> ## Documentation Index
> Fetch the complete documentation index at: https://docs.minns.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# Security & resilience
> Built-in protections against circular references, prototype pollution, oversized payloads, and queue overflow.
The SDK is designed to be resilient against erratic LLM outputs and large data structures. Every event passes through multiple safety checks before submission.
## Circular reference protection
LLM-generated objects can accidentally contain circular references. The SDK automatically detects these and replaces them with `"[Circular]"` during serialization.
```typescript theme={null}
const obj: any = { name: "test" };
obj.self = obj; // Circular!
// SDK handles this gracefully
await client.event("agent")
.action("process", obj)
.send(); // Serialized with "[Circular]" placeholder
```
## Prototype pollution protection
The SDK strips dangerous JavaScript keys during serialization to prevent prototype pollution attacks:
* `__proto__`
* `constructor`
* `prototype`
This is especially important when processing untrusted LLM output that could contain adversarial payloads.
## Size guarding
Payloads exceeding `maxPayloadSize` (default: **1 MB**) are rejected **after** serialization. This prevents accidental submission of oversized events that could degrade server performance.
The size check happens after JSON serialization, so the actual byte size of the serialized payload is what's measured — not the in-memory object size.
## Queue management
The local auto-batch queue has a configurable maximum depth (`maxQueueSize`, default: **1000**). If the queue is full, `enqueue()` throws an error rather than silently dropping events.
```typescript theme={null}
const client = createClient({
baseUrl: "https://minns.ai/api/",
autoBatch: true,
maxQueueSize: 5000, // Increase for high-throughput agents
});
```
## Telemetry
The SDK includes opt-in, fire-and-forget telemetry to monitor performance and LLM token usage.
| What's collected | What's NOT collected |
| :--------------------- | :------------------- |
| Latency metrics | Request bodies |
| HTTP status codes | Raw event content |
| Estimated token counts | User data |
| Error messages | Metadata values |
Telemetry is enabled by default. To opt out:
```typescript theme={null}
const client = createClient({
baseUrl: "https://minns.ai/api/",
enableDefaultTelemetry: false,
});
```
## Next steps
Review all client configuration options.
Explore the full REST API.