From c280683a9d3367708d928ff0ab70d196b2e2422e Mon Sep 17 00:00:00 2001
From: techjarves <usaramwasi99@Gmail.com>
Date: Mon, 22 Jun 2026 10:34:02 +0530
Subject: [PATCH]  implement modular web search capabilities with provider
 support, content fetching, and file-based caching.

---
 .gitignore                               |   1 +
 app/frontend/src/components/TextChat.jsx |  65 ++++++++-
 app/frontend/src/services/api.js         |  30 ++++-
 scripts/search/cache.js                  |  37 ++++++
 scripts/search/content.js                | 159 +++++++++++++++++++++++
 scripts/search/core.js                   |  88 +++++++++++++
 scripts/search/providers.js              | 106 +++++++++++++++
 scripts/server/serve.cjs                 |  99 +++++++++++++-
 8 files changed, 580 insertions(+), 5 deletions(-)
 create mode 100644 scripts/search/cache.js
 create mode 100644 scripts/search/content.js
 create mode 100644 scripts/search/core.js
 create mode 100644 scripts/search/providers.js

diff --git a/.gitignore b/.gitignore
index 1cb9fa2..7cb43f3 100644
--- a/.gitignore
+++ b/.gitignore
@@ -14,6 +14,7 @@ app/tools/
 
 # Generated files
 app/config/
+app/cache/
 app/dist/
 app/outputs/
 app/transcriptions/
diff --git a/app/frontend/src/components/TextChat.jsx b/app/frontend/src/components/TextChat.jsx
index eb576f1..2fcf7d3 100644
--- a/app/frontend/src/components/TextChat.jsx
+++ b/app/frontend/src/components/TextChat.jsx
@@ -1,5 +1,5 @@
 import React, { useCallback, useEffect, useLayoutEffect, useRef, useState } from "react";
-import { Bot, LoaderCircle, Send, Trash2, Square, History, Paperclip, X, ChevronDown } from "lucide-react";
+import { Bot, LoaderCircle, Send, Trash2, Square, History, Paperclip, X, ChevronDown, Globe2 } from "lucide-react";
 import {
   getDownloadProgress,
   getLlmStatus,
@@ -149,6 +149,8 @@ function TextChat({
   const [messages, setMessages] = useState([]);
   const [input, setInput] = useState("");
   const [isBusy, setIsBusy] = useState(false);
+  const [useWebSearch, setUseWebSearch] = useState(false);
+  const [webTimeFilter, setWebTimeFilter] = useState("any");
   const [loadingModel, setLoadingModel] = useState(null);
   const [tokenUsage, setTokenUsage] = useState({
     prompt_tokens: 0,
@@ -639,7 +641,7 @@ function TextChat({
     setMessages([...nextMessages, {
       role: "assistant",
       content: "",
-      generationStats: { status: "starting", tokens: 0, tokensPerSecond: 0, seconds: 0, vision: requestHasImage },
+      generationStats: { status: "starting", tokens: 0, tokensPerSecond: 0, seconds: 0, vision: requestHasImage, web: useWebSearch },
     }]);
 
     const controller = new AbortController();
@@ -691,6 +693,8 @@ function TextChat({
         repeatPenalty: textSettings?.repeatPenalty,
         seed: textSettings?.seed,
         enableThinking: thinkingEnabled,
+        useWeb: useWebSearch,
+        timeFilter: webTimeFilter,
         signal: controller.signal,
       }, (_token, fullText, _reasoningToken, fullReasoning) => {
         const now = performance.now();
@@ -790,6 +794,7 @@ function TextChat({
         reasoning: processed.reasoning,
         thinkingDuration: finalThinkingDuration,
         generationStats,
+        webSources: response.webSources || [],
       }];
       setMessages(finalMessages);
       saveConversationState(convId, finalMessages, selectedModel);
@@ -1073,6 +1078,34 @@ function TextChat({
                           )}
                         </div>
                       )}
+                      {message.role === "assistant" && Array.isArray(message.webSources) && message.webSources.length > 0 && (
+                        <div style={{
+                          marginTop: "8px",
+                          padding: "10px 12px",
+                          border: "1px solid var(--border-color)",
+                          borderRadius: "8px",
+                          background: "var(--md-sys-color-surface-container)",
+                          fontSize: "0.8rem",
+                        }}>
+                          <div style={{ display: "flex", alignItems: "center", gap: "6px", fontWeight: 700, marginBottom: "6px" }}>
+                            <Globe2 size={13} />
+                            <span>Sources</span>
+                          </div>
+                          <div style={{ display: "flex", flexDirection: "column", gap: "5px" }}>
+                            {message.webSources.map((source, sourceIndex) => (
+                              <a
+                                key={`${source.url}-${sourceIndex}`}
+                                href={source.url}
+                                target="_blank"
+                                rel="noreferrer"
