fix(ui): show team BYOK models in team fallback settings

[devin-ai-integration]

TLDR

Problem this solves:

• Team fallback UI hid the team's own BYOK models
• Team settings loaded the proxy wide model listing

How it solves it:

• Team settings now passes its team id down
• Fallback options come from /models?team_id=...

Relevant issues

Linear ticket

Resolves LIT-5299

Pre-Submission checklist

Please complete all items before asking a LiteLLM maintainer to review your PR

• 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 (Greptile reviews automatically once the PR is opened; only comment @greptileai to re-request a review after pushing changes)

Delays in PR merge?

If you're seeing a delay in your PR being merged, ping the LiteLLM Team on Slack (#pr-review).

Screenshots / Proof of Fix

Repro and verification against a live proxy on localhost:4000 with the Admin UI dev server on localhost:3000, using a team that owns a BYOK wildcard deployment (model_name: openai/* with model_info.team_id set)

Before, at commit ca2ed49e2a:

1. Go to http://localhost:3000/teams/?team=<TEAM_ID>, Settings, Edit Settings, Router Settings, Fallbacks
2. Open the Primary Model dropdown. Only gpt-5-mini is listed, none of the team's openai/* models appear

The backend already exposes both listings, which is what the dropdown difference comes down to:

curl -s http://localhost:4000/models -H "Authorization: Bearer $LITELLM_MASTER_KEY" | jq '.data | length'
1
curl -s "http://localhost:4000/models?team_id=<TEAM_ID>" -H "Authorization: Bearer $LITELLM_MASTER_KEY" | jq '.data | length'
213
curl -s "http://localhost:4000/models?team_id=<TEAM_ID>" -H "Authorization: Bearer $LITELLM_MASTER_KEY" | jq -r '.data[].id' | grep -E '^openai/(\*|gpt-5\.2)$'
openai/*
openai/gpt-5.2

After, at commit 4166054, same steps:

3. Typing openai in Primary Model now lists openai/* and every expansion; pick openai/gpt-5.2
4. In Fallback Chain, type openai/* and select it
5. Click Save Changes. The UI shows "Team settings updated successfully" and the saved router settings persist on reload

Re-verified the same flow at commit 49efcfb after the react-query refactor, this time picking openai/gpt-5.6 with an openai/* fallback chain: the dropdown lists the team's openai/* expansions, the save succeeds, and /team/info returns the persisted fallbacks. Screenshots are in the PR comments

Type

🐛 Bug Fix
🧹 Refactoring

Changes

RouterSettingsAccordion is shared by the global Router Settings page and the per team settings form, and it always built its fallback options from fetchAvailableModels, which hits /model_group/info. That listing resolves with no team id, so Router.get_model_names drops team scoped rows and a team's own BYOK models never showed up, even while you were editing that exact team.

The accordion now takes an optional teamId and, when present, loads options from the team aware /models?team_id=... route through a new fetchAvailableModelsForTeam helper. TeamInfo passes the team id it already has. Global Router Settings passes nothing, so its behavior is unchanged.

The model fetch now goes through react-query instead of a hand-rolled effect: useQuery keyed on the access token and team id replaces the local state plus manual stale-response guard, since the dashboard already ships @tanstack/react-query with a provider at the app root. Out of order responses are handled by query keying, and the component tests render under a QueryClientProvider

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/c2770b07352e4073b24895ec26693f2b

[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

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you all sign our Contributor License Agreement before we can accept your contribution.
1 out of 2 committers have signed the CLA.

