@retrivora-ai/rag-engine 2.2.4 → 2.2.5

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@retrivora-ai/rag-engine",
-  "version": "2.2.4",
+  "version": "2.2.5",
   "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",
@@ -37,6 +37,8 @@
       "import": "./dist/index.mjs"
     },
     "./style.css": "./dist/index.css",
+    "./index.css": "./dist/index.css",
+    "./styles.css": "./dist/index.css",
     "./handlers": {
       "types": "./dist/handlers/index.d.ts",
       "require": "./dist/handlers/index.js",

package/src/config/UniversalProfiles.ts

@@ -14,8 +14,7 @@ const OPENAI_BASE = {
 export const LLM_PROFILES = {
   'openai-compatible': OPENAI_BASE,
   'litellm': { 
-    ...OPENAI_BASE, 
-    chatPayloadTemplate: '{"model":"{{model}}","messages":{{messages}},"max_tokens":{{maxTokens}},"temperature":{{temperature}},"stream":false}' 
+    ...OPENAI_BASE,
   },
   'anthropic-claude': {
     chatPath: '/v1/messages',

package/src/config/serverConfig.ts

@@ -187,8 +187,10 @@ export function getEnvConfig(env: Record<string, string | undefined> = process.e
 
   // Default gateway: retrivora.com/api/v1 acts as a secure proxy that validates the license key
   // and forwards to llm.retrivora.com with LITELLM_MASTER_KEY (which end-users never need to know).
-  // Developers who self-host can override with LITELLM_BASE_URL.
-  const defaultGatewayUrl = readString(env, 'LITELLM_BASE_URL') ?? readString(env, 'LLM_BASE_URL') ?? 'https://www.retrivora.com/api/v1';
+  const defaultGatewayUrl =
+    readString(env, 'LITELLM_BASE_URL') ??
+    readString(env, 'LLM_BASE_URL') ??
+    'https://www.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;
@@ -277,7 +279,7 @@ export function getEnvConfig(env: Record<string, string | undefined> = process.e
     },
     telemetry: {
       enabled: telemetryEnabled,
-      url: readString(env, 'TELEMETRY_URL') ?? readString(env, 'NEXT_PUBLIC_TELEMETRY_URL') ?? base?.telemetry?.url ?? (process.env.NODE_ENV === 'development' ? 'http://localhost:3001/api/telemetry' : 'https://retrivora.com/api/telemetry'),
+      url: readString(env, 'TELEMETRY_URL') ?? readString(env, 'NEXT_PUBLIC_TELEMETRY_URL') ?? base?.telemetry?.url ?? 'https://www.retrivora.com/api/telemetry',
     },
     // Optional graph DB — driven by GRAPH_DB_PROVIDER env var
     ...(readString(env, 'GRAPH_DB_PROVIDER') ? {

package/src/core/ConfigFetcher.ts

@@ -57,8 +57,8 @@ export class ConfigFetcher {
       process.env.NEXT_PUBLIC_RETRIVORA_CONTROL_PLANE_URL,
       'https://www.retrivora.com',
       'https://retrivora.com',
-      'http://localhost:3001',
       'http://localhost:3000',
+      'http://localhost:3001',
     ].filter(Boolean) as string[];
 
     for (const baseUrl of controlPlaneUrls) {

package/src/core/Pipeline.ts

@@ -586,16 +586,16 @@ You are a helpful assistant. Use the provided context to answer questions accura
         ? await this.vectorDB.query(queryVector, retrievalLimit, ns, filter)
         : []).filter((s): s is VectorMatch => Boolean(s && typeof s === 'object'));
 
