Getting started¶
Install¶
pip install latenzy
Configure¶
Write a latenzy.yaml. API keys are read from environment variables only — they
have no place in a config file.
providers:
- provider: anthropic # key from ANTHROPIC_API_KEY
endpoint: direct
models:
- claude-sonnet-4-6
- claude-haiku-4-5-20251001
- provider: openai # key from OPENAI_API_KEY
endpoint: direct
models:
- gpt-4o
- gpt-4o-mini
- provider: gemini # key from GEMINI_API_KEY
endpoint: direct
models:
- gemini-2.0-flash
probe:
interval_seconds: 300 # one probe cycle every 5 minutes
max_output_tokens: 16 # keep probe cost negligible
prompt_classes: [small]
exporter:
host: 127.0.0.1 # loopback needs no auth
port: 9877
See Configuration for every field.
Run¶
export ANTHROPIC_API_KEY=... OPENAI_API_KEY=... GEMINI_API_KEY=...
latenzy doctor -c latenzy.yaml # validate config, check keys are present
latenzy once -c latenzy.yaml # one probe cycle, human-readable results
latenzy run -c latenzy.yaml # probe on an interval + serve /metrics
latenzy once prints a per-model table; latenzy run serves Prometheus metrics
on http://127.0.0.1:9877/metrics. Point Prometheus at it and import the
dashboard.
No existing Prometheus/Grafana? The standalone bundle runs the whole stack with one command.