@retrivora-ai/rag-engine 2.0.4 → 2.0.6

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@retrivora-ai/rag-engine",
-  "version": "2.0.4",
+  "version": "2.0.6",
   "description": "Retrivora AI is a plug-and-play AI engine for RAG chat experiences — generic vector DB + LLM provider, embeddable or standalone.",
   "author": "Abhinav Alkuchi",
   "license": "UNLICENSED",
@@ -75,7 +75,7 @@
   "scripts": {
     "dev": "npx @tailwindcss/cli -i src/tailwind.css -o src/index.css && tsup src/index.ts src/handlers/index.ts src/server.ts --format cjs,esm --watch --tsconfig tsconfig.build.json --external react,react-dom,next,mongodb,pg,openai,@anthropic-ai/sdk,@pinecone-database/pinecone,axios,lucide-react,mammoth,pdf-parse,xlsx,react-markdown,remark-gfm,next-themes,langchain,@langchain/core,@langchain/openai,llamaindex --inject-style",
     "build": "npm run build:pkg",
-    "build:pkg": "npx @tailwindcss/cli -i src/tailwind.css -o src/index.css && tsup src/index.ts src/handlers/index.ts src/server.ts --format cjs,esm --dts --clean --no-splitting --tsconfig tsconfig.build.json --external react,react-dom,next,mongodb,pg,openai,@anthropic-ai/sdk,@pinecone-database/pinecone,axios,lucide-react,mammoth,pdf-parse,xlsx,react-markdown,remark-gfm,next-themes,langchain,@langchain/core,@langchain/openai,llamaindex --inject-style && npx @tailwindcss/cli -i src/tailwind.css -o dist/index.css && rm src/index.css",
+    "build:pkg": "npx @tailwindcss/cli -i src/tailwind.css -o src/index.css && tsup src/index.ts src/handlers/index.ts src/server.ts --format cjs,esm --dts --clean --no-splitting --tsconfig tsconfig.build.json --external react,react-dom,next,mongodb,pg,openai,@anthropic-ai/sdk,@pinecone-database/pinecone,axios,lucide-react,mammoth,pdf-parse,xlsx,react-markdown,remark-gfm,next-themes,langchain,@langchain/core,@langchain/openai,llamaindex --inject-style && npx @tailwindcss/cli -i src/tailwind.css -o dist/index.css",
     "lint": "eslint",
     "clean": "rm -rf dist"
   },

package/src/config/serverConfig.ts

@@ -18,7 +18,8 @@
  *   universal_rest→ VECTOR_BASE_URL, VECTOR_DB_REST_API_KEY
  */
 
-import { RagConfig } from './RagConfig';
+import { RagConfig, LLMProvider, EmbeddingProvider } from './RagConfig';
+import { LicenseVerifier } from '../core/LicenseVerifier';
 import {
   VECTOR_DB_PROVIDERS,
   LLM_PROVIDERS,
@@ -78,9 +79,7 @@ export function getEnvConfig(env: Record<string, string | undefined> = process.e
     Boolean(licenseKey);
 
   const vectorProvider = readEnum(env, 'VECTOR_DB_PROVIDER', 'pinecone', VECTOR_DB_PROVIDERS);
-  const llmProvider = readEnum(env, 'LLM_PROVIDER', 'openai', LLM_PROVIDERS);
-  const embeddingProvider = readEnum(env, 'EMBEDDING_PROVIDER', 'openai', EMBEDDING_PROVIDERS);
-  const embeddingDimensions = readNumber(env, 'EMBEDDING_DIMENSIONS', 1536);
+  const embeddingDimensions = readNumber(env, 'EMBEDDING_DIMENSIONS', 768);
 
   // Build provider-specific vectorDb options
   const vectorDbOptions: Record<string, unknown> = {};
@@ -160,9 +159,37 @@ export function getEnvConfig(env: Record<string, string | undefined> = process.e
     custom: 'default',
   };
 
