How It Works
From API key to useful memory in a few controlled steps.
The integration model is intentionally simple for the product team and intentionally strict under the hood: authenticated writes, queued processing, isolated collections, and runtime retrieval over stable routes.
Your first memory-powered flow in under 15 minutes.
Three steps to go from zero to a product that remembers users across sessions.
Create your workspace
Generate an API key from the console and set your project-level auth in minutes.
Send memories from your app
Ingest user events, preferences, and conversation snippets through one consistent endpoint.
Retrieve context on every response
Query by meaning and metadata to feed the right memory back into your prompt pipeline.
Already have an account? Go to dashboard ->
Developer experience built for speed.
The API is straightforward for the first prototype and stable enough for production traffic.
Simple endpoints
Ingest and retrieve through clean REST routes that are easy to test and maintain.
Fast integration
Drop into existing Node, Python, and serverless backends without rewriting your app architecture.
Production auth
Use scoped API keys for runtime traffic and signed sessions for dashboard access.
- 01await fetch("https://api.neuralbase.cloud/v1/memories", {
- 02method: "POST",
- 03headers: { Authorization: "Bearer nb_live_..." },
- 04body: JSON.stringify({ userId, content, metadata })
- 05});
- 01import requests
- 02result
- 03= requests.get(
- 04"https://api.neuralbase.cloud/v1/memories",
- 05params={"query": "refund policy preference"}
- 06)
- 07# result["items"][0]["score"] -> 0.94
Runtime flow
What happens after your app sends the request.
Neuralbase is not just a thin proxy to a vector database. The platform adds ingestion control, quota enforcement, document processing, and isolation before anything is written or returned.
| Stage | What your app does | What Neuralbase handles |
|---|---|---|
| 1. Write | Your backend sends memories or documents to Neuralbase over authenticated API routes. | Requests are attached to a workspace, rate limited, and counted against plan quotas immediately after auth. |
| 2. Process | Content is normalized, chunked, parsed, or queued for asynchronous processing depending on the route. | BullMQ and Redis handle background work, while document extraction and status updates remain visible in the dashboard. |
| 3. Index | Embeddings are generated and stored in a per-user Qdrant collection rather than a shared tenant bucket. | This keeps the retrieval boundary aligned with the account the platform is billing and protecting. |
| 4. Retrieve | Your app calls list or search routes when it needs context for a response or workflow decision. | Neuralbase returns the most relevant memory and metadata while keeping vector internals hidden behind the API. |
Keep control of your memory infrastructure.
Neuralbase works the way serious teams prefer: a public API layer in front, a private vector store behind it, and nothing exposed that should not be.
Public API layer
Expose only your API domain to clients. Internal services stay protected with no direct vector access from the outside.
Private vector runtime
Run your vector layer on the same infrastructure as your backend for lower latency and tighter data control.
Managed embedding layer
Use Neuralbase's managed embedding layer now and keep room to tune retrieval as traffic scales.