+                                style={{ color: "var(--md-sys-color-primary)", textDecoration: "none", overflowWrap: "anywhere" }}
+                              >
+                                [{source.index || sourceIndex + 1}] {source.title || source.url}
+                              </a>
+                            ))}
+                          </div>
+                        </div>
+                      )}
 
                       {/* Generation stats pill */}
                       {message.role === "assistant" && message.generationStats && !message.error && (
@@ -1084,7 +1117,7 @@ function TextChat({
                           )}
                           <div className={`chat-generation-stats ${message.generationStats.status}`}>
                           {message.generationStats.status === "starting" ? (
-                            <><LoaderCircle size={11} className="progress-spinner" /> {message.generationStats.vision ? "Processing image..." : "Waiting for first token..."}</>
+                            <><LoaderCircle size={11} className="progress-spinner" /> {message.generationStats.web ? "Searching web..." : message.generationStats.vision ? "Processing image..." : "Waiting for first token..."}</>
                           ) : message.generationStats.status === "streaming" ? (
                             <><span style={{ opacity: 0.7 }}>⚡</span> {message.generationStats.tokensPerSecond.toFixed(1)} tok/s</>
                           ) : (
@@ -1150,6 +1183,32 @@ function TextChat({
             >
               <Paperclip size={17} />
             </button>
+            <button
+              className={`chat-composer-deepthink-btn ${useWebSearch ? "active" : ""}`}
+              onClick={() => setUseWebSearch((value) => !value)}
+              disabled={!status.ready || isBusy}
+              title={useWebSearch ? "Disable web search" : "Enable web search"}
+              style={{ marginBottom: "2px" }}
+            >
+              <Globe2 size={14} />
+              <span>Web</span>
+            </button>
+            {useWebSearch && (
+              <select
+                className="m3-input"
+                value={webTimeFilter}
+                onChange={(event) => setWebTimeFilter(event.target.value)}
+                disabled={isBusy}
+                title="Search recency"
+                style={{ width: "116px", height: "38px", marginBottom: "2px", fontSize: "0.78rem" }}
+              >
+                <option value="any">Any time</option>
+                <option value="day">Past day</option>
+                <option value="week">Past week</option>
+                <option value="month">Past month</option>
+                <option value="year">Past year</option>
+              </select>
+            )}
 
             <div className="chat-composer-middle" style={{ flex: 1, display: "flex", minWidth: 0 }}>
               <textarea
diff --git a/app/frontend/src/services/api.js b/app/frontend/src/services/api.js
index 049e658..062557a 100644
--- a/app/frontend/src/services/api.js
+++ b/app/frontend/src/services/api.js
@@ -412,6 +412,8 @@ export async function chatWithLlm(messages, options = {}) {
       presence_penalty: options.presencePenalty,
       seed: options.seed,
       stop: options.stop,
+      useWeb: options.useWeb === true,
+      timeFilter: options.timeFilter || "any",
     }),
   });
   const data = await readJsonResponse(res, "The local text model returned an invalid response.");
@@ -441,6 +443,8 @@ export async function streamChatWithLlm(messages, options = {}, onToken = () =>
       presence_penalty: options.presencePenalty,
       seed: options.seed,
       stop: options.stop,
+      useWeb: options.useWeb === true,
+      timeFilter: options.timeFilter || "any",
     }),
   };
   if (options.signal) {
@@ -462,6 +466,7 @@ export async function streamChatWithLlm(messages, options = {}, onToken = () =>
   let usage = null;
   let timings = null;
   let finishReason = null;
+  let webSources = [];
 
   const normalizeTextDelta = (value) => {
     if (typeof value === "string") return value;
@@ -490,7 +495,16 @@ export async function streamChatWithLlm(messages, options = {}, onToken = () =>
         .trim();
       if (!data || data === "[DONE]") return data === "[DONE]";
 
+      const eventName = eventText
+        .split(/\r?\n/)
+        .find((line) => line.startsWith("event:"))
+        ?.slice(6)
+        .trim();
       const parsed = JSON.parse(data);
+      if (eventName === "web_sources") {
+        webSources = parsed.sources || [];
+        return false;
+      }
       const choice = parsed.choices?.[0];
       const token = normalizeTextDelta(
         choice?.delta?.content ??
