingCache;
-    private initialised;
-    constructor(config: RagConfig);
-    /**
-     * Expose the underlying LLM provider (set after initialize()).
-     * Used by the stream handler to pass to UITransformer.analyzeAndDecide().
-     */
-    getLLMProvider(): ILLMProvider | undefined;
-    initialize(): Promise<void>;
-    /**
-     * Ingest documents with automatic chunking, embedding, and batch upsert.
-     */
-    ingest(documents: IngestDocument[], namespace?: string): Promise<Array<{
-        docId: string | number;
-        chunksIngested: number;
-    }>>;
-    /** Step 1: Chunk the document content. */
-    private prepareChunks;
-    /**
-     * Step 2: Generate embeddings for chunks with retry logic.
-     * Uses batchEmbed when available for efficiency; falls back to sequential embedding.
-     */
-    private processEmbeddings;
-    /** Step 3: Upsert chunks to vector database with retry logic. */
-    private processUpserts;
-    /** Step 4: Optional graph-based entity extraction and ingestion. */
-    private processGraphIngestion;
-    ask(question: string, history?: ChatMessage[], namespace?: string): Promise<ChatResponse>;
-    /**
-     * High-performance streaming RAG flow.
-     * Yields text chunks first, then the retrieval metadata + observability trace at the end.
-     *
-     * Latency optimizations:
-     * - Strategy classification runs in parallel with query embedding (saves ~400ms)
-     * - Hallucination scoring is fire-and-forget (doesn't block metadata yield)
-     * - UITransformation is computed after text streaming and emitted with metadata
-     * - SchemaMapper.train runs while answer generation streams
-     */
-    askStream(question: string, history?: ChatMessage[], namespace?: string): AsyncIterable<string | ChatResponse>;
-    /**
-     * Universal retrieval method combining all enabled providers.
-     * Uses an LRU-bounded embedding cache to avoid re-embedding the same query.
-     */
-    private generateUiTransformation;
-    private applyStructuredFilters;
-    private resolveNumericPredicateValue;
-    private extractNumericValueFromContent;
-    private matchesNumericPredicate;
-    private normalizeComparableField;
-    private fieldSimilarityScore;
-    private fieldTokens;
-    private toFiniteNumber;
-    retrieve(query: string, options: {
-        namespace?: string;
-        topK?: number;
-        filter?: Record<string, unknown>;
-    }): Promise<RetrievalResult>;
-    /** Rewrite the user query for better retrieval performance. */
-    private rewriteQuery;
-    /** Generate 3-5 short, relevant questions based on the vector database content. */
-    getSuggestions(query: string, namespace?: string): Promise<string[]>;
-}
-
 /**
  * ConfigResolver — validates and normalizes host configuration.
  * It merges host-provided config with environment defaults.
@@ -198,10 +111,17 @@ declare class ConfigResolver {
      * @param hostConfig - Partial configuration passed from the host application.
      */
     static resolve(hostConfig?: Partial<RagConfig>, env?: Record<string, string | undefined>): RagConfig;
+    /**
+     * Resolves the public SDK config shape used by `new Retrivora({...})`.
+     * Supports aliases from the product prompt while preserving existing env
+     * fallback behavior.
+     */
+    static resolveUniversal(hostConfig?: UniversalRagConfig, env?: Record<string, string | undefined>): RagConfig;
     /**
      * Validates the configuration for required fields.
      */
     static validate(config: RagConfig): void;
+    private static mergeRetrievalWorkflow;
 }
 
 /**
@@ -313,6 +233,8 @@ declare class ProviderRegistry {
     static createVectorProvider(config: VectorDBConfig): Promise<BaseVectorProvider>;
     static createGraphProvider(config: GraphDBConfig): Promise<BaseGraphProvider>;
     static createLLMProvider(llmConfig: LLMConfig, embeddingConfig?: EmbeddingConfig): ILLMProvider;
+    static registerLLMProvider(name: string, factory: (config: LLMConfig) => ILLMProvider): void;
+    static createEmbeddingProvider(embeddingConfig: EmbeddingConfig): ILLMProvider;
 }
 
 /**
@@ -974,7 +896,6 @@ declare class OllamaProvider implements ILLMProvider {
     static getHealthChecker(): IProviderHealthChecker;
     chat(messages: ChatMessage[], context: string, options?: ChatOptions): Promise<string>;
     chatStream(messages: ChatMessage[], context: string, options?: ChatOptions): AsyncIterable<string>;
-    private buildSystemPrompt;
     embed(text: string, options?: EmbedOptions): Promise<number[]>;
     batchEmbed(texts: string[], options?: EmbedOptions): Promise<number[][]>;
     ping(): Promise<boolean>;
@@ -1003,4 +924,4 @@ declare class UniversalLLMAdapter implements ILLMProvider {
     ping(): Promise<boolean>;
 }
 
-export { AnthropicProvider, BaseVectorProvider, type BatchOptions, BatchProcessor, type BatchResult, ChatMessage, ChatOptions, ChatResponse, ChromaDBProvider, ConfigBuilder, ConfigResolver, DocumentParser, EmbedOptions, EmbeddingConfig, EmbeddingProvider, EmbeddingStrategy, EmbeddingStrategyResolver, HealthCheckResult, ILLMProvider, IProviderHealthChecker, IProviderValidator, IngestDocument, LLMConfig, LLMFactory, LLMProvider, LLM_PROFILES, MilvusProvider, MongoDBProvider, MultiTablePostgresProvider, OllamaProvider, OpenAIProvider, PRESETS, PineconeProvider, Pipeline, PostgreSQLProvider, ProviderHealthCheck, ProviderRegistry, QdrantProvider, RAGConfig, RagConfig, RedisProvider, UIConfig, UniversalLLMAdapter, UniversalVectorProvider, UpsertDocument, VECTOR_PROFILES, VectorDBConfig, VectorDBProvider, VectorMatch, WeaviateProvider, createFromPreset, getRagConfig };
+export { AnthropicProvider, BaseVectorProvider, type BatchOptions, BatchProcessor, type BatchResult, ChatMessage, ChatOptions, ChromaDBProvider, ConfigBuilder, ConfigResolver, DocumentParser, EmbedOptions, EmbeddingConfig, EmbeddingProvider, EmbeddingStrategy, EmbeddingStrategyResolver, HealthCheckResult, ILLMProvider, IProviderHealthChecker, IProviderValidator, LLMConfig, LLMFactory, LLMProvider, LLM_PROFILES, MilvusProvider, MongoDBProvider, MultiTablePostgresProvider, OllamaProvider, OpenAIProvider, PRESETS, PineconeProvider, PostgreSQLProvider, ProviderHealthCheck, ProviderRegistry, QdrantProvider, RAGConfig, RagConfig, RedisProvider, UIConfig, UniversalLLMAdapter, UniversalRagConfig, UniversalVectorProvider, UpsertDocument, VECTOR_PROFILES, VectorDBConfig, VectorDBProvider, VectorMatch, WeaviateProvider, createFromPreset, getRagConfig };