> ## Documentation Index
> Fetch the complete documentation index at: https://docs.statebase.org/llms.txt
> Use this file to discover all available pages before exploring further.

# Reliability Guarantees

> What happens when things go wrong?

Founders choose StateBase because they need a system that doesn't fail when their agent gets popular.

## 1. High Availability (The "What if" factor)

StateBase uses a distributed architecture across multiple GCP regions.

* **Redis Layer**: Sub-10ms state retrieval.
* **Postgres Layer**: Guaranteed durability for long-term history.
* **Failover**: If our main API is unreachable, our SDKs are designed with **Graceful Degradation**.

## 2. Graceful Degradation

If the StateBase API times out, the SDK will:

1. Log a warning to your console.
2. Return an empty context/memory object.
3. Allow the LLM call to proceed (Stateless mode).
   **Result**: Your agent keeps talking, but momentarily loses its "memory" until the connection is restored.

## 3. Data Isolation

Every API Key operates in a strictly isolated environment. Sessions, memories, and traces are nunca shared across tenants.

## 4. Disaster Recovery

We perform backups every 4 hours. In the event of a total system failure, we can restore your agent's state to within the last few minutes of their interaction history.