fix(proxy): list public team model name in /v1/models by ishaan-berri · Pull Request #30588 · BerriAI/litellm

ishaan-berri · 2026-06-17T01:08:02Z

What changed

/v1/models and /models build their list from access-group / key-model expansion, which surfaces the internal routing key model_name_{team_id}_{uuid} for team-scoped (BYOK) models rather than an OpenAI-compatible model id. /v1/model/info already translates these to the public team_public_model_name; this makes the listing endpoints consistent.

By default the listing now returns the human-friendly team_public_model_name. Operators that scripted against the legacy internal keys can opt out with general_settings.use_team_public_model_name: false; the routing key still works for API calls regardless of the flag (routing is unchanged). This is a presentation-layer change only; access-group and auth semantics are untouched.

GET /v1/models/{model_id} now accepts the public name too, resolving it back to the internal routing key for the deployment lookup so a name advertised by the listing does not 404 on retrieve. When include_metadata=true, the fallback lookup runs against the internal routing key (what the router indexes fallbacks by) while the response id stays the public name, so configured fallbacks for team models are not silently dropped.

To keep listing and retrieve fully consistent, the retrieve path mirrors the listing's caller context: it accepts the same team_id and healthy_only query params and applies the same blocked/unhealthy hiding, so a public id advertised by /v1/models resolves to the same deployment on retrieve. Sibling deployments that share a public name collapse to a single entry by first occurrence, and a team row with an empty team_public_model_name is left untranslated (the internal key is shown) rather than emitting a blank id. Retrieve also echoes the public id /v1/models advertises even when the caller addresses the deployment by its internal routing key, so both endpoints report the same id.

Relevant issues

Fixes #30496

Pre-Submission checklist

I have added meaningful tests
My PR passes all unit tests on make test-unit
My PR's scope is as isolated as possible; it only solves 1 specific problem

Type

🐛 Bug Fix

Changes

The translation logic lives in litellm/proxy/common_utils/model_listing_utils.py as TeamModelNameTranslator (static methods, general_settings injected) so proxy_server.py stays thin. model_list builds each entry from listing_entries(...), which returns a (public response id, internal lookup id) pair: the public name is what the client sees, the internal key drives the metadata/fallback lookup. model_info resolves a requested public name to the internal key, scoped to the caller's accessible models so colliding public names across teams never cross an access boundary, then echoes the public id back in the response.

Tests cover access-group translation, the legacy opt-out flag, sibling de-dupe, no-router passthrough, retrieve-by-public-name (and the access-scoped 404), retrieve-by-internal-key echoing the public id, the empty public-name skip, and fallback metadata resolving via the internal routing key.

Screenshots / Proof of Fix

Local DB-backed proxy on localhost:4000 with two team-scoped BYOK deployments (team-gpt-4o-mini-a, team-gpt-4o-mini-b) tagged with access group grp-a, plus a virtual key scoped to grp-a. That key lists models through access-group expansion, which is exactly where the internal routing key used to leak.

After (this PR, default behavior):

curl -s http://localhost:4000/v1/models -H "Authorization: Bearer $KEY"

{
  "data": [
    {"id": "team-gpt-4o-mini-a", "object": "model", "created": 1677610602, "owned_by": "openai"},
    {"id": "team-gpt-4o-mini-b", "object": "model", "created": 1677610602, "owned_by": "openai"}
  ],
  "object": "list"
}

Before, reproduced on the same proxy by setting the legacy opt-out general_settings.use_team_public_model_name: false, the same key sees the internal routing keys leak:

{
  "data": [
    {"id": "model_name_c52aee85-07c7-492a-8337-01f1cd85c376_EAD5537F-3A6E-41F1-BE16-DB66CCD5F76A", "object": "model", "created": 1677610602, "owned_by": "openai"},
    {"id": "model_name_c52aee85-07c7-492a-8337-01f1cd85c376_BF3363CA-1DDB-4F3A-9A14-4D2EEEFCABA5", "object": "model", "created": 1677610602, "owned_by": "openai"}
  ],
  "object": "list"
}

Note

Medium Risk
Changes model id presentation and retrieve resolution across auth-scoped listings; mistakes could leak wrong-team metadata or break clients expecting internal keys, though routing and access checks are preserved and heavily tested.

Overview
/v1/models and /models now show team BYOK deployments as team_public_model_name instead of internal model_name_{team_id}_{uuid} keys (aligned with /v1/model/info). Routing is unchanged; operators can restore internal ids with general_settings.use_team_public_model_name: false.

A new TeamModelNameTranslator builds listing entries as (public response id, internal lookup id) so clients see friendly names while metadata/fallback lookups still use router keys. Shared public names dedupe with first-wins ordering, stay scoped to the caller’s accessible models, and skip empty public names.