-      console.log('[Pipeline] Raw vectorDB.query response:', {
-        rawCount: rawSources.length,
-        sample: rawSources.slice(0, 2).map((s) => ({
-          id: s?.id,
-          score: s?.score,
-          contentType: typeof s?.content,
-          contentSnippet: String(s?.content ?? '').substring(0, 150),
-          metadataKeys: Object.keys(s?.metadata || {}),
-        })),
-      });
+      // console.log('[Pipeline] Raw vectorDB.query response:', {
+      //   rawCount: rawSources.length,
+      //   sample: rawSources.slice(0, 2).map((s) => ({
+      //     id: s?.id,
+      //     score: s?.score,
+      //     contentType: typeof s?.content,
+      //     contentSnippet: String(s?.content ?? '').substring(0, 150),
+      //     metadataKeys: Object.keys(s?.metadata || {}),
+      //   })),
+      // });
 
       const retrieveEnd = performance.now();
       const embedMs = retrieveEnd - embedStart;
@@ -628,18 +628,18 @@ You are a helpful assistant. Use the provided context to answer questions accura
         .filter((s): s is VectorMatch => Boolean(s && typeof s === 'object'))
         .sort((a, b) => (b?.score ?? 0) - (a?.score ?? 0));
 
-      console.log('[Pipeline] Final sources for prompt & UI:', {
-        count: sources.length,
-        totalRawCount: rawSources.length,
-        sources: sources.slice(0, 10).map((s, idx) => ({
-          rank: idx + 1,
-          id: s?.id,
-          score: s?.score,
-          contentType: typeof s?.content,
-          contentSnippet: String(s?.content ?? '').substring(0, 150),
-          metadata: s?.metadata,
-        })),
-      });
+      // console.log('[Pipeline] Final sources for prompt & UI:', {
+      //   count: sources.length,
+      //   totalRawCount: rawSources.length,
+      //   sources: sources.slice(0, 10).map((s, idx) => ({
+      //     rank: idx + 1,
+      //     id: s?.id,
+      //     score: s?.score,
+      //     contentType: typeof s?.content,
+      //     contentSnippet: String(s?.content ?? '').substring(0, 150),
+      //     metadata: s?.metadata,
+      //   })),
+      // });
 
       if (graphData && graphData.nodes.length > 0) {
         const graphContext = graphData.nodes.map(n =>
@@ -912,14 +912,11 @@ You are a helpful assistant. Use the provided context to answer questions accura
       // Asynchronously trigger telemetry logging without blocking response yield
       const isTelemetryActive = this.config.telemetry?.enabled ?? Boolean(this.config.licenseKey);
       if (isTelemetryActive) {
-        const defaultUrl = process.env.NODE_ENV === 'development' && !process.env.TELEMETRY_URL && !process.env.NEXT_PUBLIC_TELEMETRY_URL
-          ? 'http://localhost:3001/api/telemetry'
-          : 'https://retrivora.com/api/telemetry';
-
+        const defaultUrl = 'https://www.retrivora.com/api/telemetry';
         const telemetryUrl = this.config.telemetry?.url || process.env.TELEMETRY_URL || process.env.NEXT_PUBLIC_TELEMETRY_URL || defaultUrl;
         const absoluteUrl = telemetryUrl.startsWith('http')
           ? telemetryUrl
-          : (process.env.NEXT_PUBLIC_APP_URL || `http://localhost:${process.env.PORT || 3000}`) + telemetryUrl;
+          : 'https://www.retrivora.com' + (telemetryUrl.startsWith('/') ? telemetryUrl : '/' + telemetryUrl);
 
         // Start background process for telemetry, awaiting hallucination scoring if it was backgrounded
         (async () => {

package/src/core/VectorPlugin.ts

@@ -155,8 +155,8 @@ export class VectorPlugin {
     }
     try {
       return await this.pipeline.getSuggestions(query, namespace);
-    } catch (err) {
-      throw wrapError(err, 'RETRIEVAL_FAILED');
+    } catch {
+      return [];
     }
   }
 }

package/src/handlers/index.ts

@@ -166,13 +166,10 @@ function reportTelemetry(
     const licenseKey = config.licenseKey || process.env.RAG_LICENSE_KEY || process.env.RETRIVORA_LICENSE_KEY || process.env.NEXT_PUBLIC_RETRIVORA_LICENSE_KEY;
     const telemetryConfig = config.telemetry;
 
