@retrivora-ai/rag-engine 1.0.4 → 1.0.5

package/dist/DocumentChunker-Dh9TvmGG.d.mts

@@ -0,0 +1,45 @@
+/**
+ * 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 | number;
+    /** Extra metadata to attach to every chunk */
+    metadata?: Record<string, unknown>;
+    /** Characters used to split text, in order of priority */
+    separators?: string[];
+}
+declare class DocumentChunker {
+    private readonly chunkSize;
+    private readonly chunkOverlap;
+    private readonly separators;
+    constructor(chunkSize?: number, chunkOverlap?: number, separators?: string[]);
+    /**
+     * Split a single text string into overlapping chunks using a recursive strategy.
+     * Preserves structural boundaries (Markdown headers) where possible.
+     */
+    chunk(text: string, options?: ChunkOptions): Chunk[];
+    private recursiveSplit;
+    chunkMany(documents: Array<{
+        content: string;
+        docId?: string | number;
+        metadata?: Record<string, unknown>;
+    }>): Chunk[];
+}
+
+export { type Chunk as C, DocumentChunker as D, type ChunkOptions as a };

package/dist/DocumentChunker-Dh9TvmGG.d.ts

@@ -0,0 +1,45 @@
+/**
+ * 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 | number;
+    /** Extra metadata to attach to every chunk */
+    metadata?: Record<string, unknown>;
+    /** Characters used to split text, in order of priority */
+    separators?: string[];
+}
+declare class DocumentChunker {
+    private readonly chunkSize;
+    private readonly chunkOverlap;
+    private readonly separators;
+    constructor(chunkSize?: number, chunkOverlap?: number, separators?: string[]);
+    /**
+     * Split a single text string into overlapping chunks using a recursive strategy.
+     * Preserves structural boundaries (Markdown headers) where possible.
+     */
+    chunk(text: string, options?: ChunkOptions): Chunk[];
+    private recursiveSplit;
+    chunkMany(documents: Array<{
+        content: string;
+        docId?: string | number;
+        metadata?: Record<string, unknown>;
+    }>): Chunk[];
+}
+
+export { type Chunk as C, DocumentChunker as D, type ChunkOptions as a };

package/dist/{RagConfig-DRJO4hGU.d.mts → RagConfig-BEmz4hVP.d.mts}

@@ -1,3 +1,60 @@
+/**
+ * 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;
+    /** Specify the task type for models that require prefixes (e.g. nomic-embed-text) */
+    taskType?: 'query' | 'document';
+}
+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>;
+    /**
+     * Send a streaming chat completion request.
+     * @returns – an async iterable of text chunks
+     */
+    chatStream?(messages: ChatMessage[], context: string, options?: ChatOptions): AsyncIterable<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 | number;
     score: number;
@@ -208,4 +265,4 @@ interface RagConfig {
     graphDb?: GraphDBConfig;
 }
 
-export type { ChatResponse as C, EmbeddingConfig as E, GraphDBConfig as G, IngestDocument as I, LLMConfig as L, RAGConfig as R, UIConfig as U, VectorMatch as V, EmbeddingProvider as a, LLMProvider as b, RagConfig as c, UpsertDocument as d, VectorDBConfig as e, VectorDBProvider as f, RetrievalResult as g, GraphNode as h, Edge as i, GraphSearchResult as j };
+export type { ChatMessage as C, EmbedOptions as E, GraphDBConfig as G, ILLMProvider as I, LLMConfig as L, RAGConfig as R, UIConfig as U, VectorMatch as V, ChatOptions as a, ChatResponse as b, EmbeddingConfig as c, EmbeddingProvider as d, IngestDocument as e, LLMProvider as f, RagConfig as g, UpsertDocument as h, VectorDBConfig as i, VectorDBProvider as j, RetrievalResult as k, GraphNode as l, Edge as m, GraphSearchResult as n };

package/dist/{RagConfig-DRJO4hGU.d.ts → RagConfig-BEmz4hVP.d.ts}

@@ -1,3 +1,60 @@
+/**
+ * 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;
+    /** Specify the task type for models that require prefixes (e.g. nomic-embed-text) */
+    taskType?: 'query' | 'document';
+}
+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>;
+    /**
+     * Send a streaming chat completion request.
+     * @returns – an async iterable of text chunks
+     */
+    chatStream?(messages: ChatMessage[], context: string, options?: ChatOptions): AsyncIterable<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 | number;
     score: number;
@@ -208,4 +265,4 @@ interface RagConfig {
     graphDb?: GraphDBConfig;
 }
 