GET /v1/models/{id} accepts public names, adds team_id and healthy_only to match list filtering, resolves to the same internal deployment as the list, and always returns the public id when translation is on (including when queried by internal key).

create_model_info_response is typed via ModelInfoResponse; OpenAPI schema and broad regression tests cover translation, metadata fallbacks, and access boundaries.

^{Reviewed by Cursor Bugbot for commit 11b70ac. Bugbot is set up for automated code reviews on this repo. Configure here.}

Behind general_settings.use_team_public_model_name (default False). When enabled, /v1/models and /models surface the public team_public_model_name for team-scoped (BYOK) models instead of the internal routing key model_name_{team_id}_{uuid} -- consistent with /v1/model/info and OpenAI-compatible. Off by default so the listing's model ids stay backward-compatible for callers that scripted against the internal name; routing by the internal name is unchanged regardless of the flag. Presentation-layer only: access-group, auth, and routing semantics are unchanged; non-team models are pass-through.

… tushar_byok_models_fix_update

greptile-apps · 2026-06-17T01:11:40Z

Greptile Summary

This PR fixes /v1/models and /models to surface the human-friendly team_public_model_name instead of the internal routing key model_name_{team_id}_{uuid} for team-scoped (BYOK) deployments, making the listing endpoints consistent with /v1/model/info. A general_settings.use_team_public_model_name: false flag lets operators preserve the legacy internal-key behavior.

TeamModelNameTranslator (new model_listing_utils.py) centralises the translation logic: listing_entries() returns (public_response_id, internal_lookup_id) pairs so fallback/metadata lookups still resolve via the router's internal key while only the public name is exposed to clients.
GET /v1/models/{id} now accepts the public name, resolves it to the internal routing key scoped to the caller's accessible models (preventing cross-team boundary leaks), and echoes the public id back in the response; new team_id and healthy_only query params mirror /v1/models so both endpoints share the same visibility filter.

Confidence Score: 5/5

Safe to merge; the change is presentation-layer only and the routing/auth paths are untouched.

The two previously flagged issues (fallback metadata being silently dropped and retrieve-by-public-name returning 404) are both addressed correctly: create_model_info_response now receives the internal routing key for fallback lookup while the public name is applied as the response id, and resolve_public_name maps the public name back to the internal key before the access check and deployment lookup. Access boundaries are preserved by restricting resolution to the caller's available_models. The only rough edge is that build_internal_to_public_map is iterated twice per retrieve call, which is a minor inefficiency rather than a correctness problem. Test coverage is thorough.

No files require special attention.

Important Files Changed

Filename	Overview
litellm/proxy/common_utils/model_listing_utils.py	New `TeamModelNameTranslator` utility — correct first-wins dedup, access-scoped resolution, empty-name skip, and flag opt-out; no issues found.
litellm/proxy/proxy_server.py	Both listing paths use `(response_id, lookup_id)` pairs correctly, and retrieve resolves public→internal before access check; `build_internal_to_public_map` is traversed twice per retrieve call.
litellm/proxy/utils.py	`create_model_info_response` now returns a typed `ModelInfoResponse` and the metadata path is cleaner; behaviour is preserved.
litellm/types/proxy/model_listing.py	New typed `ModelInfoResponse` / `ModelInfoMetadata` TypedDicts — correct shape, `metadata` correctly marked `NotRequired`.
tests/test_litellm/proxy/proxy_server/test_team_model_name_translation.py	Comprehensive new tests cover access-group translation, legacy opt-out, sibling de-dup, cross-team boundary isolation, retrieve-by-public-name, retrieve echoing public id, fallback metadata via internal key, and 404 on inaccessible name.
tests/llm_translation/base_llm_unit_tests.py	Image URL replaced with a jsdelivr-hosted sha-pinned asset to avoid gstatic's robots.txt blocking server-side fetchers — legitimate test reliability fix, coverage unchanged.
ui/litellm-dashboard/src/lib/http/schema.d.ts	Generated schema updated to reflect new `team_id` and `healthy_only` query params on both retrieve operations; no functional issues.

_{Reviews (12): Last reviewed commit: "test: avoid gstatic image url in image_u..." | Re-trigger Greptile}

ishaan-berri · 2026-06-17T01:22:23Z

@greptile review

ishaan-berri · 2026-06-17T01:22:49Z

@greptile review

codecov · 2026-06-17T01:25:59Z

Codecov Report

❌ Patch coverage is 96.70330% with 3 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
litellm/proxy/common_utils/model_listing_utils.py	96.07%	2 Missing ⚠️
litellm/proxy/proxy_server.py	95.00%	1 Missing ⚠️

📢 Thoughts on this report? Let us know!

The listing endpoints advertise team_public_model_name, but the retrieve endpoint validated and looked up by the raw id, so a public name 404'd. Resolve the public name back to the internal routing key (scoped to the caller's accessible models so colliding names never cross teams), look up by it, and echo the public name back as the response id.

