@retrivora-ai/rag-engine 0.1.3 → 0.1.4

package/dist/ChromaDBProvider-QNI7UCX4.mjs

@@ -0,0 +1,8 @@
+import {
+  ChromaDBProvider
+} from "./chunk-HUGLYKD6.mjs";
+import "./chunk-VOIWNO5O.mjs";
+import "./chunk-I4E63NIC.mjs";
+export {
+  ChromaDBProvider
+};

package/dist/DocumentChunker-BQ5kQD7B.d.ts

@@ -0,0 +1,155 @@
+import { e as RagConfig, C as ChatMessage, b as ChatResponse, d as IngestDocument, g as VectorDBConfig, f as UpsertDocument, V as VectorMatch, L as LLMConfig, c as EmbeddingConfig, I as ILLMProvider } from './Pipeline-Bo6CUCox.js';
+
+/**
+ * VectorPlugin — main orchestrator class.
+ * This is the primary interface for host applications.
+ */
+declare class VectorPlugin {
+    private config;
+    private pipeline;
+    /**
+     * Initializes the plugin with the host configuration.
+     * @param hostConfig - Configuration object passed from the host application.
+     */
+    constructor(hostConfig?: Partial<RagConfig>);
+    /**
+     * Get the current resolved configuration.
+     */
+    getConfig(): RagConfig;
+    /**
+     * Run a chat query.
+     */
+    chat(message: string, history?: ChatMessage[], namespace?: string): Promise<ChatResponse>;
+    /**
+     * Ingest documents into the vector database.
+     */
+    ingest(documents: IngestDocument[], namespace?: string): Promise<Array<{
+        docId: string;
+        chunksIngested: number;
+    }>>;
+    /**
+     * Check the health of the connected providers.
+     */
+    health(): Promise<Record<string, boolean>>;
+}
+
+/**
+ * ConfigResolver — validates and normalizes host configuration.
+ * It merges host-provided config with environment defaults.
+ */
+declare class ConfigResolver {
+    /**
+     * Resolves the final configuration by merging host-provided config with defaults.
+     * @param hostConfig - Partial configuration passed from the host application.
+     */
+    static resolve(hostConfig?: Partial<RagConfig>): RagConfig;
+    /**
+     * Validates the configuration for required fields.
+     */
+    static validate(config: RagConfig): void;
+}
+
+/**
+ * BaseVectorProvider — Abstract base class for all vector database providers.
+ * Each provider (Pinecone, Milvus, Redis, etc.) must extend this class.
+ */
+declare abstract class BaseVectorProvider {
+    protected readonly config: VectorDBConfig;
+    protected readonly indexName: string;
+    constructor(config: VectorDBConfig);
+    /**
+     * Initialise the connection (create tables, verify index, etc.)
+     */
+    abstract initialize(): Promise<void>;
+    /**
+     * Insert or update a single vector + content pair.
+     */
+    abstract upsert(doc: UpsertDocument, namespace?: string): Promise<void>;
+    /**
+     * Batch upsert for efficient ingestion of many documents.
+     */
+    abstract batchUpsert(docs: UpsertDocument[], namespace?: string): Promise<void>;
+    /**
+     * Find the top-K most similar vectors to the query vector.
+     */
+    abstract query(vector: number[], topK: number, namespace?: string, filter?: Record<string, unknown>): Promise<VectorMatch[]>;
+    /**
+     * Delete a stored vector by ID.
+     */
+    abstract delete(id: string, namespace?: string): Promise<void>;
+    /**
+     * Delete all vectors in a namespace (full data reset for a project).
+     */
+    abstract deleteNamespace(namespace: string): Promise<void>;
+    /**
+     * Check if the underlying DB is reachable.
+     */
+    abstract ping(): Promise<boolean>;
+    /**
+     * Gracefully close the connection.
+     */
+    abstract disconnect(): Promise<void>;
+}
+
+/**
+ * ProviderRegistry — dynamic provider loader for Vector DBs and LLMs.
+ */
+declare class ProviderRegistry {
+    private static vectorProviders;
+    /**
+     * Register a custom vector provider.
+     */
+    static registerVectorProvider(name: string, providerClass: new (config: VectorDBConfig) => BaseVectorProvider): void;
+    /**
+     * Creates a vector database provider based on the configuration.
+     */
+    static createVectorProvider(config: VectorDBConfig): Promise<BaseVectorProvider>;
+    /**
+     * Creates an LLM provider based on the configuration.
+     */
+    static createLLMProvider(llmConfig: LLMConfig, embeddingConfig?: EmbeddingConfig): ILLMProvider;
+}
+
+/**
+ * DocumentChunker — splits long text into overlapping chunks
+ * suitable for vector embedding and retrieval.
+ *
+ * Strategy: sentence-aware sliding window
+ *   1. Split on sentence boundaries
+ *   2. Accumulate sentences into chunks up to `chunkSize` characters
+ *   3. Carry over `chunkOverlap` characters from the previous chunk
+ */
+interface Chunk {
+    id: string;
+    content: string;
+    metadata?: Record<string, unknown>;
+}
+interface ChunkOptions {
+    /** Target chunk size in characters (default 1000) */
+    chunkSize?: number;
+    /** Overlap between consecutive chunks in characters (default 200) */
+    chunkOverlap?: number;
+    /** Source document identifier used as ID prefix */
+    docId?: string;
+    /** Extra metadata to attach to every chunk */
+    metadata?: Record<string, unknown>;
+}
+declare class DocumentChunker {
+    private readonly chunkSize;
+    private readonly chunkOverlap;
+    constructor(chunkSize?: number, chunkOverlap?: number);
+    /**
+     * Split a single text string into overlapping chunks.
+     */
+    chunk(text: string, options?: ChunkOptions): Chunk[];
+    /**
+     * Chunk multiple documents at once.
+     */
+    chunkMany(documents: Array<{
+        content: string;
+        docId?: string;
+        metadata?: Record<string, unknown>;
+    }>): Chunk[];
+}
+
+export { BaseVectorProvider as B, type Chunk as C, DocumentChunker as D, ProviderRegistry as P, VectorPlugin as V, type ChunkOptions as a, ConfigResolver as b };

