@retrivora-ai/rag-engine 2.2.4 → 2.2.5

package/dist/handlers/index.mjs

@@ -106,644 +106,6 @@ var init_templateUtils = __esm({
   }
 });
 
-// ../../services/llm-gateway/providers/groq.ts
-async function handleGroqRequest(req, apiKeyOverride) {
-  const apiKey = apiKeyOverride || process.env.GROQ_API_KEY;
-  if (!apiKey) {
-    throw new Error("LLM service credentials not configured.");
-  }
-  const modelName = req.model.replace(/^groq\//i, "");
-  const payload = __spreadProps(__spreadValues({}, req), {
-    model: modelName
-  });
-  const response = await fetch("https://api.groq.com/openai/v1/chat/completions", {
-    method: "POST",
-    headers: {
-      "Content-Type": "application/json",
-      "Authorization": `Bearer ${apiKey}`
-    },
-    body: JSON.stringify(payload)
-  });
-  if (!response.ok) {
-    const errorText = await response.text();
-    throw new Error(`Groq API Error (${response.status}): ${errorText}`);
-  }
-  if (req.stream && response.body) {
-    return { stream: response.body };
-  }
-  const json = await response.json();
-  return { response: json };
-}
-var init_groq = __esm({
-  "../../services/llm-gateway/providers/groq.ts"() {
-    "use strict";
-  }
-});
-
-// ../../services/llm-gateway/providers/openai.ts
-async function handleOpenAIRequest(req, apiKeyOverride) {
-  const apiKey = apiKeyOverride || process.env.OPENAI_API_KEY;
-  if (!apiKey) {
-    throw new Error("OpenAI API Key missing. Please set OPENAI_API_KEY in environment variables.");
-  }
-  const modelName = req.model.replace(/^(openai|gpt)\//i, "");
-  const payload = __spreadProps(__spreadValues({}, req), {
-    model: modelName
-  });
-  const response = await fetch("https://api.openai.com/v1/chat/completions", {
-    method: "POST",
-    headers: {
-      "Content-Type": "application/json",
-      "Authorization": `Bearer ${apiKey}`
-    },
-    body: JSON.stringify(payload)
-  });
-  if (!response.ok) {
-    const errorText = await response.text();
-    throw new Error(`OpenAI API Error (${response.status}): ${errorText}`);
-  }
-  if (req.stream && response.body) {
-    return { stream: response.body };
-  }
-  const json = await response.json();
-  return { response: json };
-}
-async function handleOpenAIEmbedding(req, apiKeyOverride) {
-  const apiKey = apiKeyOverride || process.env.OPENAI_API_KEY;
-  if (!apiKey) {
-    throw new Error("OpenAI API Key missing. Please set OPENAI_API_KEY in environment variables.");
-  }
-  const modelName = req.model.replace(/^(openai)\//i, "");
-  const response = await fetch("https://api.openai.com/v1/embeddings", {
-    method: "POST",
-    headers: {
-      "Content-Type": "application/json",
-      "Authorization": `Bearer ${apiKey}`
-    },
-    body: JSON.stringify({
-      model: modelName,
-      input: req.input
-    })
-  });
-  if (!response.ok) {
-    const errorText = await response.text();
-    throw new Error(`OpenAI Embedding API Error (${response.status}): ${errorText}`);
-  }
-  return await response.json();
-}
-var init_openai = __esm({
-  "../../services/llm-gateway/providers/openai.ts"() {
-    "use strict";
-  }
-});
-
-// ../../services/llm-gateway/providers/gemini.ts
-async function handleGeminiRequest(req, apiKeyOverride) {
-  const apiKey = apiKeyOverride || process.env.GEMINI_API_KEY;
-  if (!apiKey) {
-    throw new Error("Gemini API Key missing. Please set GEMINI_API_KEY in environment variables.");
-  }
-  let modelName = req.model.replace(/^(google|gemini)\//i, "");
-  if (!modelName.startsWith("gemini-")) {
-    modelName = `gemini-${modelName}`;
-  }
-  const payload = __spreadProps(__spreadValues({}, req), {
-    model: modelName
-  });
-  const response = await fetch("https://generativelanguage.googleapis.com/v1beta/openai/chat/completions", {
-    method: "POST",
-    headers: {
-      "Content-Type": "application/json",
-      "Authorization": `Bearer ${apiKey}`
-    },
-    body: JSON.stringify(payload)
-  });
-  if (!response.ok) {
-    const errorText = await response.text();
-    throw new Error(`Google Gemini API Error (${response.status}): ${errorText}`);
-  }
-  if (req.stream && response.body) {
-    return { stream: response.body };
-  }
-  const json = await response.json();
-  return { response: json };
-}
-async function handleGeminiEmbedding(req, apiKeyOverride) {
-  const apiKey = apiKeyOverride || process.env.GEMINI_API_KEY;
-  if (!apiKey) {
-    throw new Error("Gemini API Key missing. Please set GEMINI_API_KEY in environment variables.");
-  }
-  let modelName = req.model.replace(/^(google|gemini)\//i, "");
-  if (!modelName || modelName === "default" || modelName.includes("text-embedding-3") || modelName.includes("ada") || !modelName.startsWith("text-embedding-00") && !modelName.startsWith("embedding-00")) {
-    modelName = "text-embedding-004";
-  }
-  const response = await fetch("https://generativelanguage.googleapis.com/v1beta/openai/embeddings", {
-    method: "POST",
-    headers: {
-      "Content-Type": "application/json",
-      "Authorization": `Bearer ${apiKey}`
-    },
-    body: JSON.stringify({
-      model: modelName,
-      input: req.input
-    })
-  });
-  if (!response.ok) {
-    const errorText = await response.text();
-    throw new Error(`Google Gemini Embedding API Error (${response.status}): ${errorText}`);
-  }
-  return await response.json();
-}
-var init_gemini = __esm({
-  "../../services/llm-gateway/providers/gemini.ts"() {
-    "use strict";
-  }
-});
-
-// ../../services/llm-gateway/providers/huggingface.ts
-async function handleHuggingFaceChatRequest(req, apiKeyOverride) {
-  var _a2, _b;
-  const apiKey = apiKeyOverride || process.env.HUGGINGFACE_API_KEY || process.env.HF_TOKEN;
-  if (!apiKey) {
-    throw new Error("HuggingFace API Key missing. Set HUGGINGFACE_API_KEY or HF_TOKEN in environment variables.");
-  }
-  let modelName = req.model.replace(/^(huggingface|hf)\//i, "");
-  if (!modelName || modelName === "default" || modelName.toLowerCase() === "qwen" || modelName.toLowerCase() === "qwen-coder") {
-    modelName = "Qwen/Qwen2.5-Coder-32B-Instruct";
-  }
-  const payload = __spreadProps(__spreadValues({}, req), {
-    model: modelName
-  });
-  try {
-    const response = await fetch("https://router.huggingface.co/v1/chat/completions", {
-      method: "POST",
-      headers: {
-        "Content-Type": "application/json",
-        "Authorization": `Bearer ${apiKey}`
-      },
-      body: JSON.stringify(payload)
-    });
-    if (response.ok) {
-      if (req.stream && response.body) {
-        return { stream: response.body };
-      }
-      const json = await response.json();
-      return { response: json };
-    }
-  } catch (e) {
-  }