-    const enabled = telemetryConfig?.enabled ?? Boolean(licenseKey);
-    const defaultUrl = process.env.NODE_ENV === 'development' && !process.env.TELEMETRY_URL && !process.env.NEXT_PUBLIC_TELEMETRY_URL
-      ? 'http://localhost:3001/api/telemetry'
-      : 'https://www.retrivora.com/api/telemetry';
+    const host = req.headers.get('host') || 'localhost';
+    const defaultUrl = 'https://www.retrivora.com/api/telemetry';
 
     const telemetryUrl = telemetryConfig?.url || process.env.TELEMETRY_URL || process.env.NEXT_PUBLIC_TELEMETRY_URL || defaultUrl;
-    const host = req.headers.get('host') || 'localhost';
     const userAgent = req.headers.get('user-agent') || `Retrivora-SDK/${SDK_VERSION}`;
 
     let absoluteUrl = telemetryUrl;
@@ -285,7 +282,7 @@ export function createChatHandler(
   const storage = new DatabaseStorage(plugin.getConfig());
   const onAuthorize = options?.onAuthorize;
 
-  return async function POST(req: NextRequest) {
+  return async function POST(req: any, context?: any) {
     const authResult = await checkAuth(req, onAuthorize);
     if (authResult) return authResult as any;
 
@@ -361,7 +358,7 @@ export function createStreamHandler(
   const storage = new DatabaseStorage(plugin.getConfig());
   const onAuthorize = options?.onAuthorize;
 
-  return async function POST(req: NextRequest) {
+  return async function POST(req: any, context?: any) {
     const authResult = await checkAuth(req, onAuthorize);
     if (authResult) return authResult as any;
 
@@ -532,7 +529,7 @@ export function createIngestHandler(
   const plugin = getOrCreatePlugin(configOrPlugin);
   const onAuthorize = options?.onAuthorize;
 
-  return async function POST(req: NextRequest) {
+  return async function POST(req: any, context?: any) {
     const authResult = await checkAuth(req, onAuthorize);
     if (authResult) return authResult as any;
 
@@ -599,7 +596,7 @@ export function createUploadHandler(
   const plugin = getOrCreatePlugin(configOrPlugin);
   const onAuthorize = options?.onAuthorize;
 
-  return async function POST(req: NextRequest) {
+  return async function POST(req: any, context?: any) {
     const authResult = await checkAuth(req, onAuthorize);
     if (authResult) return authResult as any;
 
@@ -926,7 +923,7 @@ export function createRagHandler(
   const feedbackHandler = createFeedbackHandler(plugin, { onAuthorize });
   const sessionsHandler = createSessionsHandler(plugin, { onAuthorize });
 
-  async function routePostRequest(req: NextRequest, segment: string) {
+  async function routePostRequest(req: any, segment: string) {
     switch (segment) {
       case 'chat':
         return streamHandler(req);
@@ -948,7 +945,7 @@ export function createRagHandler(
     }
   }
 
-  async function routeGetRequest(req: NextRequest, segment: string) {
+  async function routeGetRequest(req: any, segment: string) {
     switch (segment) {
       case 'health':
         return healthHandler(req);
@@ -965,7 +962,7 @@ export function createRagHandler(
     }
   }
 
-  async function getSegment(req: NextRequest, context: any): Promise<string> {
+  async function getSegment(req: any, context: any): Promise<string> {
     // 1. If content-type is multipart/form-data, it MUST be an upload request!
     const contentType = req.headers.get('content-type') || '';
     if (contentType.includes('multipart/form-data')) {
@@ -995,7 +992,7 @@ export function createRagHandler(
   }
 
   return {
-    GET: async (req: NextRequest, context: any) => {
+    GET: async (req: any, context?: any) => {
       try {
         const segment = await getSegment(req, context);
         return await routeGetRequest(req, segment);
@@ -1004,7 +1001,7 @@ export function createRagHandler(
         return NextResponse.json({ error: msg }, { status: 500 });
       }
     },
-    POST: async (req: NextRequest, context: any) => {
+    POST: async (req: any, context?: any) => {
       try {
         const segment = await getSegment(req, context);
         return await routePostRequest(req, segment);

package/src/llm/providers/UniversalLLMAdapter.ts

@@ -43,10 +43,7 @@ export class UniversalLLMAdapter implements ILLMProvider {
   constructor(config: LLMConfig | EmbeddingConfig) {
     this.model = config.model;
 
