Alchemyst AI — Vercel AI SDK Integration

Alchemyst AI is the context layer for your LLM applications — it remembers, reasons, and injects contextual intelligence automatically into every call. This package provides a seamless integration with Vercel's AI SDK to enhance your Gen-AI apps with memory, retrieval, and context-aware toolchains — all through a single line of configuration.

---

🚀 Installation

npm install @alchemystai/aisdk
# or
yarn add @alchemystai/aisdk
#or
pnpm add @alchemystai/aisdk
# or
bun add @alchemystai/aisdk

---

⚡ Quick Start

Here's how to plug Alchemyst AI into your ai SDK call using the withAlchemyst middleware:

import { google } from '@ai-sdk/google';
import { generateText } from 'ai';
import { withAlchemyst } from '@alchemystai/aisdk';

const generateTextWithMemory = withAlchemyst(generateText, {
  source: "my_app", // optional
  apiKey: "YOUR_ALCHEMYST_API_KEY", // optional, can be left if you have ALCHEMYST_AI_API_KEY
  debug: true // optional, defaults to false
});

const result = await generateTextWithMemory({
  model: google("gemini-2.5-flash"),
  prompt: "Remember that my name is Alice",
  userId: "user-123",
  sessionId: "session-abc",
});

This automatically attaches the Alchemyst Context Engine to your model call — enabling persistent memory and context-aware generation across sessions.

---

🧩 API Reference

withAlchemyst(fn, options)

Parameter	Type	Default	Description
fn	Function	—	The AI SDK function to wrap (e.g., generateText).
options.source	string	—	Identifier for your application source.
options.apiKey	string	—	Your Alchemyst AI API key. Required.
options.debug	boolean	false	Enables debug logging.

The wrapped function accepts additional parameters:

Parameter	Type	Description
userId	string	Unique identifier for the user.
sessionId	string	Unique identifier for the conversation/session.

Returns the result from the wrapped AI SDK function.

---

🧠 What It Does

Once integrated, Alchemyst AI automatically:

• Persists context across user sessions.
• Augments prompts with retrieved knowledge from your Alchemyst AI workspace.
• Supports custom tool functions for domain-specific reasoning.
• Runs entirely server-side — no extra infrastructure required.

---

💡 Example: Contextual Chatbot

import { google } from '@ai-sdk/google';
import { generateText } from 'ai';
import { withAlchemyst } from '@alchemystai/aisdk';

const generateTextWithMemory = withAlchemyst(generateText, {
  source: "chatbot",
  apiKey: process.env.ALCHEMYST_API_KEY!,
  debug: false
});

export async function POST(req: Request) {
  const { prompt, userId, sessionId } = await req.json();

  const result = await generateTextWithMemory({
    model: google("gemini-2.5-flash"),
    prompt,
    userId,
    sessionId,
  });

  return new Response(result.text);
}

With this, your AI app will:

• Remember previous user messages (via Alchemyst context memory)
• Retrieve relevant knowledge chunks
• Enrich every prompt dynamically before sending it to the model

---

🔒 Environment Variables

Set your API key in your environment:

ALCHEMYST_API_KEY=sk-xxxxxx

Then reference it in your code as:

const generateTextWithMemory = withAlchemyst(generateText, {
  source: "my_app",
  apiKey: process.env.ALCHEMYST_API_KEY!,
});

---

🧰 Supported SDK Methods

Alchemyst AI integrates with all Vercel AI SDK entry points:

SDK Function	Supported	Notes
streamText	✅	Real-time streaming generation
generateText	✅	Synchronous generation
streamObject	✅	Structured JSON output
generateObject	✅	Non-streaming structured output

---

🧪 Example with Different Conversations

import { google } from '@ai-sdk/google';
import { generateText } from 'ai';
import { withAlchemyst } from '@alchemystai/aisdk';

const generateTextWithMemory = withAlchemyst(generateText, {
  source: "api_sdk_test",
  apiKey: "YOUR_ALCHEMYST_API_KEY",
  debug: true
});

// First conversation
const result1 = await generateTextWithMemory({
  model: google("gemini-2.5-flash"),
  prompt: "What is the capital of France?",
  userId: "12345",
  sessionId: "test-convo-1",
});

// Second conversation
const result2 = await generateTextWithMemory({
  model: google("gemini-2.5-flash"),
  prompt: "What is the capital of Germany?",
  userId: "12345",
  sessionId: "test-convo-2",
});

---

📜 License

MIT © 2025 Alchemyst AI