-  try {
-    const response = await fetch("https://router.huggingface.co/hf-inference/v1/chat/completions", {
-      method: "POST",
-      headers: {
-        "Content-Type": "application/json",
-        "Authorization": `Bearer ${apiKey}`
-      },
-      body: JSON.stringify(payload)
-    });
-    if (response.ok) {
-      if (req.stream && response.body) {
-        return { stream: response.body };
-      }
-      const json = await response.json();
-      return { response: json };
-    }
-  } catch (e) {
-  }
-  const lastUserMsg = ((_a2 = req.messages.filter((m) => m.role === "user").pop()) == null ? void 0 : _a2.content) || "Hello";
-  const pipelineRes = await fetch(`https://router.huggingface.co/hf-inference/models/${modelName}`, {
-    method: "POST",
-    headers: {
-      "Content-Type": "application/json",
-      "Authorization": `Bearer ${apiKey}`
-    },
-    body: JSON.stringify({
-      inputs: lastUserMsg,
-      parameters: { max_new_tokens: req.max_tokens || 256, temperature: req.temperature || 0.7 },
-      options: { wait_for_model: true }
-    })
-  });
-  if (!pipelineRes.ok) {
-    const errorText = await pipelineRes.text();
-    throw new Error(`HuggingFace API Error (${pipelineRes.status}): ${errorText}`);
-  }
-  const raw = await pipelineRes.json();
-  const textOutput = Array.isArray(raw) ? ((_b = raw[0]) == null ? void 0 : _b.generated_text) || JSON.stringify(raw) : JSON.stringify(raw);
-  const formattedResponse = {
-    id: `hf-${Date.now()}`,
-    object: "chat.completion",
-    created: Math.floor(Date.now() / 1e3),
-    model: modelName,
-    choices: [
-      {
-        index: 0,
-        message: {
-          role: "assistant",
-          content: textOutput
-        },
-        finish_reason: "stop"
-      }
-    ],
-    usage: {
-      prompt_tokens: lastUserMsg.length,
-      completion_tokens: textOutput.length,
-      total_tokens: lastUserMsg.length + textOutput.length
-    }
-  };
-  return { response: formattedResponse };
-}
-async function handleHuggingFaceEmbedding(req, apiKeyOverride) {
-  const apiKey = apiKeyOverride || process.env.HUGGINGFACE_API_KEY || process.env.HF_TOKEN;
-  if (!apiKey) {
-    throw new Error("HuggingFace API Key missing. Set HUGGINGFACE_API_KEY or HF_TOKEN in environment variables.");
-  }
-  let modelName = req.model.replace(/^(huggingface|hf)\//i, "");
-  if (!modelName || modelName === "default" || modelName.includes("text-embedding-3") || modelName.includes("ada") || !modelName.includes("/") && !modelName.includes("bge") && !modelName.includes("sentence-transformers")) {
-    modelName = "BAAI/bge-base-en-v1.5";
-  }
-  const inputs = Array.isArray(req.input) ? req.input : [req.input];
-  const fetchEmbeddingFromModel = async (targetModel) => {
-    try {
-      const routerRes = await fetch("https://router.huggingface.co/hf-inference/v1/embeddings", {
-        method: "POST",
-        headers: {
-          "Content-Type": "application/json",
-          "Authorization": `Bearer ${apiKey}`
-        },
-        body: JSON.stringify({
-          model: targetModel,
-          input: inputs
-        })
-      });
-      if (routerRes.ok) {
-        const json = await routerRes.json();
-        if (json && json.data) return json;
-      }
-    } catch (e) {
-    }
-    const response = await fetch(`https://router.huggingface.co/hf-inference/models/${targetModel}`, {
-      method: "POST",
-      headers: {
-        "Content-Type": "application/json",
-        "Authorization": `Bearer ${apiKey}`
-      },
-      body: JSON.stringify({
-        inputs,
-        options: { wait_for_model: true }
-      })
-    });
-    if (!response.ok) {
-      const errorText = await response.text();
-      throw new Error(`HuggingFace API Error (${response.status}): ${errorText}`);
-    }
-    const rawData = await response.json();
-    const embeddings = Array.isArray(rawData[0]) ? typeof rawData[0][0] === "number" ? rawData : rawData.map((arr) => arr[0]) : [rawData];
-    return {
-      object: "list",
-      data: embeddings.map((vec, idx) => ({
-        object: "embedding",
-        embedding: vec,
-        index: idx
-      })),
-      model: targetModel,
-      usage: {
-        prompt_tokens: inputs.reduce((acc, str) => acc + str.length, 0),
-        completion_tokens: 0,
-        total_tokens: inputs.reduce((acc, str) => acc + str.length, 0)
-      }
-    };
-  };
-  try {
-    return await fetchEmbeddingFromModel(modelName);
-  } catch (err) {
-    if (modelName !== "BAAI/bge-base-en-v1.5") {
-      console.warn(`[LLM Gateway] HuggingFace model ${modelName} failed. Falling back to BAAI/bge-base-en-v1.5...`, err);
-      return await fetchEmbeddingFromModel("BAAI/bge-base-en-v1.5");
-    }
-    throw err;
-  }
-}
-var init_huggingface = __esm({
-  "../../services/llm-gateway/providers/huggingface.ts"() {
-    "use strict";
-  }
-});
-
-// ../../services/llm-gateway/providers/anthropic.ts
-async function handleAnthropicRequest(req, apiKeyOverride) {
-  var _a2, _b;
-  const apiKey = apiKeyOverride || process.env.ANTHROPIC_API_KEY;
-  if (!apiKey) {
-    throw new Error("Anthropic API Key missing. Please set ANTHROPIC_API_KEY in environment variables.");
-  }
-  const modelName = req.model.replace(/^(anthropic|claude)\//i, "");
-  let systemPrompt = void 0;
-  const anthropicMessages = [];
-  for (const msg of req.messages) {
-    if (msg.role === "system") {
-      systemPrompt = systemPrompt ? `${systemPrompt}
-${msg.content}` : msg.content;
-    } else {
-      anthropicMessages.push({
-        role: msg.role === "assistant" ? "assistant" : "user",
-        content: msg.content
-      });
-    }
-  }
-  if (anthropicMessages.length === 0) {
-    anthropicMessages.push({ role: "user", content: "Hello" });
-  }
-  const payload = {
-    model: modelName,
-    messages: anthropicMessages,
-    max_tokens: req.max_tokens || req.max_completion_tokens || 1024,
-    stream: req.stream || false
-  };
-  if (systemPrompt) {
-    payload.system = systemPrompt;
-  }
-  if (req.temperature !== void 0) {
-    payload.temperature = req.temperature;
-  }
-  const response = await fetch("https://api.anthropic.com/v1/messages", {
-    method: "POST",
-    headers: {
-      "Content-Type": "application/json",
-      "x-api-key": apiKey,
-      "anthropic-version": "2023-06-01"
-    },
-    body: JSON.stringify(payload)
-  });
-  if (!response.ok) {
-    const errorText = await response.text();
-    throw new Error(`Anthropic API Error (${response.status}): ${errorText}`);
-  }
-  if (req.stream && response.body) {
-    return { stream: response.body };
-  }
-  const json = await response.json();
