feat: Exa.ai web search as supplementary context in the answer pipeline

[Copilot]

Adds opt-in web search via Exa.ai so the grounded-answer agent can draw on live web content when the knowledge base lacks coverage.

Changes

lib/exa-search.ts (new)

• Wraps exa.searchAndContents() — neural search, useAutoprompt, up to 3 results × 1,500 chars each
• Returns '' (safe no-op) when EXA_API_KEY is absent or the call fails

Reply pipeline (webhooks/reply + sandbox)

• searchWeb() runs in parallel with MCP/embedding work — no serial latency added
• webContext threaded into runGroundedAnswerAgent and injected into the prompt between live API data and KB sections
• Sandbox early-return (no KB content → short-circuit) now gates on !retrievedContext.trim() && !webContext.trim(), allowing pure-web answers when the KB is empty

Dashboard

• WelcomeDashboard gains a webSearchEnabled: boolean prop (both org pages pass !!process.env.EXA_API_KEY)
• Renders a "🌐 Web Search On/Off" status chip alongside the existing stat chips

Config

• EXA_API_KEY= added to env.example

Enabling

Set EXA_API_KEY in your environment. No DB migrations required — feature is entirely env-gated.

// lib/exa-search.ts
export async function searchWeb(query: string): Promise<string> {
  const apiKey = process.env.EXA_API_KEY;
  if (!apiKey) return ''; // silently disabled

  const exa = new Exa(apiKey);
  const result = await exa.searchAndContents(query, {
    numResults: 3,
    type: 'neural',
    useAutoprompt: true,
    text: { maxCharacters: 1500 },
  });
  // ...formatted citation blocks
}

---

💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more Copilot coding agent tips in the docs.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
answerify	Error		Mar 6, 2026 3:24am