@@ -540,7 +554,21 @@ export async function streamChatWithLlm(messages, options = {}, onToken = () =>
 
   if (!finished && buffer.trim()) consumeEvent(buffer);
   if (!content && !reasoningContent) throw new Error("The text model returned an empty streamed response.");
-  return { content, reasoningContent, usage, timings, finishReason };
+  return { content, reasoningContent, usage, timings, finishReason, webSources };
+}
+
+export async function searchWeb(query, options = {}) {
+  const res = await fetch("/api/web-search", {
+    method: "POST",
+    headers: { "Content-Type": "application/json" },
+    body: JSON.stringify({
+      query,
+      timeFilter: options.timeFilter || "any",
+      resultLimit: options.resultLimit,
+      fetchLimit: options.fetchLimit,
+    }),
+  });
+  return await readJsonResponse(res, "The local server returned an invalid web search response.");
 }
 
 export async function downloadLlmModel(url, filename = null, companion = null) {
diff --git a/scripts/search/cache.js b/scripts/search/cache.js
new file mode 100644
index 0000000..fec832f
--- /dev/null
+++ b/scripts/search/cache.js
@@ -0,0 +1,37 @@
+const fs = require("fs");
+const path = require("path");
+const crypto = require("crypto");
+
+const DEFAULT_TTL_MS = 30 * 60 * 1000;
+
+function ensureDir(dir) {
+  fs.mkdirSync(dir, { recursive: true });
+}
+
+function keyFor(value) {
+  return crypto.createHash("sha256").update(String(value || "")).digest("hex");
+}
+
+function readCache(cacheDir, key, ttlMs = DEFAULT_TTL_MS) {
+  try {
+    const file = path.join(cacheDir, `${keyFor(key)}.json`);
+    const stat = fs.statSync(file);
+    if (Date.now() - stat.mtimeMs > ttlMs) return null;
+    return JSON.parse(fs.readFileSync(file, "utf8"));
+  } catch (_) {
+    return null;
+  }
+}
+
+function writeCache(cacheDir, key, value) {
+  try {
+    ensureDir(cacheDir);
+    const file = path.join(cacheDir, `${keyFor(key)}.json`);
+    fs.writeFileSync(file, JSON.stringify(value, null, 2), "utf8");
+  } catch (_) {}
+}
+
+module.exports = {
+  readCache,
+  writeCache,
+};
diff --git a/scripts/search/content.js b/scripts/search/content.js
new file mode 100644
index 0000000..e6d5b60
--- /dev/null
+++ b/scripts/search/content.js
@@ -0,0 +1,159 @@
+const http = require("http");
+const https = require("https");
+const dns = require("dns").promises;
+const net = require("net");
+const zlib = require("zlib");
+
+const USER_AGENT = "Mozilla/5.0 (Windows NT 10.0; Win64; x64) LocalAIStudio/1.0";
+
+function decodeHtml(value) {
+  return String(value || "")
+    .replace(/&nbsp;/g, " ")
+    .replace(/&amp;/g, "&")
+    .replace(/&quot;/g, '"')
+    .replace(/&#39;|&apos;/g, "'")
+    .replace(/&lt;/g, "<")
+    .replace(/&gt;/g, ">")
+    .replace(/&#x([0-9a-f]+);/gi, (_, hex) => String.fromCodePoint(parseInt(hex, 16)))
+    .replace(/&#(\d+);/g, (_, num) => String.fromCodePoint(parseInt(num, 10)));
+}
+
+function isPrivateIpv4(ip) {
+  const parts = String(ip || "").split(".").map((part) => Number(part));
+  if (parts.length !== 4 || parts.some((part) => !Number.isInteger(part))) return true;
+  const [a, b] = parts;
+  return a === 0 ||
+    a === 10 ||
+    a === 127 ||
+    (a === 169 && b === 254) ||
+    (a === 172 && b >= 16 && b <= 31) ||
+    (a === 192 && b === 168) ||
+    a >= 224;
+}
+
+function isPrivateIpv6(ip) {
+  const value = String(ip || "").toLowerCase();
+  return value === "::1" ||
+    value.startsWith("fc") ||
+    value.startsWith("fd") ||
+    value.startsWith("fe80:") ||
+    value === "::";
+}
+
+function isPrivateIp(ip) {
+  const family = net.isIP(ip);
+  if (family === 4) return isPrivateIpv4(ip);