-  const textContent = ((_b = (_a2 = json.content) == null ? void 0 : _a2[0]) == null ? void 0 : _b.text) || "";
-  const openAIResponse = {
-    id: json.id || `chatcmpl-${Date.now()}`,
-    object: "chat.completion",
-    created: Math.floor(Date.now() / 1e3),
-    model: json.model || modelName,
-    choices: [
-      {
-        index: 0,
-        message: {
-          role: "assistant",
-          content: textContent
-        },
-        finish_reason: json.stop_reason === "end_turn" ? "stop" : json.stop_reason || "stop"
-      }
-    ],
-    usage: json.usage ? {
-      prompt_tokens: json.usage.input_tokens,
-      completion_tokens: json.usage.output_tokens,
-      total_tokens: json.usage.input_tokens + json.usage.output_tokens
-    } : void 0
-  };
-  return { response: openAIResponse };
-}
-var init_anthropic = __esm({
-  "../../services/llm-gateway/providers/anthropic.ts"() {
-    "use strict";
-  }
-});
-
-// ../../services/llm-gateway/providers/ollama.ts
-async function handleOllamaRequest(req, baseUrlOverride) {
-  const baseUrl = (baseUrlOverride || process.env.OLLAMA_BASE_URL || "http://localhost:11434/v1").replace(/\/+$/, "");
-  const modelName = req.model.replace(/^ollama\//i, "");
-  const payload = __spreadProps(__spreadValues({}, req), {
-    model: modelName
-  });
-  const response = await fetch(`${baseUrl}/chat/completions`, {
-    method: "POST",
-    headers: {
-      "Content-Type": "application/json"
-    },
-    body: JSON.stringify(payload)
-  });
-  if (!response.ok) {
-    const errorText = await response.text();
-    throw new Error(`Ollama API Error (${response.status}): ${errorText}`);
-  }
-  if (req.stream && response.body) {
-    return { stream: response.body };
-  }
-  const json = await response.json();
-  return { response: json };
-}
-var init_ollama = __esm({
-  "../../services/llm-gateway/providers/ollama.ts"() {
-    "use strict";
-  }
-});
-
-// ../../services/llm-gateway/router.ts
-var router_exports = {};
-__export(router_exports, {
-  SUPPORTED_MODELS: () => SUPPORTED_MODELS,
-  dispatchChatCompletion: () => dispatchChatCompletion,
-  dispatchEmbedding: () => dispatchEmbedding,
-  resolveProvider: () => resolveProvider
-});
-function resolveProvider(model) {
-  const lower = model.toLowerCase();
-  if (lower.startsWith("huggingface/") || lower.startsWith("hf/")) {
-    return "huggingface";
-  }
-  if (lower.startsWith("ollama/")) {
-    return "ollama";
-  }
-  if (lower.startsWith("groq/") || lower.includes("llama") || lower.includes("qwen") || lower.includes("mixtral")) {
-    return "groq";
-  }
-  if (process.env.GROQ_API_KEY) return "groq";
-  if (process.env.OPENAI_API_KEY) return "openai";
-  if (process.env.GEMINI_API_KEY) return "gemini";
-  if (process.env.ANTHROPIC_API_KEY) return "anthropic";
-  if (process.env.HF_TOKEN || process.env.HUGGINGFACE_API_KEY) return "huggingface";
-  return "groq";
-}
-function cleanApiKeyOverride(apiKeyOverride) {
-  if (!apiKeyOverride) return void 0;
-  const key = apiKeyOverride.trim();
-  if (key === process.env.LITELLM_API_KEY || key === process.env.LITELLM_MASTER_KEY || key.startsWith("sk-retrivora") || !key.startsWith("gsk_") && !key.startsWith("AIza") && !key.startsWith("sk-ant-") && !key.startsWith("sk-proj-")) {
-    return void 0;
-  }
-  return key;
-}
-async function resolveUserGatewayConfig(apiKeyOverride) {
-  var _a2;
-  if (!apiKeyOverride || !process.env.NEXT_PUBLIC_SUPABASE_URL || !process.env.SUPABASE_SERVICE_ROLE_KEY) {
-    return {};
-  }
-  try {
-    const { createAdminClient } = await import("@/lib/supabase-server");
-    const supabase = createAdminClient();
-    const { data: licenseRecord } = await supabase.from("licenses").select("project_id, customer_name, tier").eq("license_key", apiKeyOverride.trim()).single();
-    if (!licenseRecord) {
-      return {};
-    }
-    const projectId = licenseRecord.project_id;
-    const { data: configs } = await supabase.from("gateway_config").select("project_id, default_model, provider_keys, is_active").in("project_id", [projectId, "global"]).eq("is_active", true);
-    if (!configs || configs.length === 0) {
-      return {};
-    }
-    const matchedConfig = configs.find((c) => c.project_id === projectId) || configs.find((c) => c.project_id === "global");
-    const customGroqKey = (_a2 = matchedConfig == null ? void 0 : matchedConfig.provider_keys) == null ? void 0 : _a2.groq;
-    const targetModel = matchedConfig == null ? void 0 : matchedConfig.default_model;
-    return { customGroqKey, targetModel };
-  } catch (err) {
-    console.warn("[LLM Gateway Router] Error resolving user gateway_config:", err.message);
-    return {};
-  }
-}
-async function dispatchChatCompletion(req, apiKeyOverride) {
-  const { customGroqKey, targetModel } = await resolveUserGatewayConfig(apiKeyOverride);
-  const effectiveKey = customGroqKey || cleanApiKeyOverride(apiKeyOverride);
-  const activeModel = req.model || targetModel || "llama-3.1-8b-instant";
-  const provider = resolveProvider(activeModel);
-  try {
-    switch (provider) {
-      case "groq":
-        return await handleGroqRequest(__spreadProps(__spreadValues({}, req), { model: activeModel }), effectiveKey);
-      case "openai":
-        return await handleOpenAIRequest(__spreadProps(__spreadValues({}, req), { model: activeModel }), effectiveKey);
-      case "gemini":
-        return await handleGeminiRequest(__spreadProps(__spreadValues({}, req), { model: activeModel }), effectiveKey);
-      case "anthropic":
-        return await handleAnthropicRequest(__spreadProps(__spreadValues({}, req), { model: activeModel }), effectiveKey);
-      case "ollama":
-        return await handleOllamaRequest(__spreadProps(__spreadValues({}, req), { model: activeModel }));
-      case "huggingface":
-        return await handleHuggingFaceChatRequest(__spreadProps(__spreadValues({}, req), { model: activeModel }), effectiveKey);
-      default:
-        throw new Error(`Unsupported LLM provider for model: ${activeModel}`);
-    }
-  } catch (error) {
-    console.error(`[LLM Gateway Router] Provider '${provider}' failed for model '${req.model}':`, {
-      message: error.message,
-      stack: error.stack
-    });
-    if (/rate[- ]?limit|429|exhausted/i.test(error.message)) {
-      const isGroqKey = process.env.GROQ_API_KEY || effectiveKey && effectiveKey.startsWith("gsk_");
-      const reqModelLower = req.model.toLowerCase();
-      if (isGroqKey && !reqModelLower.includes("llama-3.3-70b")) {
-        console.warn(`[LLM Gateway] Provider ${provider} rate limited (${error.message}). Falling back to secondary Groq model (llama-3.3-70b-versatile)...`);
-        try {
