feat(frontend): admin LLM registry editing

[ntindle]

Why

Part 9 of 9 of Phase A of the LLM registry restack (#13605→#13606→#13607→#13608→#13609→#13610→#13611→#13612→this). With this PR the whole workflow that #13596 tried to do by hand-editing a machine-generated JSON becomes: open /admin/llms, add the model (HIDDEN if pre-launch), set routing cells when ready — validated, audited, propagated to every install within a day.

What

Editing controls on the read-only dashboard from #13612, re-imagined from @Bentlybro's #12468 modals under current conventions:

• Models: add/edit dialogs covering every column (visibility, min subscription tier, fallback-model picker, capability flags, price tier, enabled/recommended); delete and disable flows with the replacement-model picker (pre-filled from fallbackModelSlug)
• Creators: add/edit/delete
• Routing matrix: each copilot cell gets a model dropdown (enabled models; HIDDEN allowed and badged) + clear-cell; the API's capability warnings (e.g. non-reasoning model in a thinking cell) surface inline as warnings per the agreed design — warn, don't block
• Migrations: revert with confirmation
• All mutations via generated Orval hooks with proper query-key invalidation and toast errors; no server actions, no legacy components

Verification

• 12/12 /admin/llms integration tests (5 read + 7 new editing scenarios: create/edit request payload assertions, disable-with-replacement, routing PUT + warning surfacing, revert confirm, 500 toast)
• pnpm format / pnpm lint / pnpm types clean; full-suite failures are the pre-existing multi-suite interference documented on feat(frontend): admin LLM registry dashboard (read-only) #13612 (worse on base without these changes, verified by stash-comparison)

Checklist

• Design-system components only; generated hooks only
• Out-of-scope changes: none

[coderabbitai]

Important

Review skipped

Draft detected.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: 3cc6feca-7128-433a-824b-d9b2e00ca6d8

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch feat/llm-registry-v2-admin-ui-write

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[codecov]

Codecov Report

❌ Patch coverage is 66.66667% with 75 lines in your changes missing coverage. Please review.
✅ Project coverage is 76.02%. Comparing base (58630e3) to head (6107c43).

Additional details and impacted files

@@                         Coverage Diff                          @@
##           feat/llm-registry-v2-admin-ui-read   #13613    +/-   ##
====================================================================
  Coverage                               76.01%   76.02%            
====================================================================
  Files                                    2694     2706    +12     
  Lines                                  204324   204542   +218     
  Branches                                19661    19693    +32     
====================================================================
+ Hits                                   155324   155500   +176     
- Misses                                  44713    44743    +30     
- Partials                                 4287     4299    +12     

Flag	Coverage Δ
platform-frontend	46.36% <66.66%> (+0.11%)	⬆️
platform-frontend-e2e	31.31% <ø> (+0.27%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Components	Coverage Δ
Platform Backend	82.83% <ø> (ø)
Platform Frontend	50.24% <66.66%> (+0.18%)	⬆️
AutoGPT Libs	∅ <ø> (∅)
Classic AutoGPT	28.43% <ø> (ø)

🚀 New features to boost your workflow:

• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[ntindle]

After a post-build design review we pivoted from the DB-registry to catalog-as-code: one canonical, schema-validated catalog file is the source of truth (models + costs + copilot routing), updated by PR (catalog-only diffs may ride hotfix→master; the /review bot covers the lane), propagated by CD, with LaunchDarkly demoted to optional cohort experiments and git history as the audit log. Rationale: the admin machinery mostly bought instant-mutation and non-engineer authoring we don't need, and LD is too unreliable to be the load-bearing control layer.

No replacement by design — editing happens in the catalog file by PR.