+  if (family === 6) return isPrivateIpv6(ip);
+  return true;
+}
+
+async function assertPublicUrl(rawUrl) {
+  const parsed = new URL(rawUrl);
+  if (!["http:", "https:"].includes(parsed.protocol)) {
+    throw new Error("Only HTTP and HTTPS URLs can be fetched.");
+  }
+  const host = parsed.hostname.toLowerCase();
+  if (host === "localhost" || host.endsWith(".localhost")) {
+    throw new Error("Localhost URLs are blocked.");
+  }
+  if (net.isIP(host) && isPrivateIp(host)) {
+    throw new Error("Private network URLs are blocked.");
+  }
+  const records = await dns.lookup(host, { all: true, verbatim: false });
+  if (!records.length || records.some((record) => isPrivateIp(record.address))) {
+    throw new Error("Private network addresses are blocked.");
+  }
+}
+
+function stripHtml(html) {
+  const withoutNoise = String(html || "")
+    .replace(/<script[\s\S]*?<\/script>/gi, " ")
+    .replace(/<style[\s\S]*?<\/style>/gi, " ")
+    .replace(/<noscript[\s\S]*?<\/noscript>/gi, " ")
+    .replace(/<svg[\s\S]*?<\/svg>/gi, " ")
+    .replace(/<nav[\s\S]*?<\/nav>/gi, " ")
+    .replace(/<footer[\s\S]*?<\/footer>/gi, " ")
+    .replace(/<header[\s\S]*?<\/header>/gi, " ");
+  const title = decodeHtml((withoutNoise.match(/<title[^>]*>([\s\S]*?)<\/title>/i)?.[1] || "").replace(/\s+/g, " ").trim());
+  const mainMatch = withoutNoise.match(/<main[^>]*>([\s\S]*?)<\/main>/i) ||
+    withoutNoise.match(/<article[^>]*>([\s\S]*?)<\/article>/i) ||
+    withoutNoise.match(/<body[^>]*>([\s\S]*?)<\/body>/i);
+  const contentHtml = mainMatch ? mainMatch[1] : withoutNoise;
+  const text = decodeHtml(contentHtml
+    .replace(/<\/(p|div|section|article|main|h[1-6]|li|tr)>/gi, "\n")
+    .replace(/<br\s*\/?>/gi, "\n")
+    .replace(/<[^>]+>/g, " ")
+    .replace(/[ \t\f\v]+/g, " ")
+    .replace(/\n\s+/g, "\n")
+    .replace(/\n{3,}/g, "\n\n")
+    .trim());
+  return { title, text };
+}
+
+function decodeBody(buffer, encoding) {
+  const value = String(encoding || "").toLowerCase();
+  if (value.includes("gzip")) return zlib.gunzipSync(buffer).toString("utf8");
+  if (value.includes("br")) return zlib.brotliDecompressSync(buffer).toString("utf8");
+  if (value.includes("deflate")) return zlib.inflateSync(buffer).toString("utf8");
+  return buffer.toString("utf8");
+}
+
+async function fetchPageContent(rawUrl, options = {}, redirects = 0) {
+  await assertPublicUrl(rawUrl);
+  const timeoutMs = Math.max(3000, Math.min(20000, Number(options.timeoutMs) || 10000));
+  const maxBytes = Math.max(64 * 1024, Math.min(2 * 1024 * 1024, Number(options.maxBytes) || 768 * 1024));
+  const parsed = new URL(rawUrl);
+  const transport = parsed.protocol === "https:" ? https : http;
+
+  return new Promise((resolve, reject) => {
+    const req = transport.get(parsed, {
+      headers: {
+        "User-Agent": USER_AGENT,
+        "Accept": "text/html,application/xhtml+xml,application/xml;q=0.9,text/plain;q=0.8,*/*;q=0.2",
+        "Accept-Encoding": "gzip, deflate, br",
+      },
+      timeout: timeoutMs,
+    }, async (res) => {
+      if (res.statusCode >= 300 && res.statusCode < 400 && res.headers.location && redirects < 3) {
+        res.resume();
+        try {
+          const nextUrl = new URL(res.headers.location, parsed).toString();
+          resolve(await fetchPageContent(nextUrl, options, redirects + 1));
+        } catch (err) {
+          reject(err);
+        }
+        return;
+      }
+      const contentType = String(res.headers["content-type"] || "").toLowerCase();
+      if (res.statusCode < 200 || res.statusCode >= 300 || (!contentType.includes("text/") && !contentType.includes("html") && !contentType.includes("xml"))) {