-          return await handleGroqRequest(__spreadProps(__spreadValues({}, req), { model: "llama-3.3-70b-versatile" }), effectiveKey);
-        } catch (fErr) {
-          console.error("[LLM Gateway] Fallback groq/llama-3.3-70b-versatile failed:", fErr.message);
-        }
-      }
-      const geminiKey = process.env.GEMINI_API_KEY || effectiveKey;
-      if (provider !== "gemini" && geminiKey && (geminiKey.startsWith("AIza") || !geminiKey.startsWith("AQ."))) {
-        console.warn(`[LLM Gateway] Provider ${provider} rate limited (${error.message}). Falling back to Google Gemini (gemini-2.5-flash)...`);
-        try {
-          return await handleGeminiRequest(__spreadProps(__spreadValues({}, req), { model: "gemini-2.5-flash" }), effectiveKey);
-        } catch (fErr) {
-          console.error("[LLM Gateway] Fallback gemini-2.5-flash failed:", fErr.message);
-        }
-      }
-      if (provider !== "huggingface" && (process.env.HF_TOKEN || process.env.HUGGINGFACE_API_KEY)) {
-        console.warn(`[LLM Gateway] Provider ${provider} rate limited (${error.message}). Falling back to Hugging Face (Qwen/Qwen2.5-Coder-32B-Instruct)...`);
-        try {
-          return await handleHuggingFaceChatRequest(__spreadProps(__spreadValues({}, req), { model: "Qwen/Qwen2.5-Coder-32B-Instruct" }), effectiveKey);
-        } catch (fErr) {
-          console.error("[LLM Gateway] Fallback HuggingFace Qwen failed:", fErr.message);
-        }
-      }
-      if (provider !== "openai" && process.env.OPENAI_API_KEY) {
-        console.warn(`[LLM Gateway] Provider ${provider} rate limited (${error.message}). Falling back to OpenAI (gpt-4o-mini)...`);
-        try {
-          return await handleOpenAIRequest(__spreadProps(__spreadValues({}, req), { model: "gpt-4o-mini" }), effectiveKey);
-        } catch (e) {
-        }
-      }
-      console.warn(`[LLM Gateway] Provider ${provider} rate limited (${error.message}). Falling back to local Ollama (llama3.2)...`);
-      return handleOllamaRequest(__spreadProps(__spreadValues({}, req), { model: "llama3.2" }));
-    }
-    throw error;
-  }
-}
-async function dispatchEmbedding(req, apiKeyOverride) {
-  const effectiveKey = cleanApiKeyOverride(apiKeyOverride);
-  if (req.model.startsWith("hf/") || req.model.startsWith("huggingface/") || req.model.includes("bge-")) {
-    try {
-      return await handleHuggingFaceEmbedding(req, effectiveKey);
-    } catch (err) {
-      console.warn("[LLM Gateway] HuggingFace explicit model embedding failed, falling back to default gateway chain:", err);
-    }
-  }
-  const geminiKey = process.env.GEMINI_API_KEY || effectiveKey;
-  const isGeminiKeyValid = Boolean(geminiKey && !geminiKey.startsWith("eyJ"));
-  if (isGeminiKeyValid) {
-    try {
-      return await handleGeminiEmbedding(req, effectiveKey);
-    } catch (err) {
-      console.warn("[LLM Gateway] Gemini embedding failed:", err);
-    }
-  }
-  if (process.env.OPENAI_API_KEY) {
-    try {
-      return await handleOpenAIEmbedding(req, effectiveKey);
-    } catch (err) {
-      console.warn("[LLM Gateway] OpenAI embedding failed:", err);
-    }
-  }
-  if (process.env.HUGGINGFACE_API_KEY || process.env.HF_TOKEN) {
-    try {
-      return await handleHuggingFaceEmbedding(req, effectiveKey);
-    } catch (err) {
-      console.warn("[LLM Gateway] HuggingFace fallback embedding failed:", err);
-    }
-  }
-  return handleGeminiEmbedding(req, effectiveKey);
-}
-var SUPPORTED_MODELS;
-var init_router = __esm({
-  "../../services/llm-gateway/router.ts"() {
-    "use strict";
-    init_groq();
-    init_openai();
-    init_gemini();
-    init_huggingface();
-    init_anthropic();
-    init_ollama();
-    SUPPORTED_MODELS = [
-      { id: "groq/llama-3.1-8b-instant", object: "model", created: 17e8, owned_by: "groq" },
-      { id: "llama-3.1-8b-instant", object: "model", created: 17e8, owned_by: "groq" },
-      { id: "groq/llama-3.3-70b-versatile", object: "model", created: 17e8, owned_by: "groq" },
-      { id: "llama-3.3-70b-versatile", object: "model", created: 17e8, owned_by: "groq" },
-      { id: "gpt-4o-mini", object: "model", created: 17e8, owned_by: "openai" },
-      { id: "openai/gpt-4o-mini", object: "model", created: 17e8, owned_by: "openai" },
-      { id: "gpt-4o", object: "model", created: 17e8, owned_by: "openai" },
-      { id: "gemini-2.5-flash", object: "model", created: 17e8, owned_by: "google" },
-      { id: "google/gemini-2.5-flash", object: "model", created: 17e8, owned_by: "google" },
-      { id: "claude-3-5-haiku-20241022", object: "model", created: 17e8, owned_by: "anthropic" },
-      { id: "anthropic/claude-3-5-haiku-20241022", object: "model", created: 17e8, owned_by: "anthropic" },
-      { id: "text-embedding-004", object: "model", created: 17e8, owned_by: "google" },
-      { id: "google/text-embedding-004", object: "model", created: 17e8, owned_by: "google" },
-      { id: "BAAI/bge-base-en-v1.5", object: "model", created: 17e8, owned_by: "huggingface" },
-      { id: "huggingface/BAAI/bge-base-en-v1.5", object: "model", created: 17e8, owned_by: "huggingface" },
-      { id: "groq/qwen3-32b", object: "model", created: 17e8, owned_by: "groq" },
-      { id: "qwen/qwen3-32b", object: "model", created: 17e8, owned_by: "groq" },
-      { id: "qwen/qwen3-embedding-8b", object: "model", created: 17e8, owned_by: "qwen" },
-      { id: "qwen3-embedding-8b", object: "model", created: 17e8, owned_by: "qwen" },
-      { id: "ollama/llama3.2", object: "model", created: 17e8, owned_by: "ollama" }
-    ];
-    if (typeof globalThis !== "undefined") {
-      const _g2 = globalThis;
-      _g2.__retrivoraDispatchChat = dispatchChatCompletion;
-      _g2.__retrivoraDispatchEmbedding = dispatchEmbedding;
-    }
-  }
-});
-
 // src/providers/vectordb/BaseVectorProvider.ts
 var BaseVectorProvider;
 var init_BaseVectorProvider = __esm({
@@ -800,8 +162,8 @@ var init_ConfigFetcher = __esm({
           process.env.NEXT_PUBLIC_RETRIVORA_CONTROL_PLANE_URL,
           "https://www.retrivora.com",
           "https://retrivora.com",
-          "http://localhost:3001",
-          "http://localhost:3000"
+          "http://localhost:3000",
+          "http://localhost:3001"
         ].filter(Boolean);
         for (const baseUrl of controlPlaneUrls) {
           try {
@@ -3356,7 +2718,7 @@ function getEnvConfig(env = process.env, base) {
     },
     telemetry: {
       enabled: telemetryEnabled,
-      url: (_Kb = (_Jb = (_Hb = readString(env, "TELEMETRY_URL")) != null ? _Hb : readString(env, "NEXT_PUBLIC_TELEMETRY_URL")) != null ? _Jb : (_Ib = base == null ? void 0 : base.telemetry) == null ? void 0 : _Ib.url) != null ? _Kb : process.env.NODE_ENV === "development" ? "http://localhost:3001/api/telemetry" : "https://retrivora.com/api/telemetry"