-    // Type narrow to extract LLM-specific fields safely
     const llmConfig = config as LLMConfig;
-
-    // Merge profile defaults if specified in options
     const options = (llmConfig.options as Record<string, unknown>) ?? {};
     let profile: Record<string, unknown> = {};
 
@@ -112,7 +109,7 @@ export class UniversalLLMAdapter implements ILLMProvider {
       };
     }
 
-    // Try direct HTTP POST request to llm.retrivora.com first
+    // Tier 1: Remote HTTP POST to Retrivora Cloud Proxy (e.g. https://www.retrivora.com/api/v1/chat/completions)
     try {
       const { data } = await this.http.post(path, payload);
       const extractPath = this.opts.responseExtractPath ?? 'choices[0].message.content';
@@ -121,20 +118,9 @@ export class UniversalLLMAdapter implements ILLMProvider {
         return String(result);
       }
     } catch (httpErr: any) {
-      // Direct POST fallback attempt
-      
+      // Tier 2: In-process fallback (only if HTTP call fails and server registered dispatch on globalThis)
       const _g = globalThis as any;
-      let dispatch = _g.__retrivoraDispatchChat;
-      if (typeof dispatch !== 'function') {
-        try {
-          const gateway = await import('../../../../../services/llm-gateway/router');
-          if (gateway?.dispatchChatCompletion) {
-            _g.__retrivoraDispatchChat = gateway.dispatchChatCompletion;
-            _g.__retrivoraDispatchEmbedding = gateway.dispatchEmbedding;
-            dispatch = gateway.dispatchChatCompletion;
-          }
-        } catch { /* proceed */ }
-      }
+      const dispatch = _g.__retrivoraDispatchChat;
 
       if (typeof dispatch === 'function') {
         const res = await dispatch({
@@ -153,7 +139,6 @@ export class UniversalLLMAdapter implements ILLMProvider {
     }
 
     const { data } = await this.http.post(path, payload);
-
     const extractPath = this.opts.responseExtractPath ?? 'choices[0].message.content';
     const result = resolvePath(data, extractPath) ?? extractContent(data);
 
@@ -167,12 +152,10 @@ export class UniversalLLMAdapter implements ILLMProvider {
   /**
    * Streaming chat using native fetch + ReadableStream.
    * Parses OpenAI-compatible SSE frames: `data: {...}\n\n`
-   * Works with vLLM, LMStudio, Together AI, Fireworks, and any OpenAI-compatible API.
    */
   async *chatStream(messages: ChatMessage[], context?: string): AsyncIterable<string> {
     const path = this.opts.chatPath ?? '/chat/completions';
     const url = `${this.baseUrl.replace(/\/$/, '')}${path}`;
-    // Delta extract path — same root as chat but with delta instead of message
     const extractPath = (this.opts.responseExtractPath ?? 'choices[0].message.content')
       .replace('message.content', 'delta.content');
 
@@ -196,7 +179,6 @@ export class UniversalLLMAdapter implements ILLMProvider {
         maxTokens: this.maxTokens,
         temperature: this.temperature,
       });
-      // Inject stream: true into the payload if it's an object
       if (typeof payload === 'object' && payload !== null) {
         (payload as Record<string, unknown>).stream = true;
       }
@@ -210,22 +192,40 @@ export class UniversalLLMAdapter implements ILLMProvider {
       };
     }
 
-    const isSelfHost = this.baseUrl.includes('retrivora.com') || this.baseUrl.includes('localhost');
     let streamBody: ReadableStream<Uint8Array> | null = null;
 