…meTranslator Move the team-scoped (BYOK) listing/retrieve name translation out of proxy_server.py into a dedicated common_utils module. Static methods with general_settings injected so the logic is unit-testable without globals and proxy_server.py stays thin.

…info

ishaan-berri · 2026-06-17T01:39:08Z

@greptile review

… lookup Add listing_entries returning (public response id, internal lookup id) so include_metadata=true resolves fallbacks against the routing key the router indexes by, instead of the translated public name (which never matches).

ishaan-berri · 2026-06-17T01:53:47Z

@greptile review

…UP006)

listing_entries built a router-wide public-name -> internal-key reverse map and used it to pick each entry's metadata_lookup_id. When two teams published the same team_public_model_name, that reverse map collapsed to whichever internal routing key appeared last in the router model list, so a caller scoped to teamX could request /v1/models?include_metadata=true and receive teamY's fallback metadata under the shared public name. The lookup id is now drawn from model_names (the caller's accessible set) directly, the same access scoping resolve_public_name already applies on the retrieve path, which also removes the global reverse map entirely. Adds a unit regression on listing_entries and an end-to-end model_list test proving the shared public name surfaces the caller's own team fallbacks, not the other team's.

create_model_info_response returned dict[str, object], which is Any-free but coarse: callers could not see that "id" is a str or that "metadata" is the fallbacks shape, and "id"/"metadata" were typed as object. Introduce a ModelInfoResponse TypedDict (with metadata as a NotRequired nested TypedDict) and build the response in one shot instead of seeding a dict and mutating it, so the OpenAI-compatible model object is precisely typed end to end while the callers' model_info["id"] = ... assignments keep checking against str.

veria-ai · 2026-06-17T04:52:31Z

PR overview

All previously flagged issues have been addressed. No open security concerns remain on this pull request.

Security review

No open security issues remain on this pull request.

Fixed/addressed: 1 · PR risk: 0/10

mateo-berri · 2026-06-17T13:37:49Z

@greptileai

Generated by Claude Code

mateo-berri · 2026-06-17T13:45:00Z

bugbot run

Generated by Claude Code

cursor

Cursor Bugbot has reviewed your changes using high effort and found 2 potential issues.

Autofix Details

Bugbot Autofix prepared fixes for both issues found in the latest run.

✅ Fixed: Retrieve resolve skips list filters
- model_info now computes the same blocked/unhealthy hidden_names set as model_list (gated by a new healthy_only query param) and filters all_models before resolve_public_name, so retrieve cannot land on a deployment the listing hid.
✅ Fixed: Retrieve ignores list team_id filter
- model_info now accepts a team_id query param and forwards it to get_available_models_for_user, so retrieve sees the same team-scoped accessible list as the corresponding /v1/models?team_id=... call.

_{You can send follow-ups to the cloud agent here.}

GET /v1/models/{model_id} loaded models without the listing's team_id scoping and without the blocked/unhealthy filter, so a public id shown by /v1/models could resolve to a different internal deployment on retrieve. The retrieve endpoint now accepts the same team_id and healthy_only params and applies the same hidden_names filter before resolving the public name, keeping list and retrieve in agreement on which accessible deployment wins under first-occurrence dedup.

mateo-berri · 2026-06-17T14:01:04Z

@greptileai

Generated by Claude Code

mateo-berri · 2026-06-17T14:06:15Z

bugbot run

Generated by Claude Code

cursor

Cursor Bugbot has reviewed your changes using high effort and found 1 potential issue.

Autofix Details

Bugbot Autofix prepared a fix for the issue found in the latest run.

✅ Fixed: Empty public name maps listing
- Added truthy checks on team_id and team_public alongside the isinstance(str) checks in _internal_public_pair so empty strings no longer map an internal routing key to an empty public id, matching the falsy-skip behavior of _translate_model_name_for_response.

_{You can send follow-ups to the cloud agent here.}

Align _internal_public_pair with the falsy check used by _translate_model_name_for_response so a misconfigured row with an empty team_public_model_name no longer emits a listing entry whose id is the empty string. The internal routing key now passes through unchanged on /v1/models, matching what /v1/model/info already does.

…uiltin set The mirrored team_id/healthy_only query params on GET /v1/models/{id} and /models/{id} changed the OpenAPI spec, so regenerate the dashboard schema types to match. Also annotate the new unhealthy_names with builtin set[str] so it stays under the UP006 ruff-strict ceiling.

mateo-berri · 2026-06-17T14:28:22Z

@greptileai

Generated by Claude Code

mateo-berri · 2026-06-17T14:32:56Z

bugbot run

Generated by Claude Code

cursor

Cursor Bugbot has reviewed your changes using high effort and found 1 potential issue.