+  // Check if the license key belongs to a Free / Trial tier
+  const isFreeTier = (() => {
+    if (!licenseKey) return true;
+    try {
+      const payload = LicenseVerifier.verify(licenseKey, projectId);
+      const tier = (payload.tier || '').toLowerCase();
+      return tier === 'free_trial' || tier === 'free_tier' || tier === 'free' || tier === 'hobby';
+    } catch {
+      return true;
+    }
+  })();
+
+  const DEFAULT_FREE_GATEWAY = 'https://llm.retrivora.com/api/v1';
+
+  const defaultLlmProvider = isFreeTier ? 'universal_rest' : 'universal_rest';
+  const llmProvider = (readEnum(env, 'LLM_PROVIDER', defaultLlmProvider, LLM_PROVIDERS) ?? base?.llm?.provider ?? defaultLlmProvider) as LLMProvider;
+  const llmBaseUrl = readString(env, 'LLM_BASE_URL') ?? base?.llm?.baseUrl ?? DEFAULT_FREE_GATEWAY;
+  const llmModel = readString(env, 'LLM_MODEL') ?? base?.llm?.model ?? (isFreeTier ? 'llama-3.1-8b-instant' : (DEFAULT_MODEL_BY_PROVIDER[llmProvider] ?? 'gpt-4o'));
+  const llmApiKey = llmApiKeyByProvider[llmProvider] ?? readString(env, 'LLM_API_KEY') ?? base?.llm?.apiKey ?? 'sk-retrivora-cancer-280590';
+  const llmProfile = readString(env, 'LLM_UNIVERSAL_PROFILE') ?? (base?.llm?.options?.profile as string) ?? 'litellm';
+
+  const defaultEmbeddingProvider = isFreeTier ? 'universal_rest' : 'universal_rest';
+  const embeddingProvider = (readEnum(env, 'EMBEDDING_PROVIDER', defaultEmbeddingProvider, EMBEDDING_PROVIDERS) ?? base?.embedding?.provider ?? defaultEmbeddingProvider) as EmbeddingProvider;
+  const embeddingBaseUrl = readString(env, 'EMBEDDING_BASE_URL') ?? base?.embedding?.baseUrl ?? DEFAULT_FREE_GATEWAY;
+  const embeddingModel = readString(env, 'EMBEDDING_MODEL') ?? base?.embedding?.model ?? 'text-embedding-004';
+  const embeddingApiKey = embeddingApiKeyByProvider[embeddingProvider] ?? readString(env, 'EMBEDDING_API_KEY') ?? readString(env, 'LLM_API_KEY') ?? base?.embedding?.apiKey ?? 'sk-retrivora-cancer-280590';
+  const embeddingProfile = readString(env, 'EMBEDDING_UNIVERSAL_PROFILE') ?? (base?.embedding?.options?.profile as string) ?? 'litellm';
+
   return {
     projectId,
-    licenseKey: readString(env, 'RETRIVORA_LICENSE_KEY') ?? readString(env, 'LICENSE_KEY') ?? base?.licenseKey,
+    licenseKey,
     vectorDb: {
       provider: vectorProvider,
       indexName:
@@ -173,28 +200,28 @@ export function getEnvConfig(env: Record<string, string | undefined> = process.e
     },
     llm: {
       provider: llmProvider,
-      model: readString(env, 'LLM_MODEL') ?? DEFAULT_MODEL_BY_PROVIDER[llmProvider] ?? 'gpt-4o',
-      apiKey: llmApiKeyByProvider[llmProvider] ?? '',
-      baseUrl: readString(env, 'LLM_BASE_URL'),
+      model: llmModel,
+      apiKey: llmApiKey,
+      baseUrl: llmBaseUrl,
       systemPrompt: readString(env, 'LLM_SYSTEM_PROMPT'),
       maxTokens: readNumber(env, 'LLM_MAX_TOKENS', 4096),
       temperature: readNumber(env, 'LLM_TEMPERATURE', 0.7),
       options: {
-        profile: readString(env, 'LLM_UNIVERSAL_PROFILE'),
+        profile: llmProfile,
         thinking: readString(env, 'LLM_THINKING') === 'true' || readString(env, 'LLM_THINKING') === 'enabled' ? true : (readString(env, 'LLM_THINKING') === 'false' ? false : undefined),
         thinkingBudget: env.LLM_THINKING_BUDGET ? parseInt(env.LLM_THINKING_BUDGET, 10) : undefined,
       },
     },
     embedding: {
       provider: embeddingProvider,
-      model: readString(env, 'EMBEDDING_MODEL') ?? 'text-embedding-3-small',
-      apiKey: embeddingApiKeyByProvider[embeddingProvider],
-      baseUrl: readString(env, 'EMBEDDING_BASE_URL'),
+      model: embeddingModel,
+      apiKey: embeddingApiKey,
+      baseUrl: embeddingBaseUrl,
       dimensions: embeddingDimensions,
       queryPrefix: readString(env, 'EMBEDDING_QUERY_PREFIX'),
       documentPrefix: readString(env, 'EMBEDDING_DOCUMENT_PREFIX'),
       options: {
-        profile: readString(env, 'EMBEDDING_UNIVERSAL_PROFILE'),
+        profile: embeddingProfile,
       },
     },
     ui: {

