GitDealFlow × Vercel AI SDK

What you can build

Tool() in a Route Handler

// app/api/chat/route.ts (Next.js App Router)
// Models route through the Vercel AI Gateway via plain string IDs.
import { tool, streamText, convertToModelMessages, type UIMessage } from "ai";
import { z } from "zod";

export const runtime = "nodejs"; // tool fetch + LLM stream

const A2A = "https://signals.gitdealflow.com/api/a2a";

const gitdealflow = tool({
  description:
    "Live VC engineering signals, trending startups, sector watchlists, named-startup profiles, dataset summaries, methodology citation.",
  inputSchema: z.object({
    skill: z.enum([
      "get_trending_startups",
      "search_startups_by_sector",
      "get_startup_signal",
      "get_signals_summary",
      "get_methodology",
    ]),
    args: z.record(z.string(), z.any()).optional(),
  }),
  execute: async ({ skill, args }) => {
    const res = await fetch(A2A, {
      method: "POST",
      headers: { "Content-Type": "application/json" },
      body: JSON.stringify({
        jsonrpc: "2.0", id: 1, method: "message/send",
        params: { message: { role: "user", parts: [
          { kind: "data", data: { skill, args: args ?? {} } },
        ]}},
      }),
    });
    return (await res.json()).result.artifacts[0].parts[0].data;
  },
});

export async function POST(req: Request) {
  const { messages }: { messages: UIMessage[] } = await req.json();
  const result = streamText({
    model: "openai/gpt-5.4",
    tools: { gitdealflow },
    messages: convertToModelMessages(messages),
  });
  return result.toUIMessageStreamResponse();
}

Server Component with cached weekly fetch

// app/dashboard/trending/page.tsx
import { unstable_cache } from "next/cache";

const A2A = "https://signals.gitdealflow.com/api/a2a";

const fetchTrending = unstable_cache(
  async () => {
    const res = await fetch(A2A, {
      method: "POST",
      headers: { "Content-Type": "application/json" },
      body: JSON.stringify({
        jsonrpc: "2.0", id: 1, method: "message/send",
        params: { message: { role: "user", parts: [
          { kind: "data", data: { skill: "get_trending_startups" } },
        ]}},
      }),
    });
    return (await res.json()).result.artifacts[0].parts[0].data.startups;
  },
  ["gitdealflow-trending"],
  { revalidate: 21600, tags: ["gitdealflow"] }, // 6h
);

export default async function TrendingPage() {
  const startups = await fetchTrending();
  return (
    <main>
      <h1>Trending this week</h1>
      <ol>
        {startups.map((s: { name: string; sector: string }) => (
          <li key={s.name}>
            <strong>{s.name}</strong>: {s.sector}
          </li>
        ))}
      </ol>
    </main>
  );
}

Five prompts to try first

• ›Pull weekly top 20 inside a Server Component to seed a dashboard.
• ›Stream a chat reply that calls gitdealflow when the user asks about a startup.
• ›Trigger from a Server Action: 'add this org to my watchlist'.
• ›Schedule a weekly Vercel Cron route that emails the top 5 breakouts.
• ›Compose with the AI Gateway: route fast lookups to a small model, memos to a large one.

FAQ

Gotchas

• tool() expects you to return JSON-serializable data; the .data field on artifacts is already a plain object, passing it through is safe.
• If you use Edge runtime, the fetch to A2A should be on regular Node runtime since you may want to hold connections longer. Set `export const runtime = 'nodejs'`.
• The Vercel AI SDK's streamText pipes tool calls + tool results into the data stream, render them with the Generative UI pattern, not as plain text.

References

Other frameworks

• LangChain
• CrewAI
• Letta
• Mastra

When the free signal isn’t enough