package/dist/DocumentChunker-D9-fObJp.d.mts

@@ -0,0 +1,155 @@
+import { e as RagConfig, C as ChatMessage, b as ChatResponse, d as IngestDocument, g as VectorDBConfig, f as UpsertDocument, V as VectorMatch, L as LLMConfig, c as EmbeddingConfig, I as ILLMProvider } from './Pipeline-Bo6CUCox.mjs';
+
+/**
+ * VectorPlugin — main orchestrator class.
+ * This is the primary interface for host applications.
+ */
+declare class VectorPlugin {
+    private config;
+    private pipeline;
+    /**
+     * Initializes the plugin with the host configuration.
+     * @param hostConfig - Configuration object passed from the host application.
+     */
+    constructor(hostConfig?: Partial<RagConfig>);
+    /**
+     * Get the current resolved configuration.
+     */
+    getConfig(): RagConfig;
+    /**
+     * Run a chat query.
+     */
+    chat(message: string, history?: ChatMessage[], namespace?: string): Promise<ChatResponse>;
+    /**
+     * Ingest documents into the vector database.
+     */
+    ingest(documents: IngestDocument[], namespace?: string): Promise<Array<{
+        docId: string;
+        chunksIngested: number;
+    }>>;
+    /**
+     * Check the health of the connected providers.
+     */
+    health(): Promise<Record<string, boolean>>;
+}
+
+/**
+ * ConfigResolver — validates and normalizes host configuration.
+ * It merges host-provided config with environment defaults.
+ */
+declare class ConfigResolver {
+    /**
+     * Resolves the final configuration by merging host-provided config with defaults.
+     * @param hostConfig - Partial configuration passed from the host application.
+     */
+    static resolve(hostConfig?: Partial<RagConfig>): RagConfig;
+    /**
+     * Validates the configuration for required fields.
+     */
+    static validate(config: RagConfig): void;
+}
+
+/**
+ * BaseVectorProvider — Abstract base class for all vector database providers.
+ * Each provider (Pinecone, Milvus, Redis, etc.) must extend this class.
+ */
+declare abstract class BaseVectorProvider {
+    protected readonly config: VectorDBConfig;
+    protected readonly indexName: string;
+    constructor(config: VectorDBConfig);
+    /**
+     * Initialise the connection (create tables, verify index, etc.)
+     */
+    abstract initialize(): Promise<void>;
+    /**
+     * Insert or update a single vector + content pair.
+     */
+    abstract upsert(doc: UpsertDocument, namespace?: string): Promise<void>;
+    /**
+     * Batch upsert for efficient ingestion of many documents.
+     */
+    abstract batchUpsert(docs: UpsertDocument[], namespace?: string): Promise<void>;
+    /**
+     * Find the top-K most similar vectors to the query vector.
+     */
+    abstract query(vector: number[], topK: number, namespace?: string, filter?: Record<string, unknown>): Promise<VectorMatch[]>;
+    /**
+     * Delete a stored vector by ID.
+     */
+    abstract delete(id: string, namespace?: string): Promise<void>;
+    /**
+     * Delete all vectors in a namespace (full data reset for a project).
+     */
+    abstract deleteNamespace(namespace: string): Promise<void>;
+    /**
+     * Check if the underlying DB is reachable.
+     */
+    abstract ping(): Promise<boolean>;
+    /**
+     * Gracefully close the connection.
+     */
+    abstract disconnect(): Promise<void>;
+}
+
+/**
+ * ProviderRegistry — dynamic provider loader for Vector DBs and LLMs.
+ */
+declare class ProviderRegistry {
+    private static vectorProviders;
+    /**
+     * Register a custom vector provider.
+     */
+    static registerVectorProvider(name: string, providerClass: new (config: VectorDBConfig) => BaseVectorProvider): void;
+    /**
+     * Creates a vector database provider based on the configuration.
+     */
+    static createVectorProvider(config: VectorDBConfig): Promise<BaseVectorProvider>;
+    /**
+     * Creates an LLM provider based on the configuration.
+     */
+    static createLLMProvider(llmConfig: LLMConfig, embeddingConfig?: EmbeddingConfig): ILLMProvider;
+}
+
+/**
+ * DocumentChunker — splits long text into overlapping chunks
+ * suitable for vector embedding and retrieval.
+ *
+ * Strategy: sentence-aware sliding window
+ *   1. Split on sentence boundaries
+ *   2. Accumulate sentences into chunks up to `chunkSize` characters
+ *   3. Carry over `chunkOverlap` characters from the previous chunk
+ */
+interface Chunk {
+    id: string;
+    content: string;
+    metadata?: Record<string, unknown>;
+}
+interface ChunkOptions {
+    /** Target chunk size in characters (default 1000) */
+    chunkSize?: number;
+    /** Overlap between consecutive chunks in characters (default 200) */
+    chunkOverlap?: number;
+    /** Source document identifier used as ID prefix */
+    docId?: string;
+    /** Extra metadata to attach to every chunk */
+    metadata?: Record<string, unknown>;
+}
+declare class DocumentChunker {
+    private readonly chunkSize;
+    private readonly chunkOverlap;
+    constructor(chunkSize?: number, chunkOverlap?: number);
+    /**
+     * Split a single text string into overlapping chunks.
+     */
+    chunk(text: string, options?: ChunkOptions): Chunk[];
+    /**
+     * Chunk multiple documents at once.
+     */
+    chunkMany(documents: Array<{
+        content: string;
+        docId?: string;
+        metadata?: Record<string, unknown>;
+    }>): Chunk[];
+}
+
+export { BaseVectorProvider as B, type Chunk as C, DocumentChunker as D, ProviderRegistry as P, VectorPlugin as V, type ChunkOptions as a, ConfigResolver as b };

