@retrivora-ai/rag-engine 0.1.3 → 0.1.5

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-cfaMidtA.d.mts

@@ -0,0 +1,93 @@
+/**
+ * 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>;
+}
+
+/**
+ * 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 { type ChatMessage as C, DocumentChunker as D, type EmbedOptions as E, type ILLMProvider as I, type ChatOptions as a, type Chunk as b, type ChunkOptions as c };

package/dist/DocumentChunker-cfaMidtA.d.ts

@@ -0,0 +1,93 @@
+/**
+ * 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>;
+}
+
+/**
+ * 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 { type ChatMessage as C, DocumentChunker as D, type EmbedOptions as E, type ILLMProvider as I, type ChatOptions as a, type Chunk as b, type ChunkOptions as c };

package/dist/LLMFactory-JFOY2V4X.mjs

@@ -0,0 +1,8 @@
+import {
+  LLMFactory
+} from "./chunk-JI6VD5TJ.mjs";
+import "./chunk-UKDXCXW7.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/PostgreSQLProvider-ISNMD3BE.mjs

@@ -0,0 +1,8 @@
+import {
+  PostgreSQLProvider
+} from "./chunk-S5DRHETN.mjs";
+import "./chunk-VOIWNO5O.mjs";
+import "./chunk-I4E63NIC.mjs";
+export {
+  PostgreSQLProvider
+};

package/dist/QdrantProvider-LJWOIGES.mjs

@@ -0,0 +1,8 @@
+import {
+  QdrantProvider
+} from "./chunk-6FODXNUF.mjs";
+import "./chunk-VOIWNO5O.mjs";
+import "./chunk-I4E63NIC.mjs";
+export {
+  QdrantProvider
+};

