AI app builders

Add Aether to a Bolt app

Wire Aether memory and retrieval into a full-stack app you're building with Bolt (bolt.new).

Aether is a managed memory and retrieval API: store text or documents, then recall the most relevant pieces by meaning. In a Bolt app, reach for it when you need persistent per-user memory that outlives the session, or RAG over your own documents — anything the app should remember or search semantically.

Prompt Bolt with this

Bolt builds and runs a full-stack project in the browser. Start your prompt with the block below, then describe the app you want — Bolt will wire Aether into the server side. Add your key to the project's .env (Bolt loads it automatically) as AETHER_API_KEY.

text
Use Aether for persistent memory and document retrieval in this app.

Setup:
1. Install the SDK: `npm install @aether-ai/sdk` (for a Python backend: `pip install aether-ai`).
2. Read the API key from the `AETHER_API_KEY` environment variable. Never hardcode it and never expose it to the browser — all Aether calls run server-side.

Per-user memory (TypeScript):
  import { Memory } from "@aether-ai/sdk";
  const memory = new Memory(userId, { apiKey: process.env.AETHER_API_KEY });
  await memory.remember("a fact about this user"); // store one memory
  const hits = await memory.recall("a question", { k: 5 }); // returns [{ text, score }], higher score = more relevant

Document search / RAG (TypeScript):
  import { AetherClient } from "@aether-ai/sdk";
  const client = new AetherClient({ apiKey: process.env.AETHER_API_KEY });
  await client.insertText("some document text"); // store a document
  const results = await client.retrieve("a question", 5); // passages with .content, ready for a prompt

Rules:
- Scope memory per user: pass a stable user id as the entity id.
- Call Aether from a server route, action, or API handler — never from client components.
- Use Memory for personalization; use AetherClient + retrieve for RAG over documents.

If the agent didn't get it right

Three manual steps recover any integration:

  1. Install the SDK — run npm install @aether-ai/sdk in Bolt's built-in terminal (or pip install aether-ai for a Python backend).
  2. Set the key — add AETHER_API_KEY=... to the project .env. Get a key from the Aether Dashboard.
  3. Call from the server — construct new Memory(userId, { apiKey: process.env.AETHER_API_KEY }) inside an API route or server handler and expose remember / recall; the frontend calls that route, never Aether directly.

Good first builds

  • A support chatbot with memoryremember each customer's past issues, then recall them so the bot has context on the next message.
  • Semantic search over product docsinsertText your docs once, then retrieve matching passages for a search box or assistant.

New to the SDK? Walk through the TypeScript quickstart first.