DEVOPS

Agent-Driven Cardinality Investigation and Sampling Proposal

An agent investigates a flagged Honeycomb cardinality spike end to end: it runs shell queries against the schema, reasons about whether the dimension is legitimately needed.

CategoryDevOps
Enginepaperclip
Difficultyadvanced
Triggermanual
Steps5
Setup~25 min

How it runs

The automated pipeline, trigger to output.

  • TriggerOperator manually starts an investigation for a dataset/dimension
  • ActionRun shell-based Honeycomb queries on distinct values and usageShell
  • LogicAgent judges whether to sample, drop, or keep the field
  • ActionDraft the matching sampling rule or column-drop changeHoneycomb
  • OutputOpen GitLab MR with the change and written rationaleGitLabGitLab

What it does

This is the judgment-heavy version of the guard. Instead of a fixed threshold, an agent investigates a flagged dimension, decides whether it is genuinely useful for queries or just noise, and proposes the right remedy: sample it, drop it, or leave it with a justification.

When to use it

Use it for the hard cases where a simple threshold over-fires, for example a `customer_id` you actually slice by versus a `trace_id` that should never be a Honeycomb dimension. The agent writes up why.

How it works

  1. 1A manual trigger starts an investigation against a named dataset and dimension.
  2. 2The agent runs shell-based Honeycomb queries to inspect distinct values, query usage of the field, and event-volume contribution.
  3. 3It reasons about whether the dimension earns its cardinality cost or should be sampled or dropped.
  4. 4It drafts the matching change: a Refinery sampling rule or a column-drop proposal.
  5. 5It opens a GitLab MR containing the change plus a written rationale and the supporting query evidence for human approval.

Set it up

What you configure once, before turning it on.

  1. 1
    Connect HoneycombDistributed traces and queries.
  2. 2
    Connect ShellRun sandboxed commands inside the workspace.
  3. 3
    Connect GitLabRepos, MRs, pipelines, registry.
  4. 4
    Set each agent's modelWe leave models unset so you pick the tier — fast + cheap, or top-quality.
  5. 5
    Tune it to your dataEdit the prompts, filters, and field mappings so it matches how your team works.
  6. 6
    Test, then turn it onRun once against a sample, confirm the output, then enable the trigger.

Run this workflow in your colony.

14-day trial. No DevOps. No Sales call. Provisioned in under a minute.