@retrivora-ai/rag-engine 1.9.8 → 1.9.9

package/dist/handlers/index.mjs

@@ -19,6 +19,12 @@ var __spreadValues = (a, b) => {
   return a;
 };
 var __spreadProps = (a, b) => __defProps(a, __getOwnPropDescs(b));
+var __require = /* @__PURE__ */ ((x) => typeof require !== "undefined" ? require : typeof Proxy !== "undefined" ? new Proxy(x, {
+  get: (a, b) => (typeof require !== "undefined" ? require : a)[b]
+}) : x)(function(x) {
+  if (typeof require !== "undefined") return require.apply(this, arguments);
+  throw Error('Dynamic require of "' + x + '" is not supported');
+});
 var __objRest = (source, exclude) => {
   var target = {};
   for (var prop in source)
@@ -2344,7 +2350,7 @@ var ConfigResolver = class {
     }
   }
   static mergeRetrievalWorkflow(rag, retrieval, workflow) {
-    var _a2, _b, _c, _d, _e;
+    var _a2, _b, _c, _d, _e, _f;
     if (!rag && !retrieval && !workflow) return void 0;
     const normalized = __spreadValues({}, rag != null ? rag : {});
     if (retrieval) {
@@ -2352,20 +2358,44 @@ var ConfigResolver = class {
       normalized.scoreThreshold = (_b = retrieval.scoreThreshold) != null ? _b : normalized.scoreThreshold;
       normalized.useReranking = (_c = retrieval.useReranking) != null ? _c : normalized.useReranking;
       if (retrieval.strategy === "hybrid") normalized.architecture = (_d = normalized.architecture) != null ? _d : "hybrid";
-      if (retrieval.strategy === "agentic") normalized.architecture = "agentic";
-      if (retrieval.strategy === "contextual-compression") normalized.useReranking = true;
+      if (retrieval.strategy === "agentic") normalized.architecture = "multi-agent";
+      if (retrieval.strategy === "contextual-compression") {
+        normalized.useReranking = true;
+        normalized.architecture = (_e = normalized.architecture) != null ? _e : "retrieve-and-rerank";
+      }
     }
     if (workflow == null ? void 0 : workflow.type) {
       const type = workflow.type;
-      if (type === "agentic" || type === "agentic-rag") normalized.architecture = "agentic";
-      else if (type === "hybrid-rag") normalized.architecture = "hybrid";
-      else if (type === "graph-rag") {
+      if (type === "naive-rag") {
+        normalized.architecture = "naive";
+        normalized.useReranking = false;
+        normalized.useGraphRetrieval = false;
+      } else if (type === "retrieve-and-rerank") {
+        normalized.architecture = "retrieve-and-rerank";
+        normalized.useReranking = true;
+      } else if (type === "multimodal-rag") {
+        normalized.architecture = "multimodal";
+      } else if (type === "graph-rag") {
         normalized.architecture = "graph";
         normalized.useGraphRetrieval = true;
+      } else if (type === "hybrid-rag") {
+        normalized.architecture = "hybrid";
+      } else if (type === "agentic-router") {
+        normalized.architecture = "agentic-router";
+      } else if (type === "multi-agent-rag" || type === "multi-agent" || type === "agentic" || type === "agentic-rag") {
+        normalized.architecture = "multi-agent";
       } else if (type === "simple-rag" || type === "rag") {
-        normalized.architecture = (_e = normalized.architecture) != null ? _e : "simple";
+        normalized.architecture = (_f = normalized.architecture) != null ? _f : "naive";
       }
     }
+    if (normalized.architecture === "retrieve-and-rerank") {
+      normalized.useReranking = true;
+    } else if (normalized.architecture === "graph") {
+      normalized.useGraphRetrieval = true;
+    } else if (normalized.architecture === "naive") {
+      normalized.useReranking = false;
+      normalized.useGraphRetrieval = false;
+    }
     return normalized;
   }
 };
@@ -7111,7 +7141,7 @@ var Pipeline = class {
     return this.initialised ? this.llmProvider : void 0;
   }
   async initialize() {
-    var _a2, _b, _c;
+    var _a2, _b, _c, _d;
     if (this.initialised) return;
     const chartInstruction = `You are a helpful assistant. Use the provided context to answer questions accurately.
 