package/dist/RagConfig-DG_0f8ka.d.mts

@@ -0,0 +1,145 @@
+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 {
+    docId: string;
+    content: string;
+    metadata?: Record<string, unknown>;
+}
+interface ChatResponse {
+    reply: string;
+    sources: VectorMatch[];
+}
+
+/**
+ * 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' | 'postgresql' | 'mongodb' | 'milvus' | 'qdrant' | 'chromadb' | 'redis' | 'weaviate' | 'rest' | 'universal_rest' | 'custom';
+interface VectorDBConfig {
+    /** Which vector database to use */
+    provider: VectorDBProvider;
+    /** The index / table name to query */
+    indexName: string;
+    /** Provider-specific options (API keys, connection strings, etc.)
+     *
+     * For 'universal_rest', the following options are supported:
+     * - baseUrl: string
+     * - headers?: Record<string, string>
+     * - queryPath?: string
+     * - upsertPath?: string
+     * - queryPayloadTemplate?: string (e.g. '{"vector": {{vector}}, "limit": {{topK}}}')
+     * - upsertPayloadTemplate?: string (e.g. '{"id": "{{id}}", "vector": {{vector}}}')
+     * - responseExtractPath?: string (e.g. 'data.results')
+     * - idPath?: string (e.g. '_id')
+     * - scorePath?: string (e.g. 'similarity')
+     * - contentPath?: string (e.g. 'text')
+     */
+    options: Record<string, unknown>;
+}
+type LLMProvider = 'openai' | 'anthropic' | 'ollama' | 'rest' | 'universal_rest' | 'custom';
+interface LLMConfig {
+    /** Which LLM provider to use */
+    provider: LLMProvider;
+    /** Model name, e.g. "gpt-4o", "claude-3-opus-20240229", "llama3" */
+    model: string;
+    /** API key for cloud providers */
+    apiKey?: string;
+    /** Base URL — required for Ollama or self-hosted endpoints */
+    baseUrl?: string;
+    /** Custom system prompt injected before every chat */
+    systemPrompt?: string;
+    /** Max tokens in the LLM response */
+    maxTokens?: number;
+    /** Sampling temperature (0–1) */
+    temperature?: number;
+    /** Provider-specific options
+     *
+     * For 'universal_rest', the following options are supported:
+     * - headers?: Record<string, string>
+     * - chatPath?: string
+     * - chatPayloadTemplate?: string (e.g. '{"prompt": "{{prompt}}", "max_tokens": {{maxTokens}}}')
+     * - responseExtractPath?: string (e.g. 'choices[0].text' or 'response')
+     */
+    options?: Record<string, unknown>;
+}
+type EmbeddingProvider = 'openai' | 'ollama' | 'rest' | 'universal_rest' | 'custom';
+interface EmbeddingConfig {
+    /** Which embedding provider to use */
+    provider: EmbeddingProvider;
+    /** Model name, e.g. "text-embedding-ada-002", "nomic-embed-text" */
+    model: string;
+    /** API key (if needed) */
+    apiKey?: string;
+    /** Base URL for custom / Ollama embedding endpoints */
+    baseUrl?: string;
+    /** Output vector dimension — must match the index dimension */
+    dimensions?: number;
+    /** Provider-specific options for custom adapters */
+    options?: Record<string, unknown>;
+}
+interface UIConfig {
+    /** Title shown in the chat header */
+    title?: string;
+    /** Subtitle / description below the title */
+    subtitle?: string;
+    /** Primary brand colour (CSS colour string) */
+    primaryColor?: string;
+    /** Accent colour for user message bubbles */
+    accentColor?: string;
+    /** URL to a logo image */
+    logoUrl?: string;
+    /** Input placeholder text */
+    placeholder?: string;
+    /** Show source cards after each answer */
+    showSources?: boolean;
+    /** Welcome message shown on first load */
+    welcomeMessage?: string;
+    /** Whether to show the floating chat widget. Defaults to true. */
+    showWidget?: boolean;
+    /** Custom 'Powered by' text shown in the header */
+    poweredBy?: string;
+    /** Visual style: 'glass' (default) or 'solid' */
+    visualStyle?: 'glass' | 'solid';
+    /** Border radius: 'none', 'sm', 'md', 'lg', 'xl', 'full' */
+    borderRadius?: 'none' | 'sm' | 'md' | 'lg' | 'xl' | 'full';
+}
+interface RAGConfig {
+    /** Number of top-K chunks retrieved per query */
+    topK?: number;
+    /** Minimum similarity score to include a chunk (0–1) */
+    scoreThreshold?: number;
+    /** Target chunk size in tokens */
+    chunkSize?: number;
+    /** Overlap between adjacent chunks in tokens */
+    chunkOverlap?: number;
+}
+interface RagConfig {
+    /**
+     * Unique identifier for the consuming project.
+     * Used as the vector DB namespace to achieve multi-tenancy.
+     */
+    projectId: string;
+    /** Vector database configuration */
+    vectorDb: VectorDBConfig;
+    /** LLM configuration */
+    llm: LLMConfig;
+    /** Embedding configuration */
+    embedding: EmbeddingConfig;
+    /** Optional UI branding overrides */
+    ui?: UIConfig;
+    /** Optional RAG pipeline tuning knobs */
+    rag?: RAGConfig;
+}
+
+export type { ChatResponse as C, EmbeddingConfig as E, IngestDocument as I, LLMConfig as L, RAGConfig as R, UIConfig as U, VectorMatch as V, RagConfig as a, UpsertDocument as b, VectorDBConfig as c };