package/src/core/DatabaseStorage.ts

@@ -1,5 +1,6 @@
 import * as fs from 'fs';
 import * as path from 'path';
+import * as os from 'os';
 import { RagConfig } from '../config/RagConfig';
 import { LicenseVerifier } from './LicenseVerifier';
 
@@ -35,7 +36,10 @@ export class DatabaseStorage {
   private feedbackTableName: string;
 
   // Filesystem fallback configuration
-  private fallbackDir = path.join(process.cwd(), '.retrivora');
+  private isServerless = Boolean(process.env.VERCEL || process.env.AWS_LAMBDA_FUNCTION_NAME || process.env.LAMBDA_TASK_ROOT);
+  private fallbackDir = this.isServerless
+    ? path.join(os.tmpdir(), '.retrivora')
+    : path.join(process.cwd(), '.retrivora');
   private historyFile = path.join(this.fallbackDir, 'history.json');
   private feedbackFile = path.join(this.fallbackDir, 'feedback.json');
 
@@ -191,10 +195,28 @@ export class DatabaseStorage {
 
   // Helper for FS fallback writes
   private writeLocalFile(filePath: string, data: any[]): void {
-    if (!fs.existsSync(this.fallbackDir)) {
-      fs.mkdirSync(this.fallbackDir, { recursive: true });
+    try {
+      const dir = path.dirname(filePath);
+      if (!fs.existsSync(dir)) {
+        fs.mkdirSync(dir, { recursive: true });
+      }
+      fs.writeFileSync(filePath, JSON.stringify(data, null, 2), 'utf-8');
+    } catch (err: any) {
+      if (err.code === 'EROFS') {
+        try {
+          const tmpDir = path.join(os.tmpdir(), '.retrivora');
+          if (!fs.existsSync(tmpDir)) {
+            fs.mkdirSync(tmpDir, { recursive: true });
+          }
+          const tmpFilePath = path.join(tmpDir, path.basename(filePath));
+          fs.writeFileSync(tmpFilePath, JSON.stringify(data, null, 2), 'utf-8');
+        } catch {
+          /* ignore write failure in read-only environment */
+        }
+      } else {
+        console.warn(`[DatabaseStorage] Failed to write local fallback file: ${err.message}`);
+      }
     }
-    fs.writeFileSync(filePath, JSON.stringify(data, null, 2), 'utf-8');
   }
 
   // ─── Message History Operations ──────────────────────────────────────────

package/src/core/Pipeline.ts

@@ -608,7 +608,7 @@ You are a helpful assistant. Use the provided context to answer questions accura
 