@@ -7154,7 +7184,7 @@ var Pipeline = class {
       this.entityExtractor = new EntityExtractor(this.llmProvider);
     }
     await this.vectorDB.initialize();
-    if (((_a2 = this.config.rag) == null ? void 0 : _a2.architecture) === "agentic" || this.config.mcpServers && this.config.mcpServers.length > 0) {
+    if (((_a2 = this.config.rag) == null ? void 0 : _a2.architecture) === "agentic" || ((_b = this.config.rag) == null ? void 0 : _b.architecture) === "multi-agent" || this.config.mcpServers && this.config.mcpServers.length > 0) {
       this.mcpRegistry = new MCPRegistry(this.config.mcpServers || []);
       this.multiAgentCoordinator = new MultiAgentCoordinator({
         llmProvider: this.llmProvider,
@@ -7166,7 +7196,7 @@ var Pipeline = class {
       this.agent = new LangChainAgent(this, this.config);
       await this.agent.initialize(this.llmProvider);
     }
-    if ((_c = (_b = this.config.rag) == null ? void 0 : _b.cag) == null ? void 0 : _c.enabled) {
+    if ((_d = (_c = this.config.rag) == null ? void 0 : _c.cag) == null ? void 0 : _d.enabled) {
       await this.loadColdContext(this.config.projectId);
     }
     this.initialised = true;
@@ -7329,9 +7359,9 @@ ${m.content}`).join("\n\n---\n\n");
     return { reply, sources };
   }
   async ask(question, history = [], namespace) {
-    var _a2;
+    var _a2, _b;
     await this.initialize();
-    if ((((_a2 = this.config.rag) == null ? void 0 : _a2.architecture) === "agentic" || this.config.mcpServers && this.config.mcpServers.length > 0) && this.multiAgentCoordinator) {
+    if ((((_a2 = this.config.rag) == null ? void 0 : _a2.architecture) === "agentic" || ((_b = this.config.rag) == null ? void 0 : _b.architecture) === "multi-agent" || this.config.mcpServers && this.config.mcpServers.length > 0) && this.multiAgentCoordinator) {
       return await this.multiAgentCoordinator.run(question, history);
     }
     const stream = this.askStream(question, history, namespace);
@@ -7366,9 +7396,9 @@ ${m.content}`).join("\n\n---\n\n");
   }
   askStream(_0) {
     return __asyncGenerator(this, arguments, function* (question, history = [], namespace) {
-      var _a2;
+      var _a2, _b;
       yield new __await(this.initialize());
-      if ((((_a2 = this.config.rag) == null ? void 0 : _a2.architecture) === "agentic" || this.config.mcpServers && this.config.mcpServers.length > 0) && this.multiAgentCoordinator) {
+      if ((((_a2 = this.config.rag) == null ? void 0 : _a2.architecture) === "agentic" || ((_b = this.config.rag) == null ? void 0 : _b.architecture) === "multi-agent" || this.config.mcpServers && this.config.mcpServers.length > 0) && this.multiAgentCoordinator) {
         const stream2 = this.multiAgentCoordinator.runStream(question, history);
         try {
           for (var iter = __forAwait(stream2), more, temp, error; more = !(temp = yield new __await(iter.next())).done; more = false) {
@@ -7417,7 +7447,7 @@ ${m.content}`).join("\n\n---\n\n");
    */
   askStreamInternal(_0) {
     return __asyncGenerator(this, arguments, function* (question, history = [], namespace) {
-      var _a2, _b, _c, _d, _e, _f, _g2, _h, _i, _j, _k, _l, _m, _n, _o, _p, _q, _r, _s, _t, _u, _v, _w, _x, _y, _z;
+      var _a2, _b, _c, _d, _e, _f, _g2, _h, _i, _j, _k, _l, _m, _n, _o, _p, _q, _r, _s, _t, _u, _v, _w, _x, _y, _z, _A;
       yield new __await(this.initialize());
       const ns = namespace != null ? namespace : this.config.projectId;
       const topK = (_b = (_a2 = this.config.rag) == null ? void 0 : _a2.topK) != null ? _b : 5;
@@ -7440,13 +7470,14 @@ ${m.content}`).join("\n\n---\n\n");
         const embedStart = performance.now();
         const cacheKey = `${ns}::${searchQuery}`;
         const cachedVector = this.embeddingCache.get(cacheKey);
-        const useGraph = this.graphDB && ((_h = this.config.rag) == null ? void 0 : _h.useGraphRetrieval);
+        const arch = (_h = this.config.rag) == null ? void 0 : _h.architecture;
+        const useGraph = arch !== "naive" && arch !== "retrieve-and-rerank" && this.graphDB && ((_i = this.config.rag) == null ? void 0 : _i.useGraphRetrieval);
         const [strategyResult, embeddedVector] = yield new __await(Promise.all([
           QueryProcessor.determineRetrievalStrategy(
             searchQuery,
             useGraph ? this.llmRouter.get("fast") : void 0,
-            (_i = this.config.rag) == null ? void 0 : _i.graphKeywords,
