Skip to main content

What is Observability?

Helicone provides end-to-end observability for your LLM applications, giving you complete visibility into every request, session, and user interaction. Track costs, latency, errors, and performance metrics in real-time.

Key Observability Features

Request Logging

Capture and analyze every LLM request with detailed metadata, inputs, and outputs

Session Tracking

Group related requests to understand multi-turn conversations and workflows

Distributed Tracing

Visualize complex LLM workflows with parent-child request relationships

Custom Properties

Add custom metadata to segment and filter your requests

User Metrics

Track costs, usage, and performance per user

Analytics Dashboard

Visualize trends, costs, and performance over time

What You Can Observe

Request-Level Metrics

  • Latency: Time to first token, total request duration
  • Costs: Token usage and dollar costs per request
  • Model Performance: Completion quality, error rates
  • Request/Response Data: Full input prompts and completions
  • Metadata: Timestamps, model versions, provider info

Aggregate Metrics

  • Total Costs: Track spending across all requests
  • Usage Patterns: Requests per day, peak usage times
  • Error Rates: Failed requests, rate limits, timeouts
  • User Analytics: Cost and usage per user
  • Session Analytics: Multi-turn conversation metrics

Dashboard Features

Helicone Dashboard

Real-Time Monitoring

The Helicone dashboard provides:
  • Live request feed - See requests as they happen
  • Cost tracking - Real-time spend monitoring
  • Performance charts - Latency and throughput metrics
  • Filtering and search - Find specific requests instantly
  • Custom time ranges - Analyze any time period

Analytics Views

Different views for different insights:
  • Requests View - Detailed log of all LLM calls
  • Sessions View - Grouped conversation flows
  • Users View - Per-user cost and usage analytics
  • Models View - Compare performance across models
  • Properties View - Segment by custom metadata

Getting Started

1

Integrate Helicone

Add Helicone headers to your LLM requests or use our proxy
2

View Requests

See all logged requests in your dashboard at helicone.ai/requests
3

Add Metadata

Enrich requests with custom properties and user IDs
4

Analyze Patterns

Use filters, sessions, and traces to understand your application

Use Cases

Track performance and costs in production. Set up alerts for errors, high costs, or latency spikes.
Inspect failed requests, compare inputs/outputs, and trace complex workflows to identify issues.
Identify expensive operations, compare model costs, and track spend per user or feature.
Understand how users interact with your AI features. Track engagement, popular flows, and drop-off points.
Maintain complete logs of all LLM interactions for compliance and security auditing.

Next Steps

Log Your First Request

Start capturing LLM requests

Track Sessions

Group related requests together

Add Custom Properties

Enrich requests with metadata

Monitor Users

Track per-user analytics

Build docs developers (and LLMs) love