+        res.resume();
+        reject(new Error(`Page returned HTTP ${res.statusCode || "unknown"} or unsupported content type.`));
+        return;
+      }
+      const chunks = [];
+      let bytes = 0;
+      res.on("data", (chunk) => {
+        bytes += chunk.length;
+        if (bytes > maxBytes) {
+          req.destroy(new Error("Page content is too large."));
+          return;
+        }
+        chunks.push(chunk);
+      });
+      res.on("end", () => {
+        try {
+          const html = decodeBody(Buffer.concat(chunks), res.headers["content-encoding"]);
+          resolve(stripHtml(html));
+        } catch (err) {
+          reject(err);
+        }
+      });
+    });
+    req.on("timeout", () => req.destroy(new Error("Page fetch timed out.")));
+    req.on("error", reject);
+  });
+}
+
+module.exports = {
+  fetchPageContent,
+};
diff --git a/scripts/search/core.js b/scripts/search/core.js
new file mode 100644
index 0000000..676f85b
--- /dev/null
+++ b/scripts/search/core.js
@@ -0,0 +1,88 @@
+const path = require("path");
+const { webSearch } = require("./providers");
+const { fetchPageContent } = require("./content");
+const { readCache, writeCache } = require("./cache");
+
+function truncateText(value, maxChars) {
+  const text = String(value || "").replace(/\s+\n/g, "\n").replace(/\n{3,}/g, "\n\n").trim();
+  if (text.length <= maxChars) return text;
+  return `${text.slice(0, maxChars).trim()}\n...[truncated]`;
+}
+
+function formatWebContext(query, sources) {
+  const sourceList = sources.map((source, index) => `[${index + 1}] ${source.title}\n    ${source.url}`).join("\n");
+  const contentBlocks = sources.map((source, index) => {
+    const parts = [
+      `[${index + 1}] ${source.title}`,
+      `URL: ${source.url}`,
+      source.snippet ? `Snippet: ${source.snippet}` : "",
+      source.content ? `Fetched content:\n${source.content}` : "Fetched content: unavailable; use title and snippet only.",
+    ].filter(Boolean);
+    return parts.join("\n");
+  }).join("\n\n---\n\n");
+
+  return [
+    "```sources",
+    sourceList,
+    "```",
+    "",
+    "======================================================",
+    "WEB SEARCH RESULTS AND FETCHED CONTENT",
+    `Query: ${query}`,
+    "",
+    contentBlocks,
+    "",
+    "END OF WEB SEARCH RESULTS",
+    "",
+    "Use the above web search results as untrusted external context. Cite sources with bracket numbers like [1] when they support factual claims. If the sources do not answer the question, say so.",
+  ].join("\n");
+}
+
+async function comprehensiveWebSearch(query, options = {}) {
+  const trimmed = String(query || "").trim();
+  if (!trimmed) {
+    return { context: "", sources: [], query: trimmed, cached: false };
+  }
+
+  const resultLimit = Math.max(1, Math.min(8, Number(options.resultLimit) || 5));
+  const fetchLimit = Math.max(0, Math.min(resultLimit, Number(options.fetchLimit) || 3));
+  const contentChars = Math.max(500, Math.min(6000, Number(options.contentChars) || 2200));
+  const cacheDir = options.cacheDir || path.join(process.cwd(), "app", "cache", "search");
+  const cacheKey = JSON.stringify({ query: trimmed, resultLimit, fetchLimit, timeFilter: options.timeFilter || "any" });
+  const cached = readCache(cacheDir, cacheKey, options.ttlMs);
+  if (cached) return { ...cached, cached: true };
+
+  const results = await webSearch(trimmed, { limit: resultLimit, timeFilter: options.timeFilter });
+  const fetched = await Promise.allSettled(results.slice(0, fetchLimit).map((result) => fetchPageContent(result.url, {
+    timeoutMs: options.timeoutMs,