-            (_j = this.config.rag) == null ? void 0 : _j.vectorKeywords
+            (_j = this.config.rag) == null ? void 0 : _j.graphKeywords,
+            (_k = this.config.rag) == null ? void 0 : _k.vectorKeywords
           ),
           // Embed immediately regardless of strategy — costs nothing if cached.
           // If strategy turns out to be 'graph'-only we just won't use the vector.
@@ -7456,7 +7487,7 @@ ${m.content}`).join("\n\n---\n\n");
         if (!cachedVector && queryVector.length > 0) {
           this.embeddingCache.set(cacheKey, queryVector);
         }
-        const graphData = (strategyResult === "graph" || strategyResult === "both") && this.graphDB && ((_k = this.config.rag) == null ? void 0 : _k.useGraphRetrieval) ? yield new __await(this.graphDB.query(searchQuery)) : void 0;
+        const graphData = (strategyResult === "graph" || strategyResult === "both") && this.graphDB && ((_l = this.config.rag) == null ? void 0 : _l.useGraphRetrieval) ? yield new __await(this.graphDB.query(searchQuery)) : void 0;
         const hasMetadataFilter = filter.metadata && Object.keys(filter.metadata).length > 0;
         const hasNumericPredicates = Array.isArray(filter.__numericPredicates) && filter.__numericPredicates.length > 0;
         const wantsExhaustiveList = hasNumericPredicates || hasMetadataFilter || /\b(list|all|show|provide|give|get|browse|find|search|display)\b/i.test(question);
@@ -7469,7 +7500,8 @@ ${m.content}`).join("\n\n---\n\n");
         const structuredSources = this.applyStructuredFilters(rawSources, filter);
         let fullSources = hasNumericPredicates ? structuredSources : structuredSources.filter((m) => m.score >= scoreThreshold);
         const rerankLimit = wantsExhaustiveList ? retrievalLimit : topK;
-        if (!hasNumericPredicates && ((_l = this.config.rag) == null ? void 0 : _l.useReranking)) {
+        const useReranking = arch !== "naive" && (arch === "retrieve-and-rerank" || ((_m = this.config.rag) == null ? void 0 : _m.useReranking));
+        if (!hasNumericPredicates && useReranking) {
           fullSources = yield new __await(this.reranker.rerank(fullSources, question, rerankLimit));
         } else if (!wantsExhaustiveList) {
           fullSources = fullSources.slice(0, topK);
@@ -7499,7 +7531,7 @@ ${graphContext}
 VECTOR CONTEXT:
 ${context}`;
         }
-        if ((_n = (_m = this.config.rag) == null ? void 0 : _m.cag) == null ? void 0 : _n.enabled) {
+        if ((_o = (_n = this.config.rag) == null ? void 0 : _n.cag) == null ? void 0 : _o.enabled) {
           if (!this.coldContexts.has(ns)) {
             yield new __await(this.loadColdContext(ns));
           }
@@ -7552,10 +7584,10 @@ ${context}`;
         let hallucinationScoringPromise = Promise.resolve(void 0);
         const restrictionSuffix = "\n\n(IMPORTANT: Format your response beautifully using rich Markdown! Use bolding for emphasis, headings (##) for structure, bullet points for lists, and proper line breaks between paragraphs to make it highly readable. However, NEVER generate Markdown tables, HTML figures, or text charts (the UI renders requested charts separately, so NEVER say you cannot create, display, draw, render, or provide a chart/visualization). If listing products, use simple markdown bullet points or comma-separated names. NEVER use plus signs (+) to separate product names, category names, or list items. For product description/detail questions, provide customer-facing prose only and do NOT list internal catalog fields such as Handle, Body (HTML), Vendor, Type, Tags, Published, Option, Variant, SKU, or raw metadata labels. You CAN use numbers for years/counts like 2006 or 5800, but NEVER put a dollar sign ($) before them.)";
         const isClaude37 = this.config.llm.model.includes("claude-3-7-sonnet");
-        const isNativeThinking = isClaude37 && (((_o = this.config.llm.options) == null ? void 0 : _o.thinking) === true || ((_p = this.config.llm.options) == null ? void 0 : _p.thinking) === "enabled" || ((_q = this.config.llm.options) == null ? void 0 : _q.thinking) !== false);
+        const isNativeThinking = isClaude37 && (((_p = this.config.llm.options) == null ? void 0 : _p.thinking) === true || ((_q = this.config.llm.options) == null ? void 0 : _q.thinking) === "enabled" || ((_r = this.config.llm.options) == null ? void 0 : _r.thinking) !== false);
         const modelNameLower = this.config.llm.model.toLowerCase();
