Bounded GenAI metrics from Otel traces with O raw prompts
Source
github.com
Author
kwisatzh
Date
Context
otelcol-genai-sketches helps operators locate heavy AI usage without creating a separate monitoring label for every user or prompt. It converts OpenTelemetry activity traces into bounded metrics and approximate summaries, using keyed hashes to group prompt signatures without exporting raw prompt text.
Comparing request counts with tokenThe chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.Full definition → per request helps distinguish more traffic from larger requests. The collector also counts missing usage reports separately, so absent measurements do not masquerade as zero consumption. These signals can locate where usage accumulates; the collector itself neither enforces spending limits nor stops AI agentAn AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.Full definition → loops.
Terms in this piece · Glossary
token — The chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