package/dist/LLMFactory-XC55FTD2.mjs

@@ -0,0 +1,7 @@
+import {
+  LLMFactory
+} from "./chunk-GD3QJFNN.mjs";
+import "./chunk-I4E63NIC.mjs";
+export {
+  LLMFactory
+};

package/dist/MilvusProvider-OO6QGZDZ.mjs

@@ -0,0 +1,8 @@
+import {
+  MilvusProvider
+} from "./chunk-7NXI6ZWX.mjs";
+import "./chunk-VOIWNO5O.mjs";
+import "./chunk-I4E63NIC.mjs";
+export {
+  MilvusProvider
+};

package/dist/MongoDBProvider-WWVJG3WT.mjs

@@ -0,0 +1,8 @@
+import {
+  MongoDBProvider
+} from "./chunk-QEYVWVT5.mjs";
+import "./chunk-VOIWNO5O.mjs";
+import "./chunk-I4E63NIC.mjs";
+export {
+  MongoDBProvider
+};

package/dist/PineconeProvider-NJ675H7U.mjs

@@ -0,0 +1,8 @@
+import {
+  PineconeProvider
+} from "./chunk-AALIF3AL.mjs";
+import "./chunk-VOIWNO5O.mjs";
+import "./chunk-I4E63NIC.mjs";
+export {
+  PineconeProvider
+};