-    if (isSelfHost || Boolean(process.env.VERCEL)) {
-      const _g = globalThis as any;
-      let dispatch = _g.__retrivoraDispatchChat;
-      if (typeof dispatch !== 'function') {
-        try {
-          const gateway = await import('../../../../../services/llm-gateway/router');
-          if (gateway?.dispatchChatCompletion) {
-            _g.__retrivoraDispatchChat = gateway.dispatchChatCompletion;
-            _g.__retrivoraDispatchEmbedding = gateway.dispatchEmbedding;
-            dispatch = gateway.dispatchChatCompletion;
+    // ── Tier 1: HTTP Stream Request to Retrivora Gateway Proxy ─────────────────
+    try {
+      const response = await fetch(url, {
+        method: 'POST',
+        headers: this.resolvedHeaders,
+        body: typeof payload === 'string' ? payload : JSON.stringify(payload),
+      });
+
+      if (response.ok) {
+        const contentType = response.headers.get('content-type') || '';
+        if (contentType.includes('application/json')) {
+          const json = await response.json();
+          const text = (resolvePath(json, extractPath) as string | undefined) ?? extractContent(json);
+          if (text && typeof text === 'string') {
+            yield text;
+            return;
           }
-        } catch { /* proceed */ }
+        } else if (response.body) {
+          streamBody = response.body;
+        }
+      } else {
+        const errorText = await response.text().catch(() => response.statusText);
+        console.warn(`[UniversalLLMAdapter] Remote HTTP stream returned ${response.status}: ${errorText}. Attempting in-process fallback...`);
       }
+    } catch (fetchErr: any) {
+      console.warn(`[UniversalLLMAdapter] Remote HTTP stream fetch warning: ${fetchErr?.message}. Attempting in-process fallback...`);
+    }
+
+    // ── Tier 2: In-process Dispatch Fallback (if remote HTTP fails) ────────────
+    if (!streamBody) {
+      const _g = globalThis as any;
+      const dispatch = _g.__retrivoraDispatchChat;
 
       if (typeof dispatch === 'function') {
         try {
@@ -245,32 +245,15 @@ export class UniversalLLMAdapter implements ILLMProvider {
               yield content;
               return;
             }
-            throw new Error(`[UniversalLLMAdapter] In-process dispatch stream returned empty response for model: ${this.model}`);
           }
         } catch (dispatchErr) {
-          throw dispatchErr;
+          console.warn('[UniversalLLMAdapter] In-process dispatch error:', dispatchErr);
         }
-      } else {
-        throw new Error(`[UniversalLLMAdapter] In-process gateway dispatch not registered. Direct self-referential HTTP calls to ${this.baseUrl} are disabled on Vercel to prevent HTTP 508 Loop Detected.`);
       }
     }
 
     if (!streamBody) {
-      const response = await fetch(url, {
-        method: 'POST',
-        headers: this.resolvedHeaders,
-        body: JSON.stringify(payload),
-      });
-
-      if (!response.ok) {
-        const errorText = await response.text().catch(() => response.statusText);
-        throw new Error(`[UniversalLLMAdapter] Streaming request failed (${response.status}): ${errorText}`);
-      }
-
-      if (!response.body) {
-        throw new Error('[UniversalLLMAdapter] Response body is null — server did not send a streaming response.');
-      }
-      streamBody = response.body;
+      throw new Error(`[UniversalLLMAdapter] Streaming request failed. Ensure RETRIVORA_LICENSE_KEY is valid or LLM service is available.`);
     }
 
     const reader = streamBody.getReader();
@@ -296,12 +279,11 @@ export class UniversalLLMAdapter implements ILLMProvider {
             const text = (resolvePath(json, extractPath) as string | undefined) ?? extractContent(json);
             if (text && typeof text === 'string') yield text;
           } catch {
-            // Malformed SSE line — skip
+            // Skip malformed SSE lines
           }
         }
       }
 