         const isReasoningModel = modelNameLower.includes("-r1") || modelNameLower.includes("deepseek-r1") || modelNameLower.includes("reasoning") || modelNameLower.includes("thinking");
-        const isSimulatedThinking = !isNativeThinking && (((_r = this.config.llm.options) == null ? void 0 : _r.thinking) === true || ((_s = this.config.llm.options) == null ? void 0 : _s.thinking) === "enabled" || isReasoningModel && ((_t = this.config.llm.options) == null ? void 0 : _t.thinking) !== false);
+        const isSimulatedThinking = !isNativeThinking && (((_s = this.config.llm.options) == null ? void 0 : _s.thinking) === true || ((_t = this.config.llm.options) == null ? void 0 : _t.thinking) === "enabled" || isReasoningModel && ((_u = this.config.llm.options) == null ? void 0 : _u.thinking) !== false);
         let finalRestrictionSuffix = restrictionSuffix;
         if (isSimulatedThinking) {
           finalRestrictionSuffix += "\n\n(IMPORTANT: You must think step-by-step before answering. Write your thinking process inside a `<think>...</think>` block. Write the `<think>` block first, then follow it with your final response. Keep the thinking process thorough and detailed. Example: `<think>Thinking details here...</think>Final answer text here.`)";
@@ -7563,7 +7595,7 @@ ${context}`;
         const hardenedHistory = [...history];
         const userQuestion = { role: "user", content: question + finalRestrictionSuffix };
         const messages = [...hardenedHistory, userQuestion];
-        const systemPrompt = (_u = this.config.llm.systemPrompt) != null ? _u : "";
+        const systemPrompt = (_v = this.config.llm.systemPrompt) != null ? _v : "";
         const userPrompt = messages.map((m) => `${m.role}: ${m.content}`).join("\n");
         let fullReply = "";
         let thinkingText = "";
@@ -7674,7 +7706,7 @@ ${context}`;
           }
           yield fullReply;
         }
-        const runHallucination = ((_v = this.config.llm.options) == null ? void 0 : _v.hallucinationScoring) === true || this.config.llm.provider !== "ollama" && ((_w = this.config.llm.options) == null ? void 0 : _w.hallucinationScoring) !== false;
+        const runHallucination = ((_w = this.config.llm.options) == null ? void 0 : _w.hallucinationScoring) === true || this.config.llm.provider !== "ollama" && ((_x = this.config.llm.options) == null ? void 0 : _x.hallucinationScoring) !== false;
         if (runHallucination) {
           hallucinationScoringPromise = scoreHallucination(this.llmProvider, fullReply, context).catch(() => void 0);
         }
@@ -7683,7 +7715,7 @@ ${context}`;
         const latency = {
           embedMs: Math.round(embedMs),
           retrieveMs: Math.round(retrieveMs),
-          rerankMs: ((_x = this.config.rag) == null ? void 0 : _x.useReranking) ? Math.round(rerankMs) : void 0,
+          rerankMs: arch !== "naive" && (arch === "retrieve-and-rerank" || ((_y = this.config.rag) == null ? void 0 : _y.useReranking)) ? Math.round(rerankMs) : void 0,
           generateMs: Math.round(generateMs),
           totalMs: Math.round(totalMs)
         };
@@ -7696,7 +7728,7 @@ ${context}`;
           totalTokens: promptTokens + completionTokens,
           estimatedCostUsd: estimateCostUsd(promptTokens, completionTokens, this.config.llm.model)
         };
-        const awaitHallucination = ((_y = this.config.llm.options) == null ? void 0 : _y.awaitHallucination) === true;
+        const awaitHallucination = ((_z = this.config.llm.options) == null ? void 0 : _z.awaitHallucination) === true;
         const [uiTransformation, hallucinationResult] = yield new __await(Promise.all([
           uiTransformationPromise,
           awaitHallucination ? hallucinationScoringPromise : Promise.resolve(void 0)
@@ -7725,7 +7757,7 @@ ${context}`;
           hallucinationReason: hScore.reason
         } : {});
         const trace = buildTrace(hallucinationResult);
-        if ((_z = this.config.telemetry) == null ? void 0 : _z.enabled) {
+        if ((_A = this.config.telemetry) == null ? void 0 : _A.enabled) {
           const telemetryUrl = this.config.telemetry.url || "/api/telemetry";
           const absoluteUrl = telemetryUrl.startsWith("http") ? telemetryUrl : (process.env.NEXT_PUBLIC_APP_URL || `http://localhost:${process.env.PORT || 3e3}`) + telemetryUrl;
           (async () => {
@@ -8101,6 +8133,19 @@ var LicenseVerifier = class {
    */