-export type { ChatResponse as C, EmbeddingConfig as E, GraphDBConfig as G, IngestDocument as I, LLMConfig as L, RAGConfig as R, UIConfig as U, VectorMatch as V, EmbeddingProvider as a, LLMProvider as b, RagConfig as c, UpsertDocument as d, VectorDBConfig as e, VectorDBProvider as f, RetrievalResult as g, GraphNode as h, Edge as i, GraphSearchResult as j };
+export type { ChatMessage as C, EmbedOptions as E, GraphDBConfig as G, ILLMProvider as I, LLMConfig as L, RAGConfig as R, UIConfig as U, VectorMatch as V, ChatOptions as a, ChatResponse as b, EmbeddingConfig as c, EmbeddingProvider as d, IngestDocument as e, LLMProvider as f, RagConfig as g, UpsertDocument as h, VectorDBConfig as i, VectorDBProvider as j, RetrievalResult as k, GraphNode as l, Edge as m, GraphSearchResult as n };

package/dist/{chunk-6MLZHQZT.mjs → chunk-MWL4AGQI.mjs}

@@ -190,13 +190,23 @@ function getRagConfig(env = process.env) {
 }
 
 // src/core/ConfigResolver.ts
+var _cachedEnvConfig = null;
+function getCachedEnvConfig() {
+  if (!_cachedEnvConfig) {
+    _cachedEnvConfig = getRagConfig();
+  }
+  return _cachedEnvConfig;
+}
+function resetConfigCache() {
+  _cachedEnvConfig = null;
+}
 var ConfigResolver = class {
   /**
-   * Resolves the final configuration by merging host-provided config with defaults.
+   * Resolves the final configuration by merging host-provided config with environment defaults.
    * @param hostConfig - Partial configuration passed from the host application.
    */
   static resolve(hostConfig) {
-    const envConfig = getRagConfig();
+    const envConfig = getCachedEnvConfig();
     if (!hostConfig) {
       return envConfig;
     }
@@ -1508,14 +1518,16 @@ var LangChainAgent = class {
     this.config = config;
   }
   /**
-   * Initializes the agent with the RAG pipeline as a primary tool.
-   * Dynamically imports LangChain dependencies to avoid build errors if missing.
+   * Initializes the LangChain ReAct agent with the RAG pipeline as a tool.
+   * Dynamically imports langchain so the package doesn't crash if it isn't installed.
    */
   async initialize(chatModel) {
     try {
-      const { DynamicTool } = await import(`${"@langchain/core/tools"}`);
-      const { ChatPromptTemplate, MessagesPlaceholder } = await import(`${"@langchain/core/prompts"}`);
-      const { AgentExecutor, createOpenAIFunctionsAgent } = await import(`${"langchain/agents"}`);
+      const [{ DynamicTool }, { HumanMessage }, { createAgent }] = await Promise.all([
+        import("@langchain/core/tools"),
+        import("@langchain/core/messages"),
+        import("langchain")
+      ]);
       const searchTool = new DynamicTool({
         name: "document_search",
         description: "Use this tool to search through the knowledge base and document repository. Input should be a specific search query.",
@@ -1524,42 +1536,53 @@ var LangChainAgent = class {
           return `Search Results:
 ${response.reply}
 