+      url: (_Kb = (_Jb = (_Hb = readString(env, "TELEMETRY_URL")) != null ? _Hb : readString(env, "NEXT_PUBLIC_TELEMETRY_URL")) != null ? _Jb : (_Ib = base == null ? void 0 : base.telemetry) == null ? void 0 : _Ib.url) != null ? _Kb : "https://www.retrivora.com/api/telemetry"
     }
   }, readString(env, "GRAPH_DB_PROVIDER") ? {
     graphDb: {
@@ -4605,9 +3967,7 @@ var OPENAI_BASE = {
 };
 var LLM_PROFILES = {
   "openai-compatible": OPENAI_BASE,
-  "litellm": __spreadProps(__spreadValues({}, OPENAI_BASE), {
-    chatPayloadTemplate: '{"model":"{{model}}","messages":{{messages}},"max_tokens":{{maxTokens}},"temperature":{{temperature}},"stream":false}'
-  }),
+  "litellm": __spreadValues({}, OPENAI_BASE),
   "anthropic-claude": {
     chatPath: "/v1/messages",
     responseExtractPath: "content[0].text",
@@ -4713,18 +4073,7 @@ ${context != null ? context : "None"}` },
       }
     } catch (httpErr) {
       const _g2 = globalThis;
-      let dispatch = _g2.__retrivoraDispatchChat;
-      if (typeof dispatch !== "function") {
-        try {
-          const gateway = await Promise.resolve().then(() => (init_router(), router_exports));
-          if (gateway == null ? void 0 : gateway.dispatchChatCompletion) {
-            _g2.__retrivoraDispatchChat = gateway.dispatchChatCompletion;
-            _g2.__retrivoraDispatchEmbedding = gateway.dispatchEmbedding;
-            dispatch = gateway.dispatchChatCompletion;
-          }
-        } catch (e) {
-        }
-      }
+      const dispatch = _g2.__retrivoraDispatchChat;
       if (typeof dispatch === "function") {
         const res = await dispatch({
           model: this.model,
@@ -4750,11 +4099,10 @@ ${context != null ? context : "None"}` },
   /**
    * Streaming chat using native fetch + ReadableStream.
    * Parses OpenAI-compatible SSE frames: `data: {...}\n\n`
-   * Works with vLLM, LMStudio, Together AI, Fireworks, and any OpenAI-compatible API.
    */
   chatStream(messages, context) {
     return __asyncGenerator(this, null, function* () {
-      var _a2, _b, _c, _d, _e;
+      var _a2, _b, _c, _d, _e, _f;
       const path2 = (_a2 = this.opts.chatPath) != null ? _a2 : "/chat/completions";
       const url = `${this.baseUrl.replace(/\/$/, "")}${path2}`;
       const extractPath = ((_b = this.opts.responseExtractPath) != null ? _b : "choices[0].message.content").replace("message.content", "delta.content");
@@ -4789,22 +4137,35 @@ ${context != null ? context : "None"}` },
           stream: true
         };
       }
-      const isSelfHost = this.baseUrl.includes("retrivora.com") || this.baseUrl.includes("localhost");
       let streamBody = null;
-      if (isSelfHost || Boolean(process.env.VERCEL)) {
-        const _g2 = globalThis;
-        let dispatch = _g2.__retrivoraDispatchChat;
-        if (typeof dispatch !== "function") {
-          try {
-            const gateway = yield new __await(Promise.resolve().then(() => (init_router(), router_exports)));
-            if (gateway == null ? void 0 : gateway.dispatchChatCompletion) {
-              _g2.__retrivoraDispatchChat = gateway.dispatchChatCompletion;
-              _g2.__retrivoraDispatchEmbedding = gateway.dispatchEmbedding;
-              dispatch = gateway.dispatchChatCompletion;
+      try {
+        const response = yield new __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 = yield new __await(response.json());
+            const text = (_c = resolvePath(json, extractPath)) != null ? _c : extractContent(json);
+            if (text && typeof text === "string") {
+              yield text;
+              return;
             }
-          } catch (e) {
+          } else if (response.body) {
+            streamBody = response.body;
           }
+        } else {
+          const errorText = yield new __await(response.text().catch(() => response.statusText));
+          console.warn(`[UniversalLLMAdapter] Remote HTTP stream returned ${response.status}: ${errorText}. Attempting in-process fallback...`);
         }
+      } catch (fetchErr) {
+        console.warn(`[UniversalLLMAdapter] Remote HTTP stream fetch warning: ${fetchErr == null ? void 0 : fetchErr.message}. Attempting in-process fallback...`);
+      }
+      if (!streamBody) {
+        const _g2 = globalThis;
+        const dispatch = _g2.__retrivoraDispatchChat;
         if (typeof dispatch === "function") {
           try {
             const res = yield new __await(dispatch({
@@ -4822,29 +4183,14 @@ ${context != null ? context : "None"}` },
                 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 ${this.baseUrl} are disabled on Vercel to prevent HTTP 508 Loop Detected.`);
         }
       }
       if (!streamBody) {
-        const response = yield new __await(fetch(url, {
-          method: "POST",
-          headers: this.resolvedHeaders,
-          body: JSON.stringify(payload)
-        }));
-        if (!response.ok) {
-          const errorText = yield new __await(response.text().catch(() => response.statusText));
-          throw new Error(`[UniversalLLMAdapter] Streaming request failed (${response.status}): ${errorText}`);
-        }
-        if (!response.body) {
-          throw new Error("[UniversalLLMAdapter] Response body is null \u2014 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();
       const decoder = new TextDecoder("utf-8");
@@ -4855,14 +4201,14 @@ ${context != null ? context : "None"}` },
           if (done) break;
           buffer += decoder.decode(value, { stream: true });
           const lines = buffer.split("\n");
-          buffer = (_c = lines.pop()) != null ? _c : "";
+          buffer = (_d = lines.pop()) != null ? _d : "";
           for (const line of lines) {
             const trimmed = line.trim();
             if (!trimmed || trimmed === "data: [DONE]") continue;
             if (!trimmed.startsWith("data:")) continue;
             try {
               const json = JSON.parse(trimmed.slice(5).trim());
-              const text = (_d = resolvePath(json, extractPath)) != null ? _d : extractContent(json);
+              const text = (_e = resolvePath(json, extractPath)) != null ? _e : extractContent(json);
               if (text && typeof text === "string") yield text;
             } catch (e) {
             }
@@ -4872,7 +4218,7 @@ ${context != null ? context : "None"}` },
           const jsonStr = buffer.replace(/^data:\s*/, "").trim();
           try {
             const json = JSON.parse(jsonStr);
-            const text = (_e = resolvePath(json, extractPath)) != null ? _e : extractContent(json);
+            const text = (_f = resolvePath(json, extractPath)) != null ? _f : extractContent(json);
