@retrivora-ai/rag-engine 1.9.2 → 1.9.6

package/README.md

@@ -103,6 +103,33 @@ await pipeline.ingest([{ docId: 'readme', content: 'Your document text here' }])
 const { reply, sources } = await pipeline.ask('What is the refund policy?');
 ```
 
+### 4. Use the Retrivora SDK facade
+
+```ts
+import { Retrivora } from '@retrivora-ai/rag-engine/server';
+
+const ai = new Retrivora({
+  llm: {
+    provider: 'openai',
+    model: 'gpt-5',
+  },
+  embedding: {
+    provider: 'openai',
+  },
+  vectorDatabase: {
+    provider: 'pinecone',
+  },
+  retrieval: {
+    strategy: 'hybrid',
+  },
+  workflow: {
+    type: 'agentic-rag',
+  },
+});
+
+const { reply, sources } = await ai.ask('What is the refund policy?');
+```
+
 ---
 
 ## Configuration Reference

package/dist/{ILLMProvider-Bw2A28nU.d.mts → ILLMProvider-DNhyOYoK.d.mts}

@@ -10,6 +10,13 @@ interface RagMessage extends ChatMessage {
     /** Full observability trace emitted by the backend. Only present on assistant messages. */
     trace?: ObservabilityTrace;
     createdAt: string;
+    /**
+     * Chain-of-thought reasoning from the LLM.
+     * Populated by native Anthropic extended thinking or simulated <think> tag stripping.
+     */
+    thinking?: string;
+    /** Wall-clock ms the model spent in the thinking phase. */
+    thinkingMs?: number;
 }
 interface ChatOptions {
     /** Override max tokens for this specific call */
@@ -227,6 +234,24 @@ interface RAGConfig {
      */
     vectorKeywords?: string[];
 }
+interface RetrievalConfig {
+    /** Retrieval strategy selected by the host app. */
+    strategy?: 'vector' | 'keyword' | 'hybrid' | 'multi-query' | 'self-query' | 'parent-document' | 'contextual-compression' | 'ensemble' | 'recursive' | 'agentic';
+    /** Number of documents/chunks retrieved per query. */
+    topK?: number;
+    /** Minimum similarity score to include a retrieved chunk. */
+    scoreThreshold?: number;
+    /** Whether to rerank retrieved chunks before generation. */
+    useReranking?: boolean;
+    /** Provider-specific retrieval options for future strategies. */
+    options?: Record<string, unknown>;
+}
+interface WorkflowConfig {
+    /** High-level workflow selector from the SDK prompt. */
+    type?: 'simple-rag' | 'rag' | 'hybrid-rag' | 'graph-rag' | 'agentic-rag' | 'agentic';
+    /** Future workflow/provider-specific options. */
+    options?: Record<string, unknown>;
+}
 interface RagConfig {
     /**
      * Unique identifier for the consuming project.
@@ -246,6 +271,21 @@ interface RagConfig {
     /** Optional Graph database configuration */
     graphDb?: GraphDBConfig;
 }
+/**
+ * Friendly SDK configuration accepted by the top-level Retrivora facade.
+ * It supports the prompt's `vectorDatabase`, `retrieval`, and `workflow`
+ * naming while normalizing to the internal RagConfig shape.