package/dist/{RAGPipeline-BmkIv1HD.d.mts → Pipeline-Bo6CUCox.d.mts}

@@ -1,120 +1,9 @@
-/**
- * Generic LLM Provider interface.
- * Covers both chat completion and embedding generation so a single
- * provider can handle both responsibilities when appropriate.
- */
-interface ChatMessage {
-    role: 'user' | 'assistant' | 'system';
-    content: string;
-}
-interface ChatOptions {
-    /** Override max tokens for this specific call */
-    maxTokens?: number;
-    /** Override temperature for this specific call */
-    temperature?: number;
-    /** Stop sequences */
-    stop?: string[];
-}
-interface EmbedOptions {
-    /** Override model for this specific embed call */
-    model?: string;
-}
-interface ILLMProvider {
-    /**
-     * Send a chat completion request.
-     * @param messages – the full conversation history
-     * @param context  – retrieved RAG context to inject
-     * @param options  – optional per-call overrides
-     * @returns        – the assistant's reply text
-     */
-    chat(messages: ChatMessage[], context: string, options?: ChatOptions): Promise<string>;
-    /**
-     * Generate an embedding vector for the given text.
-     * @param text    – text to embed
-     * @param options – optional overrides
-     * @returns       – float array (the embedding)
-     */
-    embed(text: string, options?: EmbedOptions): Promise<number[]>;
-    /**
-     * Generate embedding vectors for multiple texts in a single batch.
-     * @param texts   – array of texts to embed
-     * @param options – optional overrides
-     * @returns       – array of float arrays
-     */
-    batchEmbed(texts: string[], options?: EmbedOptions): Promise<number[][]>;
-    /**
-     * Check if the provider endpoint is reachable.
-     */
-    ping(): Promise<boolean>;
-}
-
-/**
- * Generic Vector Database interface.
- * Any adapter (Pinecone, pgVector, REST, Chroma, Qdrant, …) must implement this.
- */
-interface VectorMatch {
-    /** Unique identifier of the stored chunk */
-    id: string;
-    /** Cosine or dot-product similarity score (0–1) */
-    score: number;
-    /** The original text content of the chunk */
-    content: string;
-    /** Arbitrary metadata attached at upsert time */
-    metadata?: Record<string, unknown>;
-}
-interface UpsertDocument {
-    id: string;
-    vector: number[];
-    content: string;
-    metadata?: Record<string, unknown>;
-}
-interface IVectorDB {
-    /**
-     * Initialise the connection (create tables, verify index, etc.)
-     * Called once during RAGPipeline construction.
-     */
-    initialize(): Promise<void>;
-    /**
-     * Insert or update a single vector + content pair.
-     * @param namespace – optional project/tenant namespace
-     */
-    upsert(doc: UpsertDocument, namespace?: string): Promise<void>;
-    /**
-     * Batch upsert for efficient ingestion of many documents.
-     */
-    batchUpsert(docs: UpsertDocument[], namespace?: string): Promise<void>;
-    /**
-     * Find the top-K most similar vectors to the query vector.
-     * @param vector     – embedding of the user query
-     * @param topK       – number of results to return
-     * @param namespace  – optional project/tenant namespace
-     * @param filter     – optional metadata filter (provider-specific shape)
-     */
-    query(vector: number[], topK: number, namespace?: string, filter?: Record<string, unknown>): Promise<VectorMatch[]>;