✅ ryan-crabbe-berri
❌ devin-ai-integration[bot]
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[greptile-apps]

Greptile Summary

The PR fixes team fallback settings by loading team-scoped models through a team-aware query while preserving the proxy-wide behavior for global router settings

• Passes the current team ID into RouterSettingsAccordion
• Adds a helper that converts /models?team_id=... results into fallback model groups
• Uses a team-specific TanStack Query key so late responses from previously selected teams cannot replace the current options
• Adds coverage for team-scoped models, global fallback behavior, and overlapping team requests

Confidence Score: 5/5

The PR appears safe to merge

No blocking failure remains

Important Files Changed

Filename	Overview
ui/litellm-dashboard/src/components/common_components/RouterSettingsAccordion.tsx	Replaces manually managed model-fetch state with a query keyed by access token and team ID, resolving the previously reported stale-response race
ui/litellm-dashboard/src/components/llm_calls/fetch_models.tsx	Adds a team-aware model-list helper that deduplicates, filters, sorts, and maps model IDs for fallback selection
ui/litellm-dashboard/src/components/team/TeamInfo.tsx	Passes the active team ID into the shared router settings accordion
ui/litellm-dashboard/src/components/common_components/RouterSettingsAccordion.test.tsx	Covers team and global model sources and verifies that a late response for an old team does not replace the current team's options
ui/litellm-dashboard/src/components/llm_calls/fetch_models.test.tsx	Verifies team-scoped endpoint arguments, sentinel filtering, deduplication, sorting, and empty responses

_{Reviews (3): Last reviewed commit: "refactor(ui): use react-query for fallba..." | Re-trigger Greptile}

[devin-ai-integration]

Tested end to end against a local proxy with a team that owns a BYOK wildcard deployment openai/* registered with model_info.team_id, hitting the real OpenAI API.

Setup: a proxy-level model demo-proxy-model (openai/gpt-4.1-mini with a deliberately broken key) assigned to that team, plus the team's own BYOK wildcard with a working key. All requests use a team-scoped virtual key.

Before configuring anything, the failing model has nothing to fall back to:

$ curl -sS -i -X POST http://localhost:4000/v1/chat/completions \
    -H "Authorization: Bearer $TK" -H "Content-Type: application/json" \
    -d '{"model":"demo-proxy-model","messages":[{"role":"user","content":"Say hello in 3 words."}]}'

HTTP/1.1 401 Unauthorized
{"error":{"message":"litellm.AuthenticationError: ... Received Model Group=demo-proxy-model\nAvailable Model Group Fallbacks=None","code":"401"}}

openai/gpt-4.1-mini is only offered in the Fallback Chain because of this PR. The old team-agnostic listing does not have it:

$ curl -s -H "Authorization: Bearer $KEY" localhost:4000/model_group/info | jq -c '[.data[].model_group]'
["gpt-5-mini","demo-proxy-model"]

In Team Settings > Edit Settings > Router Settings > Fallbacks, the team's BYOK model shows up and is selectable:

Same request now succeeds, and the headers prove it landed on the team's BYOK deployment at real OpenAI

HTTP/1.1 200 OK
x-litellm-model-id: 225329bb-bb05-4697-8d40-faf416f2bb92
x-litellm-model-api-base: https://api.openai.com
x-litellm-model-group: openai/gpt-4.1-mini
x-litellm-attempted-fallbacks: 1
{"model":"gpt-4.1-mini-2025-04-14","choices":[{"message":{"content":"Hello there, friend!","role":"assistant"}}],...}

x-litellm-model-id is the team-owned BYOK wildcard deployment, so the fallback configured in the UI is what served the request

Persistence

$ curl -s -H "Authorization: Bearer $KEY" "localhost:4000/team/info?team_id=$TEAM" | jq -c .team_info.router_settings.fallbacks
[{"demo-proxy-model":["openai/gpt-4.1-mini"]}]

Regression: global admin Router Settings unchanged

Still loads the proxy-wide listing only (demo-proxy-model, gpt-5-mini), searching openai returns "No data", no console errors

Dashboard unit tests: 7 passed, including the new stale-response case

[ryan-crabbe-berri]

@greptileai re review

[ryan-crabbe-berri]

Re-QA at 49efcfb after the react-query refactor: the dropdown lists the team's openai/* expansions, the fallback saves and persists after reload

[ryan-crabbe-berri]

@greptileai

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!