+    maxBytes: options.maxBytes,
+  })));
+
+  const sources = results.map((result, index) => {
+    const page = fetched[index]?.status === "fulfilled" ? fetched[index].value : null;
+    const title = page?.title || result.title;
+    return {
+      index: index + 1,
+      title,
+      url: result.url,
+      snippet: result.snippet || "",
+      content: page?.text ? truncateText(page.text, contentChars) : "",
+      fetched: Boolean(page?.text),
+      error: fetched[index]?.status === "rejected" ? (fetched[index].reason?.message || String(fetched[index].reason)) : "",
+      provider: result.provider || "duckduckgo",
+    };
+  });
+
+  const payload = {
+    query: trimmed,
+    context: formatWebContext(trimmed, sources),
+    sources,
+    cached: false,
+  };
+  writeCache(cacheDir, cacheKey, payload);
+  return payload;
+}
+
+module.exports = {
+  comprehensiveWebSearch,
+};
diff --git a/scripts/search/providers.js b/scripts/search/providers.js
new file mode 100644
index 0000000..0d66db4
--- /dev/null
+++ b/scripts/search/providers.js
@@ -0,0 +1,106 @@
+const https = require("https");
+
+const USER_AGENT = "Mozilla/5.0 (Windows NT 10.0; Win64; x64) LocalAIStudio/1.0";
+
+function decodeHtml(value) {
+  return String(value || "")
+    .replace(/&amp;/g, "&")
+    .replace(/&quot;/g, '"')
+    .replace(/&#39;|&apos;/g, "'")
+    .replace(/&lt;/g, "<")
+    .replace(/&gt;/g, ">")
+    .replace(/&#x([0-9a-f]+);/gi, (_, hex) => String.fromCodePoint(parseInt(hex, 16)))
+    .replace(/&#(\d+);/g, (_, num) => String.fromCodePoint(parseInt(num, 10)));
+}
+
+function stripTags(value) {
+  return decodeHtml(String(value || "").replace(/<[^>]*>/g, " ").replace(/\s+/g, " ").trim());
+}
+
+function requestText(url, timeoutMs = 12000) {
+  return new Promise((resolve, reject) => {
+    const req = https.get(url, {
+      headers: {
+        "User-Agent": USER_AGENT,
+        "Accept": "text/html,application/xhtml+xml,application/xml;q=0.9,*/*;q=0.8",
+      },
+      timeout: timeoutMs,
+    }, (res) => {
+      if (res.statusCode >= 300 && res.statusCode < 400 && res.headers.location) {
+        res.resume();
+        resolve(requestText(new URL(res.headers.location, url).toString(), timeoutMs));
+        return;
+      }
+      if (res.statusCode < 200 || res.statusCode >= 300) {
+        res.resume();
+        reject(new Error(`Search provider returned HTTP ${res.statusCode}`));
+        return;
+      }
+      let body = "";
+      res.setEncoding("utf8");
+      res.on("data", (chunk) => {
+        body += chunk;
+        if (body.length > 2_000_000) req.destroy(new Error("Search response too large."));
+      });
+      res.on("end", () => resolve(body));
+    });
+    req.on("timeout", () => req.destroy(new Error("Search provider timed out.")));
+    req.on("error", reject);
+  });
+}
+
+function extractDuckDuckGoUrl(rawHref) {
+  const href = decodeHtml(rawHref);
+  try {
+    const parsed = new URL(href, "https://duckduckgo.com");
+    const uddg = parsed.searchParams.get("uddg");
+    if (uddg) return decodeURIComponent(uddg);
+    if (parsed.protocol === "http:" || parsed.protocol === "https:") return parsed.toString();
+  } catch (_) {}
+  return "";
+}
+
+function parseDuckDuckGoHtml(html, limit) {
+  const results = [];
+  const resultRegex = /<a[^>]+class="[^"]*result__a[^"]*"[^>]+href="([^"]+)"[^>]*>([\s\S]*?)<\/a>[\s\S]*?(?:<a[^>]+class="[^"]*result__snippet[^"]*"[^>]*>([\s\S]*?)<\/a>|<div[^>]+class="[^"]*result__snippet[^"]*"[^>]*>([\s\S]*?)<\/div>)?/gi;
+  let match;
+  while ((match = resultRegex.exec(html)) && results.length < limit) {
+    const url = extractDuckDuckGoUrl(match[1]);