-    /**
-     * Delete a stored vector by ID.
-     */
-    delete(id: string, namespace?: string): Promise<void>;
-    /**
-     * Delete all vectors in a namespace (full data reset for a project).
-     */
-    deleteNamespace(namespace: string): Promise<void>;
-    /**
-     * Check if the underlying DB is reachable.
-     */
-    ping(): Promise<boolean>;
-    /**
-     * Gracefully close the connection.
-     */
-    disconnect(): Promise<void>;
-}
-
 /**
  * Master configuration interface for Retrivora AI.
  * Each consuming project provides one RagConfig object to drive
  * vector DB selection, LLM selection, embedding, and UI branding.
  */
-type VectorDBProvider = 'pinecone' | 'pgvector' | 'mongodb' | 'rest' | 'universal_rest' | 'custom';
+type VectorDBProvider = 'pinecone' | 'pgvector' | 'postgresql' | 'mongodb' | 'milvus' | 'qdrant' | 'chromadb' | 'redis' | 'weaviate' | 'rest' | 'universal_rest' | 'custom';
 interface VectorDBConfig {
     /** Which vector database to use */
     provider: VectorDBProvider;
@@ -232,67 +121,95 @@ interface RagConfig {
 }
 
 /**
- * RAGPipeline — the central orchestrator for the RAG workflow.
- *
- * Responsibilities:
- *  1. INGEST:  chunk → embed → upsert into vector DB
- *  2. ASK:     embed query → retrieve top-K chunks → call LLM with context
- *
- * It is LLM-agnostic and vector-DB-agnostic; concrete implementations are
- * injected via ILLMProvider and IVectorDB.
+ * Generic LLM Provider interface.
+ * Covers both chat completion and embedding generation so a single
+ * provider can handle both responsibilities when appropriate.
  */
+interface ChatMessage {
+    role: 'user' | 'assistant' | 'system';
+    content: string;
+}
+interface ChatOptions {
+    /** Override max tokens for this specific call */
+    maxTokens?: number;
+    /** Override temperature for this specific call */
+    temperature?: number;
+    /** Stop sequences */
+    stop?: string[];
+}
+interface EmbedOptions {
+    /** Override model for this specific embed call */
+    model?: string;
+}
+interface ILLMProvider {
+    /**
+     * Send a chat completion request.
+     * @param messages – the full conversation history
+     * @param context  – retrieved RAG context to inject
+     * @param options  – optional per-call overrides
+     * @returns        – the assistant's reply text
+     */
+    chat(messages: ChatMessage[], context: string, options?: ChatOptions): Promise<string>;
+    /**
+     * Generate an embedding vector for the given text.
+     * @param text    – text to embed
+     * @param options – optional overrides
+     * @returns       – float array (the embedding)
+     */
+    embed(text: string, options?: EmbedOptions): Promise<number[]>;
+    /**
+     * Generate embedding vectors for multiple texts in a single batch.
+     * @param texts   – array of texts to embed
+     * @param options – optional overrides
+     * @returns       – array of float arrays
+     */
+    batchEmbed(texts: string[], options?: EmbedOptions): Promise<number[][]>;
+    /**
+     * Check if the provider endpoint is reachable.
+     */
+    ping(): Promise<boolean>;
+}
+
+interface VectorMatch {
+    id: string;
+    score: number;
+    content: string;
+    metadata?: Record<string, unknown>;
+}
+interface UpsertDocument {
+    id: string;
+    vector: number[];
+    content: string;
+    metadata?: Record<string, unknown>;
+}
 
