llm-sentinel-storage

InfluxDB time-series storage and multi-layer caching for LLM-Sentinel.

Overview

High-performance storage layer with multiple caching tiers:

• InfluxDB v3: Time-series storage for telemetry and anomalies
• Moka Cache: In-memory cache (10,000 entry capacity)
• Redis: Distributed cache for multi-instance deployments
• Query API: Historical data retrieval with time-range queries

Features

• 8,000+ writes/second (batched)
• Dual-layer caching (memory + distributed)
• Automatic TTL management
• Batch writes for efficiency
• Query API for analytics
• Configurable retention policies

Usage

Add this to your Cargo.toml:

[dependencies]
llm-llm-sentinel-storage = "0.1.0"

Example

use llm_sentinel_storage::{StorageManager, StorageConfig};

#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
    let config = StorageConfig {
        influxdb_url: "http://localhost:8086".to_string(),
        influxdb_org: "sentinel".to_string(),
        influxdb_token: std::env::var("INFLUXDB_TOKEN")?,
        telemetry_bucket: "telemetry".to_string(),
        ..Default::default()
    };

    let manager = StorageManager::new(config).await?;

    // Store telemetry
    manager.store_telemetry(&event).await?;

    // Query historical data
    let results = manager.query_telemetry(
        "service_name = 'chat-api'",
        chrono::Duration::hours(24)
    ).await?;

    Ok(())
}

Caching Strategy

1. L1 Cache (Moka): Hot data in memory, sub-microsecond access
2. L2 Cache (Redis): Distributed cache for multi-instance setups
3. L3 Storage (InfluxDB): Persistent time-series storage

License

Apache-2.0