> ## Documentation Index
> Fetch the complete documentation index at: https://docs.cyberup24.com/llms.txt
> Use this file to discover all available pages before exploring further.
# Overview
The Lightweight AI Reconnaissance Kit (LARK) is a self-contained cybersecurity platform that combines network traffic analysis, threat detection, and edge AI in a single portable appliance. It delivers full visibility into network activity with minimal setup and without adding load or latency to the network being monitored.
Each LARK is delivered as a fully assembled, pre-configured unit. It requires no internal assembly and no software stack to build. Once powered on and connected to a traffic source, the appliance brings its services online automatically and is ready for analysis within minutes.
## Capabilities
The LARK monitors network traffic passively. A copy of traffic is delivered to the appliance from a TAP or a switch SPAN (mirror) port, allowing the appliance to observe all traffic crossing the monitored link without sitting inline and without risk to production traffic.
From that traffic, the appliance generates structured telemetry, detects suspicious activity, retains full packet data for forensic review, and presents the results through a set of web-based tools. An onboard AI layer supports summarization, natural-language investigation, and workflow automation across the collected data, running entirely on the appliance.
Continuous monitoring of east-west and north-south traffic, with full packet capture and protocol-level logging.
Signature-based intrusion detection and file analysis applied to network traffic as it is observed.
Analyst tools for querying, correlating, and visualizing collected data to investigate anomalies.
Local large language model inference for summarization, investigative assistance, and automation.
## Architecture
The LARK runs its entire software stack as containerized services managed by a Kubernetes-based orchestration layer. Every capability, from the network sensors through storage, analytics, and AI, runs locally on the appliance. No external compute or cloud connectivity is required for normal operation, so the appliance can operate in isolated or air-gapped environments provided a traffic source is available at the capture interface.
The containerized design isolates components from one another, allows individual services to be added, updated, or removed without disturbing the rest of the system, and enforces resource boundaries so that no single workload can starve the others.
## The LARK product line
The LARK is available in three models. All three run the same software and operate identically. They differ in compute, memory, and storage capacity, and the largest model runs across multiple nodes rather than one.
| Model | Nodes | Positioning |
| --------- | -------------- | ---------------------------------------------------------------------- |
| SWIFT | Single node | Entry capacity for smaller networks and focused monitoring |
| PEREGRINE | Single node | Mid-range capacity for sustained monitoring and heavier analysis |
| APEX | Multiple nodes | Highest capacity and throughput for large or high-traffic environments |
Because all three models behave identically in operation, this documentation applies to every model unless a page indicates otherwise. Capacity and hardware differences between models are detailed on the models and capacity page.
## Data retention and offload
The LARK retains all captured and processed data locally while in operation. Data may be offloaded to an external destination, such as a central storage location or analytics environment, to retain it beyond the appliance's onboard capacity or to consolidate it with data from other sources. Data handling on the appliance is automated; offloading is described in the data management section.