TestRelic AI
Go to App
Cloud Platform

Modules

Automatic, AI-assisted grouping of your test cases by functional area — Authentication, Dashboard, Checkout, and beyond.

Modules is a fourth view alongside Suites in a repository's Test Cases tab. Where Suites reflects how your test files are organized, Modules groups tests by functional area — Authentication, Dashboard, Products, Checkout — regardless of which file or suite they live in.

Test run ingested
Local embedding classifier
Modules view + Overview panel

How grouping works

Every test case is classified automatically, for free, with no configuration required:

  • Local embedding classifier — Runs in-process on the platform, not against an LLM. There's no external API call and no token cost. Tests are placed by nearest-centroid similarity against your repo's existing modules.
  • Deterministic naming — When a new module is created automatically, its name is derived from the test's file folder, describe block, and a keyword dictionary — not an AI guess.
  • Runs on every ingest — Classification happens right after a run lands, so new tests are grouped without you doing anything.
  • Per-repo toggle — Automatic classification can be turned off in Repo Settings. It's on by default.

Manual edits always win

Any assignment you make by hand — moving a test, renaming a module, or merging two modules — is recorded as a manual change. The automatic classifier never overwrites a manual assignment, even if it reclassifies the repo later.

Managing modules

From the Modules view you can:

  • Create, rename, merge, or delete a module.
  • Reassign tests individually or in bulk, from the row-level menu or a multi-select action.
  • See module origin at a glance — every module header carries a badge:
BadgeMeaning
ManualCreated, or renamed, by a person
AI generatedCreated by the local classifier or the AI grouper

Renaming an AI-generated module immediately flips it to Manual — once you've curated a name, the platform treats it as yours.

Tests the classifier can't confidently place — messy titles, no discoverable file path — are left in Ungrouped rather than forced into the wrong bucket.

Auto Group with AI

For the tests left in Ungrouped, the Auto Group with AI button (in the Modules toolbar) hands them to an LLM as a one-shot pass:

  • Batches Ungrouped tests and asks the model to place them, giving it your repo's existing module names as preferred vocabulary — so it reuses "Authentication" instead of minting "Auth".
  • Writes land as AI-generated assignments, so a later manual correction still takes precedence.
  • Counts against your organization's AI usage allowance, including calls that don't return a usable classification — the cost is incurred either way.
  • Rate-limited per user, since a single press can trigger several model calls.

The toast after a run distinguishes "nothing could be grouped confidently" from "everything is already grouped", so you know whether it's worth pressing again after new tests land.

Ask AI: chat to change the grouping

Auto Group with AI is one-shot — you take the result. For anything more targeted, use Ask AI from the Modules toolbar to open a docked assistant panel scoped to this repo's modules. You can ask it directly, for example:

  • "Merge Checkout into Products"
  • "Create a Notifications module for anything under tests/notifications/"
  • "Move the three flaky login tests into Authentication"

The assistant has real write access — list, create, rename, merge, delete, and move — not just the ability to suggest changes. It resolves modules by name (not ID), so you never need to look up a module identifier to ask for a change. Like manual edits, everything it does is recorded as a human-directed change and won't be undone by the automatic classifier. Deleting a module never deletes its tests — they move to Ungrouped.

See Ask AI for more on how the assistant works across the platform.

Grouping coverage and insights on Overview

The repo Overview tab includes a Modules insights panel that answers "which group should I look at first?" — a different question from the Modules view itself, which answers "what's in each group?"

  • Grouping coverage — Grouped vs. Ungrouped, shown first. If a meaningful share of tests are Ungrouped, every metric below it only describes part of the repo, and the panel says so explicitly.
  • Per-module health — Pass rate, failure count, and flakiness per module, sorted by failure count rather than lowest pass rate — one failure in a two-test module is a 50% failure rate, but it's rarely the repo's biggest problem.
  • Suggested merges — Modules whose content looks similar (by embedding similarity) are flagged as probable duplicates — for example, "Cart" and "Checkout" testing overlapping flows — so you can merge them in one click from the Modules view.

Where else Modules shows up

  • APIGET /repos/:repoId/tests includes each test's moduleId, so any client — the CLI Studio Modules view, the MCP server, or your own integration — can render module placement without a separate lookup.
  • CLI / Studio — Tests can be grouped by the same AI-classified Modules in both Local and Cloud mode. See Modes & engines.

Beta feature

Modules is a recent addition and still evolving — expect refinements to grouping accuracy and the insights panel in upcoming releases.

Was this page helpful?

On this page