+ */
+type UniversalRagConfig = Partial<Omit<RagConfig, 'vectorDb' | 'rag' | 'llm' | 'embedding' | 'graphDb'>> & {
+    vectorDb?: Partial<VectorDBConfig>;
+    vectorDatabase?: Partial<VectorDBConfig>;
+    llm?: Partial<LLMConfig>;
+    embedding?: Partial<EmbeddingConfig>;
+    graphDb?: Partial<GraphDBConfig>;
+    retrieval?: RetrievalConfig;
+    rag?: RAGConfig;
+    workflow?: WorkflowConfig;
+};
 
 /** Per-stage latency breakdown for a single RAG request. */
 interface LatencyBreakdown {
@@ -389,4 +429,4 @@ interface ILLMProvider {
     ping(): Promise<boolean>;
 }
 
-export type { ChatMessage as C, EmbedOptions as E, GraphDBConfig as G, ILLMProvider as I, LLMConfig as L, ObservabilityTrace as O, Product as P, RagMessage as R, TokenUsage as T, UIConfig as U, VectorMatch as V, UseRagChatOptions as a, UseRagChatReturn as b, ChatOptions as c, ChatResponse as d, EmbeddingConfig as e, EmbeddingProvider as f, IngestDocument as g, LLMProvider as h, LatencyBreakdown as i, RAGConfig as j, RagConfig as k, RetrievedChunk as l, UpsertDocument as m, VectorDBConfig as n, VectorDBProvider as o, RetrievalResult as p, GraphNode as q, Edge as r, GraphSearchResult as s };
+export type { ChatMessage as C, EmbedOptions as E, GraphDBConfig as G, ILLMProvider as I, LLMConfig as L, ObservabilityTrace as O, Product as P, RagMessage as R, TokenUsage as T, UIConfig as U, VectorMatch as V, WorkflowConfig as W, UseRagChatOptions as a, UseRagChatReturn as b, ChatOptions as c, ChatResponse as d, EmbeddingConfig as e, EmbeddingProvider as f, IngestDocument as g, LLMProvider as h, LatencyBreakdown as i, RAGConfig as j, RagConfig as k, RetrievalConfig as l, RetrievedChunk as m, UniversalRagConfig as n, UpsertDocument as o, VectorDBConfig as p, VectorDBProvider as q, GraphNode as r, Edge as s, GraphSearchResult as t, RetrievalResult as u };

package/dist/{ILLMProvider-Bw2A28nU.d.ts → ILLMProvider-DNhyOYoK.d.ts}

@@ -10,6 +10,13 @@ interface RagMessage extends ChatMessage {
     /** Full observability trace emitted by the backend. Only present on assistant messages. */
     trace?: ObservabilityTrace;
     createdAt: string;
+    /**
+     * Chain-of-thought reasoning from the LLM.
+     * Populated by native Anthropic extended thinking or simulated <think> tag stripping.
+     */
+    thinking?: string;
+    /** Wall-clock ms the model spent in the thinking phase. */
+    thinkingMs?: number;
 }
 interface ChatOptions {
     /** Override max tokens for this specific call */
@@ -227,6 +234,24 @@ interface RAGConfig {
      */
     vectorKeywords?: string[];
 }
+interface RetrievalConfig {
+    /** Retrieval strategy selected by the host app. */
+    strategy?: 'vector' | 'keyword' | 'hybrid' | 'multi-query' | 'self-query' | 'parent-document' | 'contextual-compression' | 'ensemble' | 'recursive' | 'agentic';
+    /** Number of documents/chunks retrieved per query. */
+    topK?: number;
+    /** Minimum similarity score to include a retrieved chunk. */
+    scoreThreshold?: number;
+    /** Whether to rerank retrieved chunks before generation. */
+    useReranking?: boolean;
+    /** Provider-specific retrieval options for future strategies. */
+    options?: Record<string, unknown>;
+}
+interface WorkflowConfig {
+    /** High-level workflow selector from the SDK prompt. */
+    type?: 'simple-rag' | 'rag' | 'hybrid-rag' | 'graph-rag' | 'agentic-rag' | 'agentic';
+    /** Future workflow/provider-specific options. */
+    options?: Record<string, unknown>;
+}
 interface RagConfig {
     /**
      * Unique identifier for the consuming project.
@@ -246,6 +271,21 @@ interface RagConfig {
     /** Optional Graph database configuration */
     graphDb?: GraphDBConfig;
 }
+/**
+ * Friendly SDK configuration accepted by the top-level Retrivora facade.
+ * It supports the prompt's `vectorDatabase`, `retrieval`, and `workflow`
+ * naming while normalizing to the internal RagConfig shape.
