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

# Observability

> Monitor, analyze, and audit platform activity using logs, metrics, analytics, and alerts.

## Overview

LLMGrid provides built‑in **observability features** to help administrators and operators understand how the platform is being used, detect issues, investigate incidents, and maintain reliability.

Observability in LLMGrid is centered around **logs**, **usage metrics**, **cost analytics**, and **health checks**, all managed through the UI and enforced consistently across models, agents, tools, and workflows.

***

## Core Observability Capabilities

LLMGrid observability is designed to answer four key questions:

* What requests are being made?
* How are models, tools, and agents behaving?
* Where is usage and cost coming from?
* Are systems healthy and compliant?

These questions are addressed through multiple observability surfaces.

***

## Request & Execution Logs

### Request Logs

The **Request Logs** page provides structured, request‑level visibility across the platform.

Each logged request includes:

* Timestamp
* Request ID and session ID
* Model or route used
* Success or failure status
* Execution duration
* Token usage
* Cost metadata
* Associated virtual key, team, agent, or tag

Logs can be filtered by time range, model, key, team, tag, or status.

***

### Audit Logs

Audit logs capture **administrative and configuration actions**, such as:

* Creating or updating models
* Modifying routing, guardrails, or budgets
* Managing keys, users, or credentials

This supports:

* Compliance audits
* Change tracking
* Incident investigation

Audit logs are read‑only and immutable.

***

## Usage & Metrics

### Usage Dashboard

The **Usage** section provides aggregated visibility into platform consumption.

Metrics include:

* Total requests
* Successful vs failed requests
* Token usage
* Average request cost
* Daily usage trends

Usage can be broken down by:

* Global tenant usage
* Organization
* Team
* User or agent
* Tag
* Virtual key
* Model

***

### Model & Key Activity

Dedicated views allow you to analyze:

* Which models are most used
* Which keys drive the highest activity
* Agent‑initiated vs user‑initiated traffic

This helps with capacity planning and optimization.

***

## Cost Observability

### Cost Tracking

Cost tracking surfaces how usage translates into cost, including:

* Base cost
* Applied discounts
* Final effective cost

Cost data is visible alongside usage metrics and can be filtered by the same dimensions (team, key, tag, model).

***

### Budgets & Limits

Observability integrates with **Budgets** and **Rate Limits** to:

* Detect approaching limits
* Identify throttled requests
* Prevent unexpected overuse

Budget enforcement events are visible in logs and metrics.

***

## Guardrails Observability

When guardrails are enabled, observability includes:

* Guardrail enforcement decisions
* Blocked or modified requests
* Logging‑only detections
* Tool and MCP validation outcomes

This allows safety and compliance teams to observe policy impact without disrupting traffic.

***

## Health & System Monitoring

### Cache Health

The **Caching Health** view validates connectivity and readiness for caching backends.

Health checks provide:

* High‑level status
* Detailed diagnostic output

***

### Model & Tool Health

Health checks and execution logs help identify:

* Unavailable models
* Tool connection failures
* Integration‑specific issues

These signals support reliability engineering and operational response.

***

## Tags & Attribution

Tags play a key role in observability by enabling attribution and segmentation.

You can observe usage, cost, and behavior by:

* Client
* Environment (prod, staging)
* Integration source
* Request context

Tags flow through logs, metrics, and analytics consistently.

***

## Data Export & Analysis

Observability data can be exported or integrated with external systems via:

* API access
* Programmatic log ingestion
* Analytics and reporting pipelines

This enables long‑term analysis and integration with SIEM or monitoring platforms.

***

## Operational Best Practices

* Set a regular cadence for reviewing usage and logs
* Use tags to maintain attribution clarity
* Monitor failure rates and latency trends
* Validate changes using logs after configuration updates
* Keep guardrail logs enabled during policy rollout
* Use budgets and alerts as early‑warning signals

***

## Observability & Governance Alignment

Observability features are tightly integrated with governance controls:

* **Virtual Keys** define access boundaries
* **Guardrails** enforce and log policy decisions
* **Budgets** define operational limits
* **Routing** affects execution paths
* **Logs** record all outcomes

This ensures observability reflects real enforcement behavior, not just metrics.

***

## Related Sections

* **Usage** – High‑level metrics and trends
* **Logs** – Request and audit logs
* **Cost Tracking** – Cost attribution and discounts
* **Budgets** – Usage thresholds and enforcement
* **Guardrails** – Safety and compliance visibility
* **Router Settings** – Routing behavior and outcomes