-      // Flush remaining buffer
       if (buffer.trim() && buffer.trim() !== 'data: [DONE]') {
         const jsonStr = buffer.replace(/^data:\s*/, '').trim();
         try {
@@ -317,85 +299,52 @@ export class UniversalLLMAdapter implements ILLMProvider {
 
   async embed(text: string): Promise<number[]> {
     const path = this.opts.embedPath ?? '/embeddings';
+    const payload = this.opts.embedPayloadTemplate
+      ? buildPayload(this.opts.embedPayloadTemplate, { input: text, model: this.model })
+      : { input: text, model: this.model };
 
-    let payload: unknown;
-    if (this.opts.embedPayloadTemplate) {
-      payload = buildPayload(this.opts.embedPayloadTemplate, {
-        model: this.model,
-        input: text,
-      });
-    } else {
-      payload = {
-        model: this.model,
-        input: text,
-      };
-    }
-
-    // Try direct HTTP POST request to llm.retrivora.com first
     try {
       const { data } = await this.http.post(path, payload);
       const extractPath = this.opts.embedExtractPath ?? 'data[0].embedding';
-      const vector = resolvePath(data, extractPath);
-      if (Array.isArray(vector)) {
-        return vector as number[];
-      }
-    } catch (httpErr: any) {
-      console.warn(`[UniversalLLMAdapter] Direct HTTP embedding POST to ${this.baseUrl}${path} failed (${httpErr.message}). Attempting in-process gateway dispatch fallback...`);
-
+      const vector = resolvePath(data, extractPath) as number[] | undefined;
+      if (Array.isArray(vector)) return vector;
+    } catch {
+      // In-process fallback for embeddings
       const _g = globalThis as any;
-      let dispatch = _g.__retrivoraDispatchEmbedding;
-      if (typeof dispatch !== 'function') {
-        try {
-          const gateway = await import('../../../../../services/llm-gateway/router');
-          if (gateway?.dispatchEmbedding) {
-            _g.__retrivoraDispatchChat = gateway.dispatchChatCompletion;
-            _g.__retrivoraDispatchEmbedding = gateway.dispatchEmbedding;
-            dispatch = gateway.dispatchEmbedding;
-          }
-        } catch { /* proceed */ }
-      }
-
-      if (typeof dispatch === 'function') {
-        const res = await dispatch({
-          model: this.model,
-          input: text,
-        }, this.apiKey);
+      const dispatchEmbed = _g.__retrivoraDispatchEmbedding;
 
+      if (typeof dispatchEmbed === 'function') {
+        const res = await dispatchEmbed({ input: text, model: this.model }, this.apiKey);
         if (res?.data?.[0]?.embedding) {
           return res.data[0].embedding;
         }
       }
-      throw httpErr;
     }
 
     const { data } = await this.http.post(path, payload);
-
     const extractPath = this.opts.embedExtractPath ?? 'data[0].embedding';
-    const vector = resolvePath(data, extractPath);
+    const vector = resolvePath(data, extractPath) as number[] | undefined;
 
     if (!Array.isArray(vector)) {
-      throw new Error(`[UniversalLLMAdapter] Expected a number array at '${extractPath}' for embeddings.`);
+      throw new Error(`[UniversalLLMAdapter] Could not extract embedding vector from path '${extractPath}' in response.`);
     }
 
     return vector;
   }
 
   async batchEmbed(texts: string[]): Promise<number[][]> {
-    const vectors: number[][] = [];
+    const results: number[][] = [];
     for (const text of texts) {
-      vectors.push(await this.embed(text));
+      results.push(await this.embed(text));
     }
-    return vectors;
+    return results;
   }
 
   async ping(): Promise<boolean> {
     try {
-      if (this.opts.pingPath) {
-        await this.http.get(this.opts.pingPath);
-      }
+      await this.embed('ping');
       return true;
-    } catch (err) {
-      console.error('[UniversalLLMAdapter] Ping failed:', err);
+    } catch {
       return false;
     }
   }