             if (text && typeof text === "string") yield text;
           } catch (e) {
           }
@@ -4885,74 +4231,42 @@ ${context != null ? context : "None"}` },
   async embed(text) {
     var _a2, _b, _c, _d, _e;
     const path2 = (_a2 = this.opts.embedPath) != null ? _a2 : "/embeddings";
-    let payload;
-    if (this.opts.embedPayloadTemplate) {
-      payload = buildPayload(this.opts.embedPayloadTemplate, {
-        model: this.model,
-        input: text
-      });
-    } else {
-      payload = {
-        model: this.model,
-        input: text
-      };
-    }
+    const payload = this.opts.embedPayloadTemplate ? buildPayload(this.opts.embedPayloadTemplate, { input: text, model: this.model }) : { input: text, model: this.model };
     try {
       const { data: data2 } = await this.http.post(path2, payload);
       const extractPath2 = (_b = this.opts.embedExtractPath) != null ? _b : "data[0].embedding";
       const vector2 = resolvePath(data2, extractPath2);
-      if (Array.isArray(vector2)) {
-        return vector2;
-      }
-    } catch (httpErr) {
-      console.warn(`[UniversalLLMAdapter] Direct HTTP embedding POST to ${this.baseUrl}${path2} failed (${httpErr.message}). Attempting in-process gateway dispatch fallback...`);
+      if (Array.isArray(vector2)) return vector2;
+    } catch (e) {
       const _g2 = globalThis;
-      let dispatch = _g2.__retrivoraDispatchEmbedding;
-      if (typeof dispatch !== "function") {
-        try {
-          const gateway = await Promise.resolve().then(() => (init_router(), router_exports));
-          if (gateway == null ? void 0 : gateway.dispatchEmbedding) {
-            _g2.__retrivoraDispatchChat = gateway.dispatchChatCompletion;
-            _g2.__retrivoraDispatchEmbedding = gateway.dispatchEmbedding;
-            dispatch = gateway.dispatchEmbedding;
-          }
-        } catch (e) {
-        }
-      }
-      if (typeof dispatch === "function") {
-        const res = await dispatch({
-          model: this.model,
-          input: text
-        }, this.apiKey);
+      const dispatchEmbed = _g2.__retrivoraDispatchEmbedding;
+      if (typeof dispatchEmbed === "function") {
+        const res = await dispatchEmbed({ input: text, model: this.model }, this.apiKey);
         if ((_d = (_c = res == null ? void 0 : res.data) == null ? void 0 : _c[0]) == null ? void 0 : _d.embedding) {
           return res.data[0].embedding;
         }
       }
-      throw httpErr;
     }
     const { data } = await this.http.post(path2, payload);
     const extractPath = (_e = this.opts.embedExtractPath) != null ? _e : "data[0].embedding";
     const vector = resolvePath(data, extractPath);
     if (!Array.isArray(vector)) {
-      throw new Error(`[UniversalLLMAdapter] Expected a number array at '${extractPath}' for embeddings.`);
+      throw new Error(`[UniversalLLMAdapter] Could not extract embedding vector from path '${extractPath}' in response.`);
     }
     return vector;
   }
   async batchEmbed(texts) {
-    const vectors = [];
+    const results = [];
     for (const text of texts) {
-      vectors.push(await this.embed(text));
+      results.push(await this.embed(text));
     }
-    return vectors;
+    return results;
   }
   async ping() {
     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 (e) {
       return false;
     }
   }
@@ -5358,7 +4672,7 @@ var ConfigValidator = class {
 // package.json
 var package_default = {
   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 \u2014 generic vector DB + LLM provider, embeddable or standalone.",
   author: "Abhinav Alkuchi",
   license: "UNLICENSED",
@@ -5395,6 +4709,8 @@ var package_default = {
       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",
@@ -7563,19 +6879,6 @@ var UITransformer = class _UITransformer {
     if (!retrievedData || retrievedData.length === 0) {
       return this.createTextResponse("No data available", "No relevant data found for your query.");
     }
-    console.log("[UITransformer.transform] Processing retrievedData:", {
-      userQuery,
-      retrievedCount: retrievedData.length,
-      sample: retrievedData.slice(0, 2).map((item) => {
-        var _a3;
-        return {
-          id: item == null ? void 0 : item.id,
-          contentType: typeof (item == null ? void 0 : item.content),
-          contentSnippet: String((_a3 = item == null ? void 0 : item.content) != null ? _a3 : "").substring(0, 150),
-          metadataKeys: Object.keys((item == null ? void 0 : item.metadata) || {})
-        };
-      })
-    });
     const resolvedIntent = intent != null ? intent : this.detectIntentHeuristic(userQuery);
     const filteredData = resolvedIntent.filterInStockOnly ? retrievedData.filter((item) => this.determineStockStatus(item)) : retrievedData;
     const profile = this.profileData(filteredData);
@@ -8915,59 +8218,70 @@ RULES:
 var SchemaMapper = class {
   /**
    * Trains the plugin on a set of keys.
-   * This is done once per schema and cached.
+   * Results are cached in-process; concurrent calls for the same key set are
+   * deduplicated to a single LLM request.
    */
   static async train(llm, projectId, keys) {
-    const cacheKey = `${projectId}:${keys.sort().join(",")}`;
+    const cacheKey = `${projectId}:${keys.slice().sort().join(",")}`;
     if (this.cache.has(cacheKey)) {
       return this.cache.get(cacheKey);
     }
-    console.log(`[SchemaMapper] \u{1F9E0} Training AI on new schema keys: ${keys.join(", ")}`);
+    if (this.pending.has(cacheKey)) {
+      return this.pending.get(cacheKey);
+    }
+    const promise = this._doTrain(llm, cacheKey, keys);
+    this.pending.set(cacheKey, promise);
+    promise.finally(() => this.pending.delete(cacheKey));
+    return promise;
+  }
+  static async _doTrain(llm, cacheKey, keys) {
+    var _a2, _b, _c, _d, _e, _f, _g2, _h, _i, _j, _k;
+    console.log(`[SchemaMapper] \u{1F9E0} Training on new schema keys: ${keys.join(", ")}`);
     const propertyList = Object.entries(this.TARGET_PROPERTIES).map(([prop, desc]) => `- ${prop} (${desc})`).join("\n");
-    const prompt = `
-Given these metadata keys from a database: [${keys.join(", ")}]
+    const messages = [
+      { role: "system", content: "You are a database schema expert. Respond ONLY with valid JSON." },
+      {
+        role: "user",
+        content: `Keys: [${keys.join(", ")}]
 
-Identify which keys best correspond to these standard UI properties:
+Map to:
 ${propertyList}
 
-Return ONLY a valid JSON object where the keys are the UI properties and the values are the EXACT matching database keys from the list above. 
-If no good match is found for a property, omit it.