   static verify(licenseKey, currentProjectId, provider, publicKeyOverride) {
     const isProduction = process.env.NODE_ENV === "production";
+    const now = Date.now();
+    if (licenseKey) {
+      const cacheKey = `${licenseKey}:${currentProjectId}:${provider || ""}:${publicKeyOverride || ""}`;
+      const cached = this.cache.get(cacheKey);
+      if (cached && now - cached.cachedAt < this.CACHE_TTL_MS) {
+        const currentTimestampSec = Math.floor(now / 1e3);
+        if (cached.payload.expiresAt && currentTimestampSec > cached.payload.expiresAt) {
+          this.cache.delete(cacheKey);
+        } else {
+          return cached.payload;
+        }
+      }
+    }
     if (!licenseKey) {
       if (isProduction) {
         throw new ConfigurationException(
@@ -8187,6 +8232,10 @@ var LicenseVerifier = class {
           }
         }
       }
+      if (licenseKey) {
+        const cacheKey = `${licenseKey}:${currentProjectId}:${provider || ""}:${publicKeyOverride || ""}`;
+        this.cache.set(cacheKey, { payload, cachedAt: now });
+      }
       return payload;
     } catch (err) {
       const message = err instanceof Error ? err.message : String(err);
@@ -8195,7 +8244,76 @@ var LicenseVerifier = class {
       );
     }
   }
+  static async verifyIP(licenseKey) {
+    if (!licenseKey) return;
+    try {
+      const parts = licenseKey.split(".");
+      if (parts.length !== 3) return;
+      const [, payloadB64] = parts;
+      const payload = JSON.parse(
+        Buffer.from(payloadB64, "base64url").toString("utf8")
+      );
+      const tier = (payload.tier || "").toLowerCase();
+      if (tier === "enterprise" || tier === "custom") {
+        return;
+      }
+      if (payload.ipv4 || payload.ipv6) {
+        let currentIpv4 = "";
+        let currentIpv6 = "";
+        try {
+          const res = await fetch("https://api4.ipify.org?format=json", { signal: AbortSignal.timeout(3e3) });
+          const data = await res.json();
+          currentIpv4 = data.ip;
+        } catch (e) {
+        }
+        try {
+          const res = await fetch("https://api6.ipify.org?format=json", { signal: AbortSignal.timeout(3e3) });
+          const data = await res.json();
+          currentIpv6 = data.ip;
+        } catch (e) {
+        }
+        const localIps = [];
+        try {
+          const osModule = __require("os");
+          const interfaces = osModule.networkInterfaces();
+          for (const name of Object.keys(interfaces)) {
+            for (const iface of interfaces[name] || []) {
+              if (!iface.internal) {
+                localIps.push(iface.address);
+              }
+            }
+          }
+        } catch (e) {
+        }
+        const isIpv4Match = payload.ipv4 && (currentIpv4 === payload.ipv4 || localIps.includes(payload.ipv4));
+        const isIpv6Match = payload.ipv6 && (currentIpv6 === payload.ipv6 || localIps.includes(payload.ipv6));
+        if (payload.ipv4 && payload.ipv6) {
+          if (!isIpv4Match && !isIpv6Match) {
+            throw new Error(
+              `License key access restricted. This license key was created for a different system (authorized IPv4: ${payload.ipv4}, IPv6: ${payload.ipv6}) and cannot be used on this machine.`
+            );
+          }
+        } else if (payload.ipv4 && !isIpv4Match) {
+          throw new Error(
+            `License key access restricted. This license key was created for a different system (authorized IPv4: ${payload.ipv4}) and cannot be used on this machine.`
+          );
+        } else if (payload.ipv6 && !isIpv6Match) {
+          throw new Error(
+            `License key access restricted. This license key was created for a different system (authorized IPv6: ${payload.ipv6}) and cannot be used on this machine.`
+          );
+        }
+      }
+    } catch (err) {
+      if (err.message.includes("License key access restricted")) {
+        throw new ConfigurationException(`[Retrivora SDK] ${err.message}`);
+      }
+    }
+  }
 };
+// A simple in-memory cache to save CPU overhead of cryptographic signature checks.
+LicenseVerifier.cache = /* @__PURE__ */ new Map();
+LicenseVerifier.CACHE_TTL_MS = 60 * 1e3;
+// Cache verified license for 60 seconds
 // Retrivora's Public Key used to verify the license key signature.
 // In production, this matches the private key held by Retrivora SaaS.
 LicenseVerifier.PUBLIC_KEY = `-----BEGIN PUBLIC KEY-----

