Data collection
vLLM
vLLM inference: TTFT, GPU utilisation, KV-cache hit rate, batch size from /metrics.
- Metrics
- Agentless
- Package 1.1.0
What it collects
- Signals
- Metrics
- How it collects
- Prometheus scraping (
prometheus) - Where it runs
- Agentless: your workspace's cloud gateway can run it, with nothing to install. Collectors you install can run it too.
- Package version
- 1.1.0
How to connect it
- 01Open the connect wizardIn the Tracix console, open Integrations and pick vLLM.
- 02Choose where it runsAgentless runs it on your workspace's cloud gateway. Pick a policy to run it on collectors you install.
- 03Fill in the connectionEnter the settings the wizard asks for. Collectors pick up the policy change on their own.
More in AI & LLM
All packages in AI & LLM
OpenAI
AI & LLM
OpenAI usage and spend from the Admin Usage & Cost API — per-model input/output/cached/audio token counts, request volume, and USD cost, as metrics and…
Metrics · logs
Anthropic
AI & LLM
Claude usage and spend from the Anthropic Admin Usage & Cost API — per-model token counts, cache hit rate, web-search calls, and daily USD cost, as metrics…
Metrics · logs
Connect vLLM to Tracix
Start free and connect it from the console, or book a demo and we will walk through your environment with you.