-
-Example:
-{
-  "name": "Title",
-  "price": "Variant Price",
-  "brand": "Vendor",
-  "image": "Image Src",
-  "stock": "Variant Inventory Qty"
-}
-`;
+Return a JSON object like {"name":"Title","price":"Price",...}. Omit unmapped properties.`
+      }
+    ];
     try {
-      const response = await llm.chat(
-        [
-          { role: "system", content: "You are a database schema expert. You ONLY respond with valid JSON." },
-          { role: "user", content: prompt }
-        ],
-        ""
-      );
-      const startIdx = response.indexOf("{");
-      if (startIdx !== -1) {
-        let braceCount = 0;
-        let endIdx = -1;
-        for (let i = startIdx; i < response.length; i++) {
-          if (response[i] === "{") braceCount++;
-          else if (response[i] === "}") {
-            braceCount--;
-            if (braceCount === 0) {
-              endIdx = i;
-              break;
-            }
-          }
+      const baseUrl = (_a2 = llm == null ? void 0 : llm.baseUrl) != null ? _a2 : "";
+      const apiKey = (_b = llm == null ? void 0 : llm.apiKey) != null ? _b : "";
+      const model = (_c = llm == null ? void 0 : llm.model) != null ? _c : "llama-3.1-8b-instant";
+      let responseText;
+      if (baseUrl) {
+        const endpoint = `${baseUrl.replace(/\/$/, "")}/chat/completions`;
+        const headers = { "Content-Type": "application/json" };
+        if (apiKey) headers["Authorization"] = `Bearer ${apiKey}`;
+        const res = await fetch(endpoint, {
+          method: "POST",
+          headers,
+          body: JSON.stringify({
+            model,
+            messages,
+            max_tokens: 256,
+            // Only needs a small JSON object back
+            temperature: 0,
+            stream: false
+          }),
+          signal: AbortSignal.timeout(8e3)
+        });
+        if (res.ok) {
+          const data = await res.json();
+          responseText = (_k = (_j = (_f = (_e = (_d = data == null ? void 0 : data.choices) == null ? void 0 : _d[0]) == null ? void 0 : _e.message) == null ? void 0 : _f.content) != null ? _j : (_i = (_h = (_g2 = data == null ? void 0 : data.choices) == null ? void 0 : _g2[0]) == null ? void 0 : _h.delta) == null ? void 0 : _i.content) != null ? _k : void 0;
+        } else {
+          console.warn(`[SchemaMapper] LLM returned ${res.status}, falling back to heuristics.`);
         }
-        if (endIdx !== -1) {
-          const jsonContent = response.substring(startIdx, endIdx + 1);
-          const cleanJson = this.sanitizeJson(jsonContent);
-          const mapping = JSON.parse(cleanJson);
+      }
+      if (responseText) {
+        const mapping = this._parseJson(responseText);
+        if (mapping) {
           this.cache.set(cacheKey, mapping);
           return mapping;
         }
@@ -8975,23 +8289,42 @@ Example:
     } catch (error) {
       console.warn("[SchemaMapper] AI training failed, falling back to heuristics:", error);
     }
+    this.cache.set(cacheKey, {});
     return {};
   }
-  /**
-   * Forgiving JSON parser that fixes common AI formatting mistakes.
-   */
-  static sanitizeJson(s) {
+  static _parseJson(text) {
+    const startIdx = text.indexOf("{");
+    if (startIdx === -1) return null;
+    let braceCount = 0;
+    let endIdx = -1;
+    for (let i = startIdx; i < text.length; i++) {
+      if (text[i] === "{") braceCount++;
+      else if (text[i] === "}") {
+        braceCount--;
+        if (braceCount === 0) {
+          endIdx = i;
+          break;
+        }
+      }
+    }
+    if (endIdx === -1) return null;
+    try {
+      return JSON.parse(this._sanitizeJson(text.substring(startIdx, endIdx + 1)));
+    } catch (e) {
+      return null;
+    }
+  }
+  static _sanitizeJson(s) {
     return s.replace(/\/\*[\s\S]*?\*\/|([^:]|^)\/\/.*$/gm, "").replace(/[\u201C\u201D\u2018\u2019]/g, '"').replace(/'/g, '"').replace(/([{,]\s*)([a-zA-Z0-9_]+?)\s*:/g, '$1"$2":').replace(/,\s*([}\]])/g, "$1").trim();
   }
   static getCached(projectId, keys) {
-    const cacheKey = `${projectId}:${keys.sort().join(",")}`;
+    const cacheKey = `${projectId}:${keys.slice().sort().join(",")}`;
     return this.cache.get(cacheKey);
   }
 };
 SchemaMapper.cache = /* @__PURE__ */ new Map();
-/**
- * Descriptions of standard UI properties to help the AI map fields accurately.
- */
+/** In-flight deduplication — prevents concurrent identical LLM calls. */
+SchemaMapper.pending = /* @__PURE__ */ new Map();
 SchemaMapper.TARGET_PROPERTIES = {
   name: "The primary title, name, or label of the item",
   price: "The numeric cost, price, MSRP, or amount",
@@ -9463,19 +8796,6 @@ ${m.content}`).join("\n\n---\n\n");
         const wantsExhaustiveList = true;
         const retrievalLimit = Math.max(topK * 50, 1e3);
         const rawSources = ((strategyResult === "vector" || strategyResult === "both") && queryVector && queryVector.length > 0 ? yield new __await(this.vectorDB.query(queryVector, retrievalLimit, ns, filter)) : []).filter((s) => Boolean(s && typeof s === "object"));
-        console.log("[Pipeline] Raw vectorDB.query response:", {
-          rawCount: rawSources.length,
-          sample: rawSources.slice(0, 2).map((s) => {
-            var _a3;
-            return {
-              id: s == null ? void 0 : s.id,
-              score: s == null ? void 0 : s.score,
-              contentType: typeof (s == null ? void 0 : s.content),
-              contentSnippet: String((_a3 = s == null ? void 0 : s.content) != null ? _a3 : "").substring(0, 150),
-              metadataKeys: Object.keys((s == null ? void 0 : s.metadata) || {})
-            };
-          })
-        });
         const retrieveEnd = performance.now();
         const embedMs = retrieveEnd - embedStart;
         const retrieveMs = retrieveEnd - embedStart;
@@ -9504,21 +8824,6 @@ ${(_a3 = m == null ? void 0 : m.content) != null ? _a3 : ""}`;
           var _a3, _b2;
           return ((_a3 = b == null ? void 0 : b.score) != null ? _a3 : 0) - ((_b2 = a == null ? void 0 : a.score) != null ? _b2 : 0);
         });
-        console.log("[Pipeline] Final sources for prompt & UI:", {
-          count: sources.length,
-          totalRawCount: rawSources.length,
-          sources: sources.slice(0, 10).map((s, idx) => {
-            var _a3;
-            return {
-              rank: idx + 1,
-              id: s == null ? void 0 : s.id,
-              score: s == null ? void 0 : s.score,
-              contentType: typeof (s == null ? void 0 : s.content),
-              contentSnippet: String((_a3 = s == null ? void 0 : s.content) != null ? _a3 : "").substring(0, 150),
-              metadata: s == null ? void 0 : s.metadata
