Build powerful AI workflows in code.

1export const userWorkflow = inngest.createFunction(
2  fnOptions, fnListener,
3  async ({ event, step, ctx }) => {
4    const similar = await step.run("query-vectordb", async () => {
5      // Query a vectorDB for similar results given input
6      const embedding = createEmedding(event.data.input);
7      return await index.query({ vector: embedding, topK: 3 }).matches;
8    });
9
10    const response = await step.run("generate-llm-response", async () => {
11      // Inject our prompt given similar search results and event.data.input
12      const prompt = createAgentPrompt(similar, event.data.input);
13      return await llm.createCompletion({
14        model: "gpt-3.5-turbo",
15        prompt,
16      });
17    });
18
19    // Run as many chains and post-AI flows as you need, with retries, state
20    // and orchestration managed for you.
21
22    await step.run("save-to-db", async () => {
23      // Connect to your standard DBs and APIs, as this is regular code that's
24      // deployed in your existing stack
25      await db.summaries.create({ requestID: event.data.requestID, response });
26    });
27  }
28);