-Sources Used: ${JSON.stringify(response.sources.map((s) => s.id))}`;
+Sources Used: ${JSON.stringify(
+            response.sources.map((s) => s.id)
+          )}`;
         }
       });
-      const tools = [searchTool];
-      const prompt = ChatPromptTemplate.fromMessages([
-        ["system", this.config.llm.systemPrompt || "You are a helpful AI assistant with access to a document search tool."],
-        new MessagesPlaceholder("chat_history"),
-        ["human", "{input}"],
-        new MessagesPlaceholder("agent_scratchpad")
-      ]);
-      const agent = await createOpenAIFunctionsAgent({
-        llm: chatModel,
-        tools,
-        prompt
-      });
-      this.executor = new AgentExecutor({
-        agent,
-        tools
+      this.agent = createAgent({
+        // eslint-disable-next-line @typescript-eslint/no-explicit-any
+        model: chatModel,
+        tools: [searchTool],
+        systemPrompt: this.config.llm.systemPrompt || "You are a helpful AI assistant with access to a document search tool."
       });
+      void HumanMessage;
     } catch (error) {
-      console.error("[LangChainAgent] Failed to initialize. Ensure 'langchain' and '@langchain/core' are installed.");
-      throw error;
+      const isMissing = error instanceof Error && error.message.includes("Cannot find module");
+      const hint = isMissing ? " Make sure 'langchain' and '@langchain/core' are installed:\n  npm install langchain @langchain/core" : "";
+      throw new Error(
+        `[LangChainAgent] Failed to initialize.${hint}
+${error instanceof Error ? error.message : String(error)}`
+      );
     }
   }
   /**
    * Run the agentic flow.
+   *
+   * LangChain v1.x: invoke takes `{ messages: [...] }` instead of `{ input, chat_history }`.
+   * The agent returns `{ messages: [...] }` — the last message is the final answer.
    */
   async run(input, chatHistory = []) {
-    if (!this.executor) {
+    var _a, _b, _c;
+    if (!this.agent) {
       throw new Error("[LangChainAgent] Agent not initialized. Call initialize() first.");
     }
-    const response = await this.executor.invoke({
-      input,
-      chat_history: chatHistory
+    const { HumanMessage, AIMessage } = await import("@langchain/core/messages");
+    const historyMessages = chatHistory.map(
+      (m) => m.role === "user" ? new HumanMessage(m.content) : new AIMessage(m.content)
+    );
+    const response = await this.agent.invoke({
+      messages: [...historyMessages, new HumanMessage(input)]
     });
-    return response.output;
+    const lastMessage = (_a = response == null ? void 0 : response.messages) == null ? void 0 : _a.at(-1);
+    if (lastMessage && typeof lastMessage.content === "string") {
+      return lastMessage.content;
+    }
+    if (Array.isArray(lastMessage == null ? void 0 : lastMessage.content)) {
+      return lastMessage.content.filter((c) => c.type === "text").map((c) => c.text).join("");
+    }
+    return String((_c = (_b = response == null ? void 0 : response.output) != null ? _b : response) != null ? _c : "");
   }
 };
 
@@ -1968,6 +1991,10 @@ var QueryProcessor = class {
   }
   /**
    * Constructs a QueryFilter object from extracted hints.
+   *
+   * Note: proper nouns are extracted inside `extractQueryFieldHints` and surfaced
+   * as field-less hints; `buildQueryFilter` adds them to `keywords` without
+   * re-running the proper noun regex to avoid duplication.
    */
   static buildQueryFilter(question, hints) {
     const filter = { metadata: {}, keywords: [], queryText: question };
@@ -1975,13 +2002,11 @@ var QueryProcessor = class {
       if (hint.field) {
         filter.metadata[hint.field] = hint.value;
       } else {
-        filter.keywords.push(hint.value);
+        if (!filter.keywords.includes(hint.value)) {
+          filter.keywords.push(hint.value);
+        }
       }
     }
-    for (const match of question.matchAll(/\b[A-Z][a-z]+(?:\s+[A-Z][a-z]+){0,3}\b/g)) {
-      const term = this.normalizeHintValue(match[0]);
-      if (term && !filter.keywords.includes(term)) filter.keywords.push(term);
-    }
     if (Object.keys(filter.metadata || {}).length === 0) delete filter.metadata;
     if (filter.keywords && filter.keywords.length === 0) delete filter.keywords;
     return filter;
@@ -1989,10 +2014,40 @@ var QueryProcessor = class {
 };
 
 // src/core/Pipeline.ts
+var LRUEmbeddingCache = class {
+  constructor(maxSize = 500) {
+    this.cache = /* @__PURE__ */ new Map();
+    this.maxSize = maxSize;
+  }
+  get(key) {
+    const value = this.cache.get(key);
+    if (value !== void 0) {
+      this.cache.delete(key);
+      this.cache.set(key, value);
+    }
+    return value;
+  }
+  set(key, value) {
+    if (this.cache.has(key)) {
+      this.cache.delete(key);
+    } else if (this.cache.size >= this.maxSize) {
+      const oldestKey = this.cache.keys().next().value;
+      if (oldestKey !== void 0) this.cache.delete(oldestKey);
+    }
+    this.cache.set(key, value);
+  }
+  clear() {
+    this.cache.clear();
+  }
+  get size() {
+    return this.cache.size;
+  }
+};
 var Pipeline = class {
   constructor(config) {
     this.config = config;
-    this.embeddingCache = /* @__PURE__ */ new Map();
+    /** LRU-bounded cache: avoids re-embedding identical queries within the same process. */
+    this.embeddingCache = new LRUEmbeddingCache(500);
     this.initialised = false;
     var _a, _b, _c, _d, _e;
     this.chunker = new DocumentChunker(
@@ -2080,6 +2135,7 @@ var Pipeline = class {
   }
   /**
    * Step 2: Generate embeddings for chunks with retry logic.
+   * Uses batchEmbed when available for efficiency; falls back to sequential embedding.
    */
   async processEmbeddings(chunks) {
     const embedBatchOptions = {
@@ -2093,7 +2149,9 @@ var Pipeline = class {
       embedBatchOptions
     );
     if (vectors.length !== chunks.length) {
-      throw new Error(`Embedding mismatch: got ${vectors.length} vectors for ${chunks.length} chunks`);
+      throw new Error(
+        `[Pipeline] Embedding mismatch: got ${vectors.length} vectors for ${chunks.length} chunks. Check embedding provider logs for individual chunk failures.`
+      );
     }
     return vectors;
   }
@@ -2247,6 +2305,7 @@ ${context}`;
   }
   /**
    * Universal retrieval method combining all enabled providers.
+   * Uses an LRU-bounded embedding cache to avoid re-embedding the same query.
    */
   async retrieve(query, options) {
     var _a, _b, _c;
@@ -2279,10 +2338,13 @@ ${history.map((m) => `${m.role}: ${m.content}`).join("\n")}
 New Question: ${question}
 