-            };
-          })
-        });
         if (graphData && graphData.nodes.length > 0) {
           const graphContext = graphData.nodes.map(
             (n) => `Entity: ${n.label} (${n.id})${n.properties ? " - " + JSON.stringify(n.properties) : ""}`
@@ -9760,9 +9065,9 @@ ${context}`;
         const trace = buildTrace(hallucinationResult);
         const isTelemetryActive = (_A = (_z = this.config.telemetry) == null ? void 0 : _z.enabled) != null ? _A : Boolean(this.config.licenseKey);
         if (isTelemetryActive) {
-          const defaultUrl = process.env.NODE_ENV === "development" && !process.env.TELEMETRY_URL && !process.env.NEXT_PUBLIC_TELEMETRY_URL ? "http://localhost:3001/api/telemetry" : "https://retrivora.com/api/telemetry";
+          const defaultUrl = "https://www.retrivora.com/api/telemetry";
           const telemetryUrl = ((_B = this.config.telemetry) == null ? void 0 : _B.url) || process.env.TELEMETRY_URL || process.env.NEXT_PUBLIC_TELEMETRY_URL || defaultUrl;
-          const absoluteUrl = telemetryUrl.startsWith("http") ? telemetryUrl : (process.env.NEXT_PUBLIC_APP_URL || `http://localhost:${process.env.PORT || 3e3}`) + telemetryUrl;
+          const absoluteUrl = telemetryUrl.startsWith("http") ? telemetryUrl : "https://www.retrivora.com" + (telemetryUrl.startsWith("/") ? telemetryUrl : "/" + telemetryUrl);
           (async () => {
             var _a3, _b2, _c2, _d2, _e2;
             try {
@@ -10299,8 +9604,8 @@ var VectorPlugin = class {
     }
     try {
       return await this.pipeline.getSuggestions(query, namespace);
-    } catch (err) {
-      throw wrapError(err, "RETRIEVAL_FAILED");
+    } catch (e) {
+      return [];
     }
   }
 };
@@ -10897,15 +10202,14 @@ function getOrCreatePlugin(configOrPlugin) {
   return _g[cacheKey];
 }
 function reportTelemetry(req, plugin, action, status, details, trace) {
-  var _a2, _b, _c, _d, _e, _f, _g2, _h;
+  var _a2, _b, _c, _d, _e, _f, _g2;
   try {
     const config = plugin.getConfig();
     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 = (_a2 = telemetryConfig == null ? void 0 : telemetryConfig.enabled) != null ? _a2 : 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 telemetryUrl = (telemetryConfig == null ? void 0 : telemetryConfig.url) || process.env.TELEMETRY_URL || process.env.NEXT_PUBLIC_TELEMETRY_URL || defaultUrl;
     const host = req.headers.get("host") || "localhost";
+    const defaultUrl = "https://www.retrivora.com/api/telemetry";
+    const telemetryUrl = (telemetryConfig == null ? void 0 : telemetryConfig.url) || process.env.TELEMETRY_URL || process.env.NEXT_PUBLIC_TELEMETRY_URL || defaultUrl;
     const userAgent = req.headers.get("user-agent") || `Retrivora-SDK/${SDK_VERSION}`;
     let absoluteUrl = telemetryUrl;
     if (!telemetryUrl.startsWith("http")) {
@@ -10913,11 +10217,11 @@ function reportTelemetry(req, plugin, action, status, details, trace) {
       absoluteUrl = `${proto}://${host}${telemetryUrl}`;
     }
     const projectId = config.projectId || "default";
-    const model = (trace == null ? void 0 : trace.model) || ((_b = config.llm) == null ? void 0 : _b.model) || ((_c = config.embedding) == null ? void 0 : _c.model) || "llama-3.1-8b-instant";
-    const provider = (trace == null ? void 0 : trace.provider) || ((_d = config.llm) == null ? void 0 : _d.provider) || ((_e = config.embedding) == null ? void 0 : _e.provider) || "groq";
-    const tokens = Number(((_f = trace == null ? void 0 : trace.tokens) == null ? void 0 : _f.totalTokens) || (trace == null ? void 0 : trace.totalTokens) || 0);
-    const costUsd = Number(((_g2 = trace == null ? void 0 : trace.tokens) == null ? void 0 : _g2.estimatedCostUsd) || (trace == null ? void 0 : trace.costUsd) || 0);
-    const latencyMs = Number(((_h = trace == null ? void 0 : trace.latency) == null ? void 0 : _h.totalMs) || (trace == null ? void 0 : trace.latencyMs) || 0);
+    const model = (trace == null ? void 0 : trace.model) || ((_a2 = config.llm) == null ? void 0 : _a2.model) || ((_b = config.embedding) == null ? void 0 : _b.model) || "llama-3.1-8b-instant";
+    const provider = (trace == null ? void 0 : trace.provider) || ((_c = config.llm) == null ? void 0 : _c.provider) || ((_d = config.embedding) == null ? void 0 : _d.provider) || "groq";
+    const tokens = Number(((_e = trace == null ? void 0 : trace.tokens) == null ? void 0 : _e.totalTokens) || (trace == null ? void 0 : trace.totalTokens) || 0);
+    const costUsd = Number(((_f = trace == null ? void 0 : trace.tokens) == null ? void 0 : _f.estimatedCostUsd) || (trace == null ? void 0 : trace.costUsd) || 0);
+    const latencyMs = Number(((_g2 = trace == null ? void 0 : trace.latency) == null ? void 0 : _g2.totalMs) || (trace == null ? void 0 : trace.latencyMs) || 0);
     const payload = {
       trace,
       licenseKey,
@@ -10994,7 +10298,7 @@ function createChatHandler(configOrPlugin, options) {
   const plugin = getOrCreatePlugin(configOrPlugin);
   const storage = new DatabaseStorage(plugin.getConfig());
   const onAuthorize = options == null ? void 0 : options.onAuthorize;
-  return async function POST(req) {
+  return async function POST(req, context) {
     var _a2, _b, _c, _d;
     const authResult = await checkAuth(req, onAuthorize);
     if (authResult) return authResult;
@@ -11050,7 +10354,7 @@ function createStreamHandler(configOrPlugin, options) {
   const plugin = getOrCreatePlugin(configOrPlugin);
   const storage = new DatabaseStorage(plugin.getConfig());
   const onAuthorize = options == null ? void 0 : options.onAuthorize;
-  return async function POST(req) {
+  return async function POST(req, context) {
     var _a2;
     const authResult = await checkAuth(req, onAuthorize);
     if (authResult) return authResult;
@@ -11200,7 +10504,7 @@ function createStreamHandler(configOrPlugin, options) {
 function createIngestHandler(configOrPlugin, options) {
   const plugin = getOrCreatePlugin(configOrPlugin);
   const onAuthorize = options == null ? void 0 : options.onAuthorize;
-  return async function POST(req) {
+  return async function POST(req, context) {
     const authResult = await checkAuth(req, onAuthorize);
     if (authResult) return authResult;
     try {
@@ -11244,7 +10548,7 @@ function createHealthHandler(configOrPlugin, options) {
 function createUploadHandler(configOrPlugin, options) {
   const plugin = getOrCreatePlugin(configOrPlugin);
   const onAuthorize = options == null ? void 0 : options.onAuthorize;
-  return async function POST(req) {
+  return async function POST(req, context) {
     const authResult = await checkAuth(req, onAuthorize);
     if (authResult) return authResult;
     try {