 interface IngestDocument {
-    /** Unique document ID (used as chunk ID prefix) */
     docId: string;
-    /** Full text content of the document */
     content: string;
-    /** Optional metadata stored alongside each chunk */
     metadata?: Record<string, unknown>;
 }
 interface ChatResponse {
-    /** The LLM's answer */
     reply: string;
-    /** Retrieved source chunks used to generate the answer */
     sources: VectorMatch[];
 }
-declare class RAGPipeline {
-    private readonly vectorDB;
-    private readonly llmProvider;
-    private readonly embeddingProvider;
-    private readonly chunker;
-    private readonly config;
+/**
+ * Pipeline — orchestrates the RAG flow: Embed → Search → Augment → Generate.
+ */
+declare class Pipeline {
+    private vectorDB;
+    private llmProvider;
+    private embeddingProvider;
+    private chunker;
+    private config;
     private initialised;
-    constructor(config: RagConfig, adapters?: {
-        vectorDB?: IVectorDB;
-        llmProvider?: ILLMProvider;
-        embeddingProvider?: ILLMProvider;
-    });
-    /** Lazily initialise the vector DB connection */
-    init(): Promise<void>;
-    /**
-     * Ingest one or more documents into the vector DB.
-     * Automatically chunks, embeds, and upserts each chunk.
-     */
-    ingest(documents: IngestDocument[], namespace?: string): Promise<{
+    constructor(config: RagConfig);
+    initialize(): Promise<void>;
+    ingest(documents: IngestDocument[], namespace?: string): Promise<Array<{
         docId: string;
         chunksIngested: number;
-    }[]>;
-    /**
-     * Run a full RAG query:
-     *  1. Embed the user question
-     *  2. Retrieve top-K relevant chunks from the vector DB
-     *  3. Filter by score threshold
-     *  4. Build context string
-     *  5. Call the LLM with chat history + context
-     */
+    }>>;
     ask(question: string, history?: ChatMessage[], namespace?: string): Promise<ChatResponse>;
-    health(): Promise<{
-        vectorDB: boolean;
-        llm: boolean;
-    }>;
-    deleteDocument(docId: string, namespace?: string): Promise<void>;
-    deleteProject(namespace?: string): Promise<void>;
+    health(): Promise<Record<string, boolean>>;
 }
 
-export { type ChatMessage as C, type EmbedOptions as E, type ILLMProvider as I, type LLMConfig as L, type RAGConfig as R, type UIConfig as U, type VectorMatch as V, type ChatOptions as a, type ChatResponse as b, type EmbeddingConfig as c, type IVectorDB as d, type IngestDocument as e, type RagConfig as f, type UpsertDocument as g, type VectorDBConfig as h, RAGPipeline as i };
+export { type ChatMessage as C, type EmbedOptions as E, type ILLMProvider as I, type LLMConfig as L, Pipeline as P, type RAGConfig as R, type UIConfig as U, type VectorMatch as V, type ChatOptions as a, type ChatResponse as b, type EmbeddingConfig as c, type IngestDocument as d, type RagConfig as e, type UpsertDocument as f, type VectorDBConfig as g };