 Optimized Search Query:`;
-    const rewrite = await this.llmProvider.chat([
-      { role: "system", content: "You are an assistant that optimizes search queries for RAG systems." },
-      { role: "user", content: prompt }
-    ], "");
+    const rewrite = await this.llmProvider.chat(
+      [
+        { role: "system", content: "You are an assistant that optimizes search queries for RAG systems." },
+        { role: "user", content: prompt }
+      ],
+      ""
+    );
     return rewrite.trim() || question;
   }
 };
@@ -2501,8 +2563,35 @@ var DocumentParser = class {
 };
 
 // src/handlers/index.ts
-function createChatHandler(config) {
-  const plugin = new VectorPlugin(config);
+function sseFrame(payload) {
+  return `data: ${JSON.stringify(payload)}
+
+`;
+}
+function sseTextFrame(text) {
+  return `data: ${JSON.stringify({ type: "text", text })}
+
+`;
+}
+function sseMetaFrame(meta) {
+  return `data: ${JSON.stringify(__spreadValues({ type: "metadata" }, meta != null ? meta : {}))}
+
+`;
+}
+function sseErrorFrame(message) {
+  return `data: ${JSON.stringify({ type: "error", error: message })}
+
+`;
+}
+var SSE_HEADERS = {
+  "Content-Type": "text/event-stream; charset=utf-8",
+  "Cache-Control": "no-cache, no-transform",
+  Connection: "keep-alive",
+  "X-Accel-Buffering": "no"
+  // Disable Nginx buffering for streaming
+};
+function createChatHandler(configOrPlugin) {
+  const plugin = configOrPlugin instanceof VectorPlugin ? configOrPlugin : new VectorPlugin(configOrPlugin);
   return async function POST(req) {
     try {
       const body = await req.json();
@@ -2518,67 +2607,64 @@ function createChatHandler(config) {
     }
   };
 }
-function createStreamHandler(config) {
-  const plugin = new VectorPlugin(config);
+function createStreamHandler(configOrPlugin) {
+  const plugin = configOrPlugin instanceof VectorPlugin ? configOrPlugin : new VectorPlugin(configOrPlugin);
   return async function POST(req) {
+    let body;
     try {
-      const body = await req.json();
-      const { message, history = [], namespace } = body;
-      const encoder = new TextEncoder();
-      const stream = new ReadableStream({
-        async start(controller) {
+      body = await req.json();
+    } catch (e) {
+      return new Response(JSON.stringify({ error: "Invalid JSON body" }), {
+        status: 400,
+        headers: { "Content-Type": "application/json" }
+      });
+    }
+    const { message, history = [], namespace } = body;
+    if (!(message == null ? void 0 : message.trim())) {
+      return new Response(JSON.stringify({ error: "message is required" }), {
+        status: 400,
+        headers: { "Content-Type": "application/json" }
+      });
+    }
+    const encoder = new TextEncoder();
+    const stream = new ReadableStream({
+      async start(controller) {
+        const enqueue = (text) => controller.enqueue(encoder.encode(text));
+        try {
+          const pipelineStream = plugin.chatStream(message, history, namespace);
           try {
-            const pipelineStream = plugin.chatStream(message, history, namespace);
-            try {
-              for (var iter = __forAwait(pipelineStream), more, temp, error; more = !(temp = await iter.next()).done; more = false) {
-                const chunk = temp.value;
-                if (typeof chunk === "string") {
-                  controller.enqueue(encoder.encode(chunk));
-                } else {
-                  controller.enqueue(encoder.encode(`
-
-__METADATA__${JSON.stringify(chunk)}`));
-                }
+            for (var iter = __forAwait(pipelineStream), more, temp, error; more = !(temp = await iter.next()).done; more = false) {
+              const chunk = temp.value;
+              if (typeof chunk === "string") {
+                enqueue(sseTextFrame(chunk));
+              } else {
+                enqueue(sseMetaFrame(chunk));
               }
-            } catch (temp) {
-              error = [temp];
+            }
+          } catch (temp) {
+            error = [temp];
+          } finally {