package/dist/{index-DNvoi-sV.d.ts → index-CfkqZd2Y.d.ts}

@@ -1,4 +1,4 @@
-import { k as RagConfig, I as ILLMProvider, g as IngestDocument, C as ChatMessage, d as ChatResponse, u as RetrievalResult, n as UniversalRagConfig } from './ILLMProvider-B8ROITYK.js';
+import { k as RagConfig, I as ILLMProvider, g as IngestDocument, C as ChatMessage, d as ChatResponse, u as RetrievalResult, n as UniversalRagConfig } from './ILLMProvider-DMxLyTdq.js';
 
 /**
  * DocumentChunker — splits long text into overlapping chunks

package/dist/{index-CrxCy36A.d.mts → index-DXd29KMq.d.mts}

@@ -1,5 +1,5 @@
 import { NextRequest } from 'next/server';
-import { k as RagConfig, I as ILLMProvider, C as ChatMessage, d as ChatResponse, g as IngestDocument } from './ILLMProvider-B8ROITYK.mjs';
+import { k as RagConfig, I as ILLMProvider, C as ChatMessage, d as ChatResponse, g as IngestDocument } from './ILLMProvider-DMxLyTdq.mjs';
 
 interface ValidationError {
     field: string;

package/dist/{index-BCPKUAVL.d.ts → index-D_bOdJML.d.ts}

@@ -1,5 +1,5 @@
 import { NextRequest } from 'next/server';
-import { k as RagConfig, I as ILLMProvider, C as ChatMessage, d as ChatResponse, g as IngestDocument } from './ILLMProvider-B8ROITYK.js';
+import { k as RagConfig, I as ILLMProvider, C as ChatMessage, d as ChatResponse, g as IngestDocument } from './ILLMProvider-DMxLyTdq.js';
 
 interface ValidationError {
     field: string;

package/dist/{index-B8PbEFSY.d.mts → index-xygonxpW.d.mts}

@@ -1,4 +1,4 @@
-import { k as RagConfig, I as ILLMProvider, g as IngestDocument, C as ChatMessage, d as ChatResponse, u as RetrievalResult, n as UniversalRagConfig } from './ILLMProvider-B8ROITYK.mjs';
+import { k as RagConfig, I as ILLMProvider, g as IngestDocument, C as ChatMessage, d as ChatResponse, u as RetrievalResult, n as UniversalRagConfig } from './ILLMProvider-DMxLyTdq.mjs';
 
 /**
  * DocumentChunker — splits long text into overlapping chunks