feat(ui): require embedding model for semantic auto router

[devin-ai-integration]

Relevant issues

Linear ticket

Pre-Submission checklist

• I have added meaningful tests
• My PR passes all CI/CD checks (e.g., lint, format, unit tests)
• My PR's scope is as isolated as possible; it only solves 1 specific problem
• I have received a Greptile Confidence Score of at least 4/5 before requesting a maintainer review

Screenshots / Proof of Fix

This is a UI-only validation change on the Admin UI Add/Edit Auto Router flow, so there is no LLM call to make. To verify against a live proxy running the Admin dev server (npm run dev in ui/litellm-dashboard, http://localhost:3000):

1. Go to Models -> Add Model -> Add Auto Router and pick "Semantic Router"
2. Fill in an Auto Router Name, a Default Model, and at least one route, but leave Embedding Model empty
3. Click "Add Auto Router" and observe the blocking error "Please select an Embedding Model"; the form does not submit
4. Select an Embedding Model and submit; the router is created as before
5. Edit an existing semantic router and clear the Embedding Model; "Save Changes" is blocked with "Embedding model is required"

Captured at commit c8b2f06

Type

🆕 New Feature

Changes

Semantic routing matches requests against example utterances via an embedding model, so the embedding model is functionally required, yet the Add/Edit Auto Router UI marked it optional (placeholder "(optional)", allowClear, no validation rule). This makes it a mandatory field for the Semantic Router type

The submit-time validations for the semantic router path were previously inline any-typed checks in add_auto_router_tab.tsx. They are extracted into a pure, strongly typed getSemanticRouterError helper (mirroring the existing getSemanticConfigError for the complexity router), which now also rejects a missing embedding model. Validation order is default model -> routes present -> embedding model -> per-route completeness

getSemanticRouterError({ defaultModel, embeddingModel, routerConfig }):
  if !defaultModel                              -> "Please select a Default Model"
  if no routes                                  -> "Please configure at least one route ..."
  if !embeddingModel                            -> "Please select an Embedding Model"
  if any route missing name/description/uttrs   -> "Please ensure all routes ..."
  else null

The Embedding Model Form.Item in both the add tab and the edit modal now carries a required rule, drops allowClear, and updates the tooltip/placeholder to no longer say "optional"

Final Attestation

• The tests check the right things, including the edge cases, and regressions in the respective real-world customer use-cases are not possible after this PR

Link to Devin session: https://app.devin.ai/sessions/609d5dc8268347baa38c74a615ddf333

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment, CI, and merge conflict monitoring

[greptile-apps]

Greptile Summary

This PR makes the Embedding Model field mandatory when creating or editing a Semantic Auto Router in the Admin UI. It extracts the formerly ad-hoc, any-typed validation into a dedicated getSemanticRouterError helper, mirrors the pattern of the existing getSemanticConfigError for the complexity router, and adds a focused test suite.

• build_semantic_router_validation.ts introduces a strongly-typed helper that enforces a deterministic validation order (default model → routes present → embedding model → per-route completeness) and returns a typed string | null.
• Both the Add tab (add_auto_router_tab.tsx) and Edit modal (edit_auto_router_modal.tsx) drop allowClear from the embedding model selector and add an Ant Design required: true rule, so the field is blocked at both the submit-handler level and the form-validation level.

Confidence Score: 4/5

Safe to merge; changes are UI-only and well-contained, with no backend impact.

The core validation logic is correct, well-tested, and isolated to a single new module. The two observations are minor: the edit modal's generic catch toast becomes slightly misleading now that a form-validation failure (missing embedding model) can trigger it, and the test suite does not exercise null/undefined routerConfig inputs even though the type signature allows them. Neither issue affects production correctness today.

edit_auto_router_modal.tsx — the catch block in handleSubmit now surfaces a misleading toast for what is actually a client-side validation rejection.

Important Files Changed

Filename	Overview
ui/litellm-dashboard/src/components/add_model/build_semantic_router_validation.ts	New strongly-typed validation helper getSemanticRouterError extracted from inline ad-hoc checks; correctly enforces embedding model requirement and validates field order as specified.
ui/litellm-dashboard/src/components/add_model/build_semantic_router_validation.test.ts	Six focused unit tests cover all validation branches; missing coverage for routerConfig: null and routerConfig: undefined inputs but those are handled by optional chaining in the implementation.
ui/litellm-dashboard/src/components/add_model/add_auto_router_tab.tsx	Inline validation replaced with getSemanticRouterError; embedding model Form.Item gains required: true rule and drops allowClear; double-validation with Ant Design is harmless given the early-return guard.
ui/litellm-dashboard/src/components/edit_auto_router/edit_auto_router_modal.tsx	Embedding model field gains required: true and loses allowClear; when Ant Design form validation rejects the empty field the catch block surfaces a misleading "Failed to update" toast rather than a validation-specific message.

Comments Outside Diff (1)

1. ui/litellm-dashboard/src/components/edit_auto_router/edit_auto_router_modal.tsx, line 283-286 (link)

   Misleading toast when form validation rejects

   Now that the embedding model carries a required: true rule, form.validateFields() can throw a validation-error object (not a network/server error) when the field is empty. The catch block unconditionally shows "Failed to update auto router configuration", which conflicts with the inline "Embedding model is required" indicator already rendered by Ant Design on the field. A user who sees both would reasonably wonder whether the save actually failed on the backend.

   Consider narrowing the toast to server errors only — Ant Design's validateFields rejection has errorFields on it, so it is distinguishable.

   Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!

_{Reviews (1): Last reviewed commit: "feat(ui): require embedding model for se..." | Re-trigger Greptile}

[greptile-apps]

null/undefined routerConfig not exercised

The type signature accepts SemanticRouterConfig | null | undefined, and the implementation handles those via routerConfig?.routes. A test passing routerConfig: null (and one for undefined) would pin down that both resolve to the "Please configure at least one route" branch and prevent a future refactor from silently breaking the null guard.

[devin-ai-integration]

✅ Devin Review: No Issues Found

Devin Review analyzed this PR and found no potential bugs to report.

View in Devin Review to see 1 additional finding.

[codspeed-hq]

Merging this PR will not alter performance

✅ 31 untouched benchmarks

---

_{Comparing litellm_semantic_router_embedding_required_ui (c8b2f06) with litellm_internal_staging (b2202cb)}

[devin-ai-integration]

Runtime QA

Tested end-to-end against a live proxy at http://localhost:4000/ui serving a production build of this branch, logged in as admin. Session: https://app.devin.ai/sessions/609d5dc8268347baa38c74a615ddf333

Add Auto Router (Semantic Router) now requires an embedding model — PASSED

With a valid name, one complete route, and a Default Model selected but Embedding Model left empty, clicking "Add Auto Router" is blocked with "Please select an Embedding Model" and no router is created.

After selecting voyage-4-large, submission succeeds and the router is created.

DB confirms auto_router_embedding_model is persisted for the created qa_semantic_router.

Edit Auto Router — untested via UI

Created auto routers do not surface in the "Models + Endpoints" table in this build (confirmed via /v1/model/info and a full scroll of "All Available Models"), so the edit modal has no row to open from. The Edit modal's required-embedding change is covered by the code and edit_auto_router_modal.test.ts, but was not runtime-verified through the UI.