+ */
+type UniversalRagConfig = Partial<Omit<RagConfig, 'vectorDb' | 'rag' | 'llm' | 'embedding' | 'graphDb'>> & {
+    vectorDb?: Partial<VectorDBConfig>;
+    vectorDatabase?: Partial<VectorDBConfig>;
+    llm?: Partial<LLMConfig>;
+    embedding?: Partial<EmbeddingConfig>;
+    graphDb?: Partial<GraphDBConfig>;
+    retrieval?: RetrievalConfig;
+    rag?: RAGConfig;
+    workflow?: WorkflowConfig;
+};
 
 /** Per-stage latency breakdown for a single RAG request. */
 interface LatencyBreakdown {
@@ -389,4 +429,4 @@ interface ILLMProvider {
     ping(): Promise<boolean>;
 }
 
-export type { ChatMessage as C, EmbedOptions as E, GraphDBConfig as G, ILLMProvider as I, LLMConfig as L, ObservabilityTrace as O, Product as P, RagMessage as R, TokenUsage as T, UIConfig as U, VectorMatch as V, UseRagChatOptions as a, UseRagChatReturn as b, ChatOptions as c, ChatResponse as d, EmbeddingConfig as e, EmbeddingProvider as f, IngestDocument as g, LLMProvider as h, LatencyBreakdown as i, RAGConfig as j, RagConfig as k, RetrievedChunk as l, UpsertDocument as m, VectorDBConfig as n, VectorDBProvider as o, RetrievalResult as p, GraphNode as q, Edge as r, GraphSearchResult as s };
+export type { ChatMessage as C, EmbedOptions as E, GraphDBConfig as G, ILLMProvider as I, LLMConfig as L, ObservabilityTrace as O, Product as P, RagMessage as R, TokenUsage as T, UIConfig as U, VectorMatch as V, WorkflowConfig as W, UseRagChatOptions as a, UseRagChatReturn as b, ChatOptions as c, ChatResponse as d, EmbeddingConfig as e, EmbeddingProvider as f, IngestDocument as g, LLMProvider as h, LatencyBreakdown as i, RAGConfig as j, RagConfig as k, RetrievalConfig as l, RetrievedChunk as m, UniversalRagConfig as n, UpsertDocument as o, VectorDBConfig as p, VectorDBProvider as q, GraphNode as r, Edge as s, GraphSearchResult as t, RetrievalResult as u };

package/dist/handlers/index.d.mts

@@ -1,3 +1,3 @@
-import '../ILLMProvider-Bw2A28nU.mjs';
-export { c as createChatHandler, d as createHealthHandler, e as createIngestHandler, f as createStreamHandler, k as createSuggestionsHandler, g as createUploadHandler, s as sseErrorFrame, h as sseFrame, i as sseMetaFrame, l as sseObservabilityFrame, j as sseTextFrame, m as sseUIFrame } from '../index-B70ZLkfG.mjs';
+import '../ILLMProvider-DNhyOYoK.mjs';
+export { c as createChatHandler, d as createHealthHandler, e as createIngestHandler, f as createStreamHandler, k as createSuggestionsHandler, g as createUploadHandler, s as sseErrorFrame, h as sseFrame, i as sseMetaFrame, l as sseObservabilityFrame, j as sseTextFrame, m as sseUIFrame } from '../index-CHL1jdYm.mjs';
 import 'next/server';

package/dist/handlers/index.d.ts

@@ -1,3 +1,3 @@
-import '../ILLMProvider-Bw2A28nU.js';
-export { c as createChatHandler, d as createHealthHandler, e as createIngestHandler, f as createStreamHandler, k as createSuggestionsHandler, g as createUploadHandler, s as sseErrorFrame, h as sseFrame, i as sseMetaFrame, l as sseObservabilityFrame, j as sseTextFrame, m as sseUIFrame } from '../index-DVu-mkAM.js';
+import '../ILLMProvider-DNhyOYoK.js';
+export { c as createChatHandler, d as createHealthHandler, e as createIngestHandler, f as createStreamHandler, k as createSuggestionsHandler, g as createUploadHandler, s as sseErrorFrame, h as sseFrame, i as sseMetaFrame, l as sseObservabilityFrame, j as sseTextFrame, m as sseUIFrame } from '../index-Hgbwl9X4.js';
 import 'next/server';