Agent auto‑scaling to zero in Codeac: impact on low‑traffic workloads

Short answer

As of the latest public information (2024), Codeac does not document an auto‑scale‑to‑zero feature for its analysis agent. The agent is either invoked on‑demand per CI job (managed SaaS) or runs as a long‑lived self‑hosted service. There is no configuration option in codeac.yml or the CI provider settings that controls an idle timeout before scaling down.

Likely explanation vs. confirmed facts

Aspect	Likely explanation (based on community & docs)	Confirmed fact
Agent lifecycle	Managed SaaS agent spins up a fresh container for each analysis job; self‑hosted runners stay alive until the host stops them.	Codeac’s public docs describe the agent as a CI step, not a persistent service.
Idle‑timeout configuration	No idleTimeout, scaleToZero, or similar keys exist in codeac.yml.	Configuration reference lists only analysis scope, triggers, and rule sets.
Billing model	Pricing is per seat / per active repository (or per analysis count), not per compute minute.	Billing dashboard shows line items for seats and repositories, not agent uptime.
Cold‑start impact	First run after a long idle period may incur container pull/start latency (seconds).	Observed in CI logs when the Codeac step pulls the Docker image each job.

Steps to verify for your organization

1. Check the official docs – search for “agent”, “scaling”, “idle”, or “billing” on docs.codeac.io.
2. Inspect your CI pipeline logs – look at the Codeac step: does it pull a new image each run? (Indicates per‑job agent.)
3. Review the billing dashboard – confirm whether compute minutes are itemized separately from seats/repos.
4. Open a support ticket or ask in the Codeac Slack/Discord – request a definitive statement on scale‑to‑zero support and any future roadmap.

One missing diagnostic detail

Are you using the managed SaaS agent (default) or a self‑hosted runner? The answer changes the recommendation: managed SaaS already behaves like scale‑to‑zero per job, while a self‑hosted runner would need explicit shutdown logic to avoid idle costs.

Practical take‑aways for low‑traffic workloads

• If you run the managed SaaS integration, you already get “scale‑to‑zero” semantics – each pipeline run starts a fresh agent, so idle time costs nothing.
• If you operate a self‑hosted agent, you must implement your own idle‑shutdown (e.g., a cron that stops the container after N minutes of inactivity) to avoid paying for the underlying VM/container runtime.
• Expect a cold‑start penalty of a few seconds on the first analysis after a long gap; this is usually negligible for intermittent pipelines.

Assumptions & uncertainty

• Information reflects Codeac’s public state as of 2024; pricing or agent features may have changed since.
• Self‑hosted deployments (if offered) have different scaling characteristics and are not covered by the managed SaaS behavior.
• No official SLA or documentation confirms an auto‑scale‑to‑zero flag; treat the above as best‑effort guidance pending vendor confirmation.