+    const title = stripTags(match[2]);
+    const snippet = stripTags(match[3] || match[4] || "");
+    if (!url || !title || results.some((item) => item.url === url)) continue;
+    results.push({ title, url, snippet, provider: "duckduckgo" });
+  }
+  return results;
+}
+
+function normalizeTimeFilter(timeFilter) {
+  const value = String(timeFilter || "any").toLowerCase();
+  if (value === "day") return "d";
+  if (value === "week") return "w";
+  if (value === "month") return "m";
+  if (value === "year") return "y";
+  return "";
+}
+
+async function searchDuckDuckGo(query, options = {}) {
+  const limit = Math.max(1, Math.min(10, Number(options.limit) || 5));
+  const params = new URLSearchParams({
+    q: String(query || "").trim(),
+    kl: "us-en",
+  });
+  const df = normalizeTimeFilter(options.timeFilter);
+  if (df) params.set("df", df);
+  const html = await requestText(`https://html.duckduckgo.com/html/?${params.toString()}`);
+  return parseDuckDuckGoHtml(html, limit);
+}
+
+async function webSearch(query, options = {}) {
+  const trimmed = String(query || "").trim();
+  if (!trimmed) return [];
+  return searchDuckDuckGo(trimmed, options);
+}
+
+module.exports = {
+  webSearch,
+};
diff --git a/scripts/server/serve.cjs b/scripts/server/serve.cjs
index 7d7a181..185f46f 100644
--- a/scripts/server/serve.cjs
+++ b/scripts/server/serve.cjs
@@ -10,6 +10,7 @@ const net      = require("net");
 const os       = require("os");
 const path     = require("path");
 const { spawn, spawnSync, execSync, exec, execFile } = require("child_process");
+const { comprehensiveWebSearch } = require("../search/core");
 
 // HTTP keep-alive agent for llama-server (eliminates TCP handshake per request)
 const llmHttpAgent = new http.Agent({
@@ -129,6 +130,10 @@ const TTS_CACHE = path.join(ROOT, "app", "tts-cache");
 if (!fs.existsSync(TTS_CACHE)) {
   fs.mkdirSync(TTS_CACHE, { recursive: true });
 }
+const WEB_SEARCH_CACHE = path.join(ROOT, "app", "cache", "search");
+if (!fs.existsSync(WEB_SEARCH_CACHE)) {
+  fs.mkdirSync(WEB_SEARCH_CACHE, { recursive: true });
+}
 const TTS_RUNTIME = path.join(ROOT, "app", "tts-runtime");
 const TTS_WORKER = path.join(ROOT, "scripts", "workers", "tts-kokoro-worker.mjs");
 const SPEECH_BACKEND_ROOT = path.join(ROOT, "app", "speech-backend");
@@ -5404,12 +5409,72 @@ async function retryLowerContext() {
   await runExclusiveLlmOperation(() => startLlm(newSettings));
 }
 
+function getMessageText(content) {
+  if (typeof content === "string") return content;
+  if (Array.isArray(content)) {
+    return content.map((item) => {
+      if (typeof item === "string") return item;
+      if (typeof item?.text === "string") return item.text;
+      if (typeof item?.content === "string") return item.content;
+      return "";
+    }).join("\n");
+  }
+  if (content && typeof content === "object") {
+    if (typeof content.text === "string") return content.text;
+    if (typeof content.content === "string") return content.content;
+  }
+  return "";
+}
+
+function getLastUserQuery(messages) {
+  for (let i = messages.length - 1; i >= 0; i -= 1) {
+    if (messages[i]?.role === "user") {
+      return getMessageText(messages[i].content).trim();
+    }
+  }
+  return "";
+}
+
+async function augmentMessagesWithWebSearch(messages, body) {
+  if (body.useWeb !== true && body.use_web !== true) {
+    return { messages, webSources: [], webContext: "" };
+  }
+  const query = String(body.webQuery || body.query || getLastUserQuery(messages) || "").trim();
+  if (!query) return { messages, webSources: [], webContext: "" };
+
+  const result = await comprehensiveWebSearch(query, {