+            try {
+              more && (temp = iter.return) && await temp.call(iter);
             } finally {
-              try {
-                more && (temp = iter.return) && await temp.call(iter);
-              } finally {
-                if (error)
-                  throw error[0];
-              }
+              if (error)
+                throw error[0];
             }
-            controller.close();
-          } catch (streamError) {
-            console.error("[createStreamHandler] Stream processing error:", streamError);
-            const errorMessage = streamError instanceof Error ? streamError.message : String(streamError);
-            controller.enqueue(encoder.encode(`
-
-__ERROR__${JSON.stringify({ error: errorMessage })}`));
-            controller.close();
           }
+        } catch (streamError) {
+          const errorMessage = streamError instanceof Error ? streamError.message : String(streamError);
+          console.error("[createStreamHandler] Stream error:", streamError);
+          enqueue(sseErrorFrame(errorMessage));
+        } finally {
+          controller.close();
         }
-      });
-      return new Response(stream, {
-        headers: {
-          "Content-Type": "text/event-stream",
-          "Cache-Control": "no-cache",
-          "Connection": "keep-alive"
-        }
-      });
-    } catch (err) {
-      const message = err instanceof Error ? err.message : "Internal server error";
-      return new Response(JSON.stringify({ error: message }), {
-        status: 500,
-        headers: { "Content-Type": "application/json" }
-      });
-    }
+      }
+    });
+    return new Response(stream, { headers: SSE_HEADERS });
   };
 }
-function createIngestHandler(config) {
-  const plugin = new VectorPlugin(config);
+function createIngestHandler(configOrPlugin) {
+  const plugin = configOrPlugin instanceof VectorPlugin ? configOrPlugin : new VectorPlugin(configOrPlugin);
   return async function POST(req) {
     try {
       const body = await req.json();
@@ -2594,8 +2680,8 @@ function createIngestHandler(config) {
     }
   };
 }
-function createHealthHandler(config) {
-  const plugin = new VectorPlugin(config);
+function createHealthHandler(configOrPlugin) {
+  const plugin = configOrPlugin instanceof VectorPlugin ? configOrPlugin : new VectorPlugin(configOrPlugin);
   return async function GET() {
     try {
       const health = await plugin.checkHealth();
@@ -2611,8 +2697,8 @@ function createHealthHandler(config) {
     }
   };
 }
-function createUploadHandler(config) {
-  const plugin = new VectorPlugin(config);
+function createUploadHandler(configOrPlugin) {
+  const plugin = configOrPlugin instanceof VectorPlugin ? configOrPlugin : new VectorPlugin(configOrPlugin);
   return async function POST(req) {
     try {
       const formData = await req.formData();
@@ -2647,6 +2733,7 @@ function createUploadHandler(config) {
 
 export {
   getRagConfig,
+  resetConfigCache,
   ConfigResolver,
   OpenAIProvider,
   AnthropicProvider,
@@ -2665,6 +2752,10 @@ export {
   ProviderHealthCheck,
   VectorPlugin,
   DocumentParser,
+  sseFrame,
+  sseTextFrame,
+  sseMetaFrame,
+  sseErrorFrame,
   createChatHandler,
   createStreamHandler,
   createIngestHandler,

package/dist/handlers/index.d.mts

@@ -1,3 +1,3 @@
-import '../RagConfig-DRJO4hGU.mjs';
-export { c as createChatHandler, b as createHealthHandler, d as createIngestHandler, f as createStreamHandler, e as createUploadHandler } from '../index-Bjy0es5a.mjs';
+import '../RagConfig-BEmz4hVP.mjs';
+export { 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-DFSy0CG6.mjs';
 import 'next/server';

package/dist/handlers/index.d.ts

@@ -1,3 +1,3 @@
-import '../RagConfig-DRJO4hGU.js';
-export { c as createChatHandler, b as createHealthHandler, d as createIngestHandler, f as createStreamHandler, e as createUploadHandler } from '../index-B2mutkgp.js';
+import '../RagConfig-BEmz4hVP.js';
+export { 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-CSfaCeux.js';
 import 'next/server';