Skip to content

fix(help): recommend a resolvable model and label bare rocm as the launcher (EAI-8011) - #296

Merged
fredespi merged 1 commit into
mainfrom
rocm-help-recommends-a-nonexistent-model-id-and
Aug 24, 2026
Merged

fix(help): recommend a resolvable model and label bare rocm as the launcher (EAI-8011)#296
fredespi merged 1 commit into
mainfrom
rocm-help-recommends-a-nonexistent-model-id-and

Conversation

@fredespi

Copy link
Copy Markdown
Collaborator

Summary

Two help-text defects surfaced by the README walkthrough (EAI-8011), fixed in apps/rocm/src/main.rs. Help/doc text only — no behaviour change.

  • The worked serve examples named a model the CLI cannot resolve. rocm serve qwen2.5-7b-instruct is neither a catalog name nor an owner/repo reference, so copying an example verbatim fails. Replaced all six help/doc occurrences with qwen, a real catalog alias for the built-in assistant (Qwen3-4B-Instruct-2507-GGUF) that rocm model resolves. Realigned the top-level EXAMPLES column for the shorter name.
  • The help described bare rocm as opening the dashboard — the same thing it says rocm dash does. Running rocm with no subcommand actually opens the launcher (launch_defaultrun_launcher), so the long_about now says "interactive launcher (TUI)". The dash command's own help row still describes the dashboard, so the two read distinctly.

These are the two EAI-8011 rows pinned as expected failures in #241; once both land those rows XPASS and retire themselves.

Test plan

  • Built rocm on Linux (macOS can't build this repo — known platform-gated st_mode failure in rocm-engine-lemonade).
  • Ran rocm help, rocm serve --help, rocm model --verbose against the built binary at wide terminal width; confirmed every rocm serve example names qwen (listed in model --verbose as aliases=[qwen, ...]) and the no-subcommand sentence no longer says "dashboard" while the dash row still does.
  • Linux workspace gate (clippy + full test suite) green apart from an unrelated full-disk environment failure.

@fredespi
fredespi requested a review from a team as a code owner August 21, 2026 14:25
…uncher (EAI-8011)

The help's worked examples named `rocm serve qwen2.5-7b-instruct`, which is
neither a catalog name nor an owner/repo reference, so copying an example
verbatim fails. Use `qwen`, a real catalog alias for the built-in assistant
that `rocm model` resolves.

The long help also described running `rocm` with no subcommand as opening the
dashboard — the same thing it says `rocm dash` does. Bare `rocm` opens the
launcher; describe it as such so the two read distinctly.

Signed-off-by: fredespi <fredrik.espinoza@gmail.com>
@fredespi
fredespi force-pushed the rocm-help-recommends-a-nonexistent-model-id-and branch from 0294ce9 to fbdee8b Compare August 21, 2026 14:35

@r0x0r r0x0r left a comment

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Reviewed. Doc/help-text only, no behaviour change, and both fixes check out against the code:

  • qwen is a real, resolvable catalog alias — BUILTIN_ASSISTANT_MODEL_ALIAS = "qwen" maps to Qwen3-4B-Instruct-2507-GGUF (apps/rocm/src/providers.rs), so every replaced rocm serve qwen / rocm chat --model qwen example now names something the CLI can actually resolve, unlike the previous qwen2.5-7b-instruct.
  • The long_about wording change is accurate: bare rocm routes to launch_defaultrun_launcher (the launcher/TUI), and the distinct dash row still describes the dashboard, so the two no longer read identically.

All six occurrences are consistent and the EXAMPLES column realignment looks right. No substantive concerns.

@r0x0r r0x0r left a comment

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Approving. The two EAI-8011 help-text defects are fixed correctly: every serve/chat example now names qwen, a real resolvable alias (BUILTIN_ASSISTANT_MODEL_ALIAS -> Qwen3-4B-Instruct-2507-GGUF), and the bare-rocm long_about now accurately says launcher (launch_default -> run_launcher) while the dash row still describes the dashboard. Doc-only, no behaviour change, 24/24 CI green.

@fredespi
fredespi added this pull request to the merge queue Aug 24, 2026
Merged via the queue into main with commit 6931b30 Aug 24, 2026
24 checks passed
@fredespi
fredespi deleted the rocm-help-recommends-a-nonexistent-model-id-and branch August 24, 2026 15:48
fredespi added a commit that referenced this pull request Sep 9, 2026
Brings the branch up to date with main and addresses the review findings that
depend on it.

Stale expectation rows deleted (their bugs are fixed on main); the scenarios
stay as guards:
- help-serve-example-names-a-resolvable-model, help-describes-the-default-command
  (#296 made the worked example 'rocm serve qwen' and relabelled bare 'rocm' the
  launcher)
- serve-rejects-no-advertised-device-policy (#304 hid cpu_only from the
  advertised list)
- diagnose-commands-name-a-real-group (#302 filters the UNKNOWN group lookup)

Verified in a Linux container on two hosts: 0 XPASS, 0 stale.

Other review fixes:
- advertised_device_policies reads clap's block rendering as well as the inline
  one, so the scenario keeps measuring the contract instead of tripping its own
  guard
- run_rocm_without_env deleted (no call sites)
- the services-stop comment now matches its expectations.toml row
- 'not the leading remedy' reworded to 'not offered as a cause', which is what
  the assertion checks
- the non-default device group is chosen from plausible device groups rather
  than whichever /etc/group lists first
- scenario numbers made unique and prefixed after main's renumbering
- new unit test rejects an expectations.toml row whose scenario no longer exists

Signed-off-by: fredespi <fredrik.espinoza@gmail.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants