AI observability
Understand what your AI application is doing.
Monitor requests, traces, latency, token usage, and costs in one place.
Instrument once. See everything.
Aetherion collects a trace for every model call your application makes. Install the SDK, point it at a project, and requests start appearing with their provider, model, token usage, latency and estimated cost — no gateway, no proxy, no changes to how you call your providers.
app.py
import meridian_ai
meridian = meridian_ai.MeridianAI()
# your existing AI calls are traced
response = client.chat.completions.create(...)Capabilities
Traces
Every model call captured as a structured trace with provider, model and status.
Latency
See how long each request takes so slow paths surface before your users find them.
Token usage
Input, output and total tokens per request, project and model.
Costs
Estimated spend broken down by project and provider, updated as traces arrive.
Start monitoring your AI application
Create a project, complete SDK onboarding, and receive your first trace.