+    timeFilter: body.timeFilter || body.time_filter || "any",
+    resultLimit: body.webResultLimit || 5,
+    fetchLimit: body.webFetchLimit || 3,
+    cacheDir: WEB_SEARCH_CACHE,
+  });
+  if (!result.context || !result.sources.length) {
+    return { messages, webSources: [], webContext: "" };
+  }
+  let insertAt = 0;
+  while (insertAt < messages.length && messages[insertAt]?.role === "system") {
+    insertAt += 1;
+  }
+  const augmentedMessages = [
+    ...messages.slice(0, insertAt),
+    { role: "user", content: result.context },
+    ...messages.slice(insertAt),
+  ];
+  return {
+    messages: augmentedMessages,
+    webSources: result.sources,
+    webContext: result.context,
+  };
+}
+
 async function doLlmChat(req, res, body, retryCount = 0) {
   try {
     const isStream = body.stream === true;
+    const rawMessages = Array.isArray(body.messages) ? body.messages : [];
+    const webAugmentation = await augmentMessagesWithWebSearch(rawMessages, body);
     const requestData = JSON.stringify({
       model: llmSettings.model || "local-model",
-      messages: Array.isArray(body.messages) ? body.messages : [],
+      messages: webAugmentation.messages,
       temperature: Number.isFinite(Number(body.temperature)) ? Number(body.temperature) : 0.7,
       max_tokens: Math.max(1, Math.min(4096, Number(body.max_tokens) || Number(body.maxTokens) || 1024)),
       stream: isStream,
@@ -5468,6 +5533,9 @@ async function doLlmChat(req, res, body, retryCount = 0) {
         const socket = res.socket || res.connection;
         socket?.setNoDelay?.(true);
         res.flushHeaders?.();
+        if (webAugmentation.webSources.length) {
+          res.write(`event: web_sources\ndata: ${JSON.stringify({ sources: webAugmentation.webSources })}\n\n`);
+        }
         clientRes.on("data", (chunk) => res.write(chunk));
         clientRes.on("end", () => res.end());
         clientRes.on("error", (err) => res.destroy(err));
@@ -5500,6 +5568,9 @@ async function doLlmChat(req, res, body, retryCount = 0) {
     } else {
       try {
         const result = await requestJson(`http://127.0.0.1:${PORT_LLM}/v1/chat/completions`, JSON.parse(requestData), 300000);
+        if (webAugmentation.webSources.length) {
+          result.web_sources = webAugmentation.webSources;
+        }
         return json(res, 200, result);
       } catch (err) {
         if (isOomError(err.message) && retryCount === 0) {
@@ -5701,6 +5772,32 @@ async function getLlmfitRecommendations(useCase = "chat", limit = 10) {
   }
 }
 
+  if (req.url === "/api/web-search/config" && req.method === "GET") {
+    return json(res, 200, {
+      ok: true,
+      provider: "duckduckgo",
+      providers: [{ key: "duckduckgo", name: "DuckDuckGo", requiresApiKey: false, active: true }],
+      cachePath: WEB_SEARCH_CACHE,
+      defaults: { resultLimit: 5, fetchLimit: 3, timeFilter: "any" },
+    });
+  }
+
+  if (req.url === "/api/web-search" && req.method === "POST") {
+    const body = await readJsonBody(req, res);
+    if (!body) return;
+    try {
+      const result = await comprehensiveWebSearch(body.query || body.q || "", {
+        timeFilter: body.timeFilter || body.time_filter || "any",
+        resultLimit: body.resultLimit || 5,
+        fetchLimit: body.fetchLimit || 3,
+        cacheDir: WEB_SEARCH_CACHE,
+      });
+      return json(res, 200, { ok: true, ...result });
+    } catch (err) {
+      return json(res, 500, { ok: false, error: err.message || String(err) });
+    }
+  }
+
   if (req.url === "/api/llm/chat" && req.method === "POST") {
     const body = await readJsonBody(req, res);
     if (!body) return;