package/dist/RagConfig-DG_0f8ka.d.ts

@@ -0,0 +1,145 @@
+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 {
+    docId: string;
+    content: string;
+    metadata?: Record<string, unknown>;
+}
+interface ChatResponse {
+    reply: string;
+    sources: VectorMatch[];
+}
+
+/**
+ * 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' | 'postgresql' | 'mongodb' | 'milvus' | 'qdrant' | 'chromadb' | 'redis' | 'weaviate' | 'rest' | 'universal_rest' | 'custom';
+interface VectorDBConfig {
+    /** Which vector database to use */
+    provider: VectorDBProvider;
+    /** The index / table name to query */
+    indexName: string;
+    /** Provider-specific options (API keys, connection strings, etc.)
+     *
+     * For 'universal_rest', the following options are supported:
+     * - baseUrl: string
+     * - headers?: Record<string, string>
+     * - queryPath?: string
+     * - upsertPath?: string
+     * - queryPayloadTemplate?: string (e.g. '{"vector": {{vector}}, "limit": {{topK}}}')
+     * - upsertPayloadTemplate?: string (e.g. '{"id": "{{id}}", "vector": {{vector}}}')
+     * - responseExtractPath?: string (e.g. 'data.results')
+     * - idPath?: string (e.g. '_id')
+     * - scorePath?: string (e.g. 'similarity')
+     * - contentPath?: string (e.g. 'text')
+     */
+    options: Record<string, unknown>;
+}
+type LLMProvider = 'openai' | 'anthropic' | 'ollama' | 'rest' | 'universal_rest' | 'custom';
+interface LLMConfig {
+    /** Which LLM provider to use */
+    provider: LLMProvider;
+    /** Model name, e.g. "gpt-4o", "claude-3-opus-20240229", "llama3" */
+    model: string;
+    /** API key for cloud providers */
+    apiKey?: string;
+    /** Base URL — required for Ollama or self-hosted endpoints */
+    baseUrl?: string;
+    /** Custom system prompt injected before every chat */
+    systemPrompt?: string;
+    /** Max tokens in the LLM response */
+    maxTokens?: number;
+    /** Sampling temperature (0–1) */
+    temperature?: number;
+    /** Provider-specific options
+     *
+     * For 'universal_rest', the following options are supported:
+     * - headers?: Record<string, string>
+     * - chatPath?: string
+     * - chatPayloadTemplate?: string (e.g. '{"prompt": "{{prompt}}", "max_tokens": {{maxTokens}}}')
+     * - responseExtractPath?: string (e.g. 'choices[0].text' or 'response')
+     */
+    options?: Record<string, unknown>;
+}
+type EmbeddingProvider = 'openai' | 'ollama' | 'rest' | 'universal_rest' | 'custom';
+interface EmbeddingConfig {
+    /** Which embedding provider to use */
+    provider: EmbeddingProvider;
+    /** Model name, e.g. "text-embedding-ada-002", "nomic-embed-text" */
+    model: string;
+    /** API key (if needed) */
+    apiKey?: string;
+    /** Base URL for custom / Ollama embedding endpoints */
+    baseUrl?: string;
+    /** Output vector dimension — must match the index dimension */
+    dimensions?: number;
+    /** Provider-specific options for custom adapters */
+    options?: Record<string, unknown>;
+}
+interface UIConfig {
+    /** Title shown in the chat header */
+    title?: string;
+    /** Subtitle / description below the title */
+    subtitle?: string;
+    /** Primary brand colour (CSS colour string) */
+    primaryColor?: string;
+    /** Accent colour for user message bubbles */
+    accentColor?: string;
+    /** URL to a logo image */
+    logoUrl?: string;
+    /** Input placeholder text */
+    placeholder?: string;
+    /** Show source cards after each answer */
+    showSources?: boolean;
+    /** Welcome message shown on first load */
+    welcomeMessage?: string;
+    /** Whether to show the floating chat widget. Defaults to true. */
+    showWidget?: boolean;
+    /** Custom 'Powered by' text shown in the header */
+    poweredBy?: string;
+    /** Visual style: 'glass' (default) or 'solid' */
+    visualStyle?: 'glass' | 'solid';
+    /** Border radius: 'none', 'sm', 'md', 'lg', 'xl', 'full' */
+    borderRadius?: 'none' | 'sm' | 'md' | 'lg' | 'xl' | 'full';
+}
+interface RAGConfig {
+    /** Number of top-K chunks retrieved per query */
+    topK?: number;
+    /** Minimum similarity score to include a chunk (0–1) */
+    scoreThreshold?: number;
+    /** Target chunk size in tokens */
+    chunkSize?: number;
+    /** Overlap between adjacent chunks in tokens */
+    chunkOverlap?: number;
+}
+interface RagConfig {
+    /**
+     * Unique identifier for the consuming project.
+     * Used as the vector DB namespace to achieve multi-tenancy.
+     */
+    projectId: string;
+    /** Vector database configuration */
+    vectorDb: VectorDBConfig;
+    /** LLM configuration */
+    llm: LLMConfig;
+    /** Embedding configuration */
+    embedding: EmbeddingConfig;
+    /** Optional UI branding overrides */
+    ui?: UIConfig;
+    /** Optional RAG pipeline tuning knobs */
+    rag?: RAGConfig;
+}
+
+export type { ChatResponse as C, EmbeddingConfig as E, IngestDocument as I, LLMConfig as L, RAGConfig as R, UIConfig as U, VectorMatch as V, RagConfig as a, UpsertDocument as b, VectorDBConfig as c };