       // 4. Context Augmentation - Review all retrieved/reranked sources for full comprehension
       let context = fullSources.length
-        ? fullSources.map((m, i) => `[Source ${i + 1}]\n${m.content}`).join('\n\n---\n\n')
+        ? fullSources.map((m, i) => `[Source ${i + 1}]\n${m?.content ?? ''}`).join('\n\n---\n\n')
         : 'No relevant context found.';
 
       // Count the records of relevant sources to pass into the slice dynamically
@@ -722,7 +722,12 @@ You are a helpful assistant. Use the provided context to answer questions accura
         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.`)';
       }
 
-      const hardenedHistory = [...history];
+      const hardenedHistory = (Array.isArray(history) ? history : [])
+        .filter((m): m is ChatMessage => Boolean(m && typeof m === 'object'))
+        .map((m) => ({
+          role: m.role || 'user',
+          content: typeof m.content === 'string' ? m.content : (m.content != null ? String(m.content) : ''),
+        }));
       const userQuestion = { role: 'user', content: question + finalRestrictionSuffix } as ChatMessage;
       const messages: ChatMessage[] = [...hardenedHistory, userQuestion];
 
@@ -880,11 +885,11 @@ You are a helpful assistant. Use the provided context to answer questions accura
         rewrittenQuery,
         systemPrompt,
         userPrompt: question + finalRestrictionSuffix,
-        chunks: sources.map((s) => ({
+        chunks: (sources || []).filter(Boolean).map((s) => ({
           id: s.id,
           score: s.score,
-          content: s.content,
-          metadata: s.metadata ?? {},
+          content: s?.content ?? '',
+          metadata: s?.metadata ?? {},
           namespace: ns,
         } satisfies RetrievedChunk)),
         latency,

package/src/handlers/index.ts

@@ -271,14 +271,17 @@ export function createChatHandler(
 
     try {
       const body = await req.json();
-      const { message: rawMessage, history = [], namespace, sessionId = 'default', messageId, userMessageId } = body as {
-        message: string;
-        history?: ChatMessage[];
-        namespace?: string;
-        sessionId?: string;
-        messageId?: string;
-        userMessageId?: string;
-      };
+      const bodyObj = (body ?? {}) as any;
+      let rawMessage = bodyObj.message;
+      if (!rawMessage && Array.isArray(bodyObj.messages) && bodyObj.messages.length > 0) {
+        const lastMsg = bodyObj.messages[bodyObj.messages.length - 1];
+        rawMessage = typeof lastMsg === 'string' ? lastMsg : (lastMsg?.content || lastMsg?.text);
+      }
+      if (!rawMessage && typeof bodyObj.prompt === 'string') {
+        rawMessage = bodyObj.prompt;
+      }
+      const history = bodyObj.history || (Array.isArray(bodyObj.messages) ? bodyObj.messages.slice(0, -1) : []);
+      const { namespace, sessionId = 'default', messageId, userMessageId } = bodyObj;
 
       // 2. Input sanitization
       const sanitized = sanitizeInput(rawMessage ?? '');
@@ -359,7 +362,17 @@ export function createStreamHandler(
       });
     }
 
-    const { message: rawMessage, history = [], namespace, sessionId = 'default', messageId, userMessageId } = body;
+    const bodyObj = (body ?? {}) as any;
+    let rawMessage = bodyObj.message;
+    if (!rawMessage && Array.isArray(bodyObj.messages) && bodyObj.messages.length > 0) {
+      const lastMsg = bodyObj.messages[bodyObj.messages.length - 1];
+      rawMessage = typeof lastMsg === 'string' ? lastMsg : (lastMsg?.content || lastMsg?.text);
+    }
+    if (!rawMessage && typeof bodyObj.prompt === 'string') {
+      rawMessage = bodyObj.prompt;
+    }
+    const history = bodyObj.history || (Array.isArray(bodyObj.messages) ? bodyObj.messages.slice(0, -1) : []);
+    const { namespace, sessionId = 'default', messageId, userMessageId } = bodyObj;
 
     // 2. Input sanitization
     const sanitized = sanitizeInput(rawMessage ?? '');