package/dist/RedisProvider-ASONNYBI.mjs

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

package/dist/WeaviateProvider-PSDCUGC7.mjs

@@ -0,0 +1,8 @@
+import {
+  WeaviateProvider
+} from "./chunk-V75V7BT2.mjs";
+import "./chunk-VOIWNO5O.mjs";
+import "./chunk-I4E63NIC.mjs";
+export {
+  WeaviateProvider
+};

package/dist/chunk-6FODXNUF.mjs

@@ -0,0 +1,91 @@
+import {
+  BaseVectorProvider
+} from "./chunk-VOIWNO5O.mjs";
+import {
+  __spreadValues
+} from "./chunk-I4E63NIC.mjs";
+
+// src/providers/vectordb/QdrantProvider.ts
+import axios from "axios";
+var QdrantProvider = class extends BaseVectorProvider {
+  constructor(config) {
+    super(config);
+    const opts = config.options;
+    const baseUrl = opts.baseUrl;
+    if (!baseUrl) throw new Error("[QdrantProvider] baseUrl is required");
+    this.http = axios.create({
+      baseURL: baseUrl,
+      headers: __spreadValues({
+        "Content-Type": "application/json"
+      }, opts.apiKey ? { "api-key": opts.apiKey } : {})
+    });
+  }
+  async initialize() {
+    await this.ping();
+  }
+  async upsert(doc, namespace) {
+    await this.batchUpsert([doc], namespace);
+  }
+  async batchUpsert(docs, namespace) {
+    const payload = {
+      points: docs.map((doc) => ({
+        id: doc.id,
+        vector: doc.vector,
+        payload: __spreadValues({
+          content: doc.content,
+          metadata: doc.metadata || {}
+        }, namespace ? { namespace } : {})
+      }))
+    };
+    await this.http.put(`/collections/${this.indexName}/points`, payload);
+  }
+  // eslint-disable-next-line @typescript-eslint/no-unused-vars
+  async query(vector, topK, namespace, _filter) {
+    const payload = {
+      vector,
+      limit: topK,
+      with_payload: true,
+      filter: namespace ? {
+        must: [{ key: "namespace", match: { value: namespace } }]
+      } : void 0
+    };
+    const { data } = await this.http.post(`/collections/${this.indexName}/points/search`, payload);
+    return (data.result || []).map((res) => {
+      var _a, _b;
+      return {
+        id: res["id"],
+        score: res["score"],
+        content: ((_a = res["payload"]) == null ? void 0 : _a.content) || "",
+        metadata: ((_b = res["payload"]) == null ? void 0 : _b.metadata) || {}
+      };
+    });
+  }
+  // eslint-disable-next-line @typescript-eslint/no-unused-vars
+  async delete(id, _namespace) {
+    await this.http.post(`/collections/${this.indexName}/points/delete`, {
+      points: [id]
+    });
+  }
+  // eslint-disable-next-line @typescript-eslint/no-unused-vars
+  async deleteNamespace(_namespace) {
+    await this.http.post(`/collections/${this.indexName}/points/delete`, {
+      filter: {
+        must: [{ key: "namespace", match: { value: _namespace } }]
+      }
+    });
+  }
+  async ping() {
+    try {
+      await this.http.get("/healthz");
+      return true;
+    } catch (e) {
+      return false;
+    }
+  }
+  async disconnect() {
+  }
+};
+
+export {
+  QdrantProvider
+};