Autofix Details

Bugbot Autofix prepared a fix for the issue found in the latest run.

✅ Fixed: Retrieve echoes path not list id
- Map the resolved internal routing key back through the same TeamModelNameTranslator internal_to_public table the listing uses, so retrieve returns the public team name for the deployment regardless of whether the path was the public name or the internal key, with a regression test covering the internal-key path.

_{You can send follow-ups to the cloud agent here.}

…path GET /v1/models/{id} accepts both the public team name and the internal routing key, but the response always echoed the path back as the response id. Retrieving by the internal key under team-public-name translation returned the internal string while /v1/models advertised the public name for the same deployment, breaking OpenAI-compatible listing/retrieve consistency. Map the resolved internal key back through the same internal-to-public table the listing uses; falls back to the path for globals and when the flag is disabled.

mateo-berri · 2026-06-17T14:46:10Z

@greptileai

Generated by Claude Code

mateo-berri · 2026-06-17T14:51:55Z

bugbot run

Generated by Claude Code

cursor

✅ Bugbot reviewed your changes and found no new issues!

Comment @cursor review or bugbot run to trigger another review on this PR

^{Reviewed by Cursor Bugbot for commit 69ed748. Configure here.}

Anthropic (and other server-side image fetchers) now get a 400 because www.gstatic.com's robots.txt disallows crawling, breaking the detail=low/high cases of the shared base test_image_url across providers. Reuse the sha-pinned in-repo logo served via jsdelivr, which the detail=None cases already fetch successfully.

mateo-berri · 2026-06-17T15:27:36Z

@greptileai

Generated by Claude Code

mateo-berri · 2026-06-17T15:32:09Z

bugbot run

Generated by Claude Code

cursor

✅ Bugbot reviewed your changes and found no new issues!

Comment @cursor review or bugbot run to trigger another review on this PR

^{Reviewed by Cursor Bugbot for commit 11b70ac. Configure here.}

mateo-berri

LGTM; thanks!

* fix(proxy): optionally surface public team model name in /v1/models Behind general_settings.use_team_public_model_name (default False). When enabled, /v1/models and /models surface the public team_public_model_name for team-scoped (BYOK) models instead of the internal routing key model_name_{team_id}_{uuid} -- consistent with /v1/model/info and OpenAI-compatible. Off by default so the listing's model ids stay backward-compatible for callers that scripted against the internal name; routing by the internal name is unchanged regardless of the flag. Presentation-layer only: access-group, auth, and routing semantics are unchanged; non-team models are pass-through. * fix(proxy): default team model listings to public names * test(proxy): cover team model listing metadata * test(proxy): cover empty team listing deployments * refactor(proxy): simplify team model listing translation * fix(proxy): resolve public team model name on GET /v1/models/{id} The listing endpoints advertise team_public_model_name, but the retrieve endpoint validated and looked up by the raw id, so a public name 404'd. Resolve the public name back to the internal routing key (scoped to the caller's accessible models so colliding names never cross teams), look up by it, and echo the public name back as the response id. * test(proxy): cover public-name resolution on model retrieve * refactor(proxy): extract team model-name translation into TeamModelNameTranslator Move the team-scoped (BYOK) listing/retrieve name translation out of proxy_server.py into a dedicated common_utils module. Static methods with general_settings injected so the logic is unit-testable without globals and proxy_server.py stays thin. * refactor(proxy): use TeamModelNameTranslator in model_list and model_info * test(proxy): target TeamModelNameTranslator for model-name translation * fix(proxy): type create_model_info_response return as dict[str, object] * fix(proxy): keep internal routing key for team model listing metadata lookup Add listing_entries returning (public response id, internal lookup id) so include_metadata=true resolves fallbacks against the routing key the router indexes by, instead of the translated public name (which never matches). * fix(proxy): build /v1/models metadata from internal key, show public id * test(proxy): cover team listing fallback metadata via internal key * fix(proxy): use builtin dict generics in create_model_info_response (UP006) --------- Co-authored-by: Tushar More <tusharmore8408@gmail.com> Co-authored-by: Ishaan Jaffer <ishaanjaffer0324@gmail.com> (cherry picked from commit 60f4c01)

chore(release): backport #30380, #30503, #30558, #30130, #30588, #30495, #30690 to stable/1.89.x and cut 1.89.2

….2) (#327) This PR contains the following updates: | Package | Update | Change | |---|---|---| | [ghcr.io/berriai/litellm](https://images.chainguard.dev/directory/image/wolfi-base/overview) ([source](https://github.com/BerriAI/litellm)) | patch | `v1.89.1` → `v1.89.2` | --- ### Release Notes <details> <summary>BerriAI/litellm (ghcr.io/berriai/litellm)</summary> ### [`v1.89.2`](https://github.com/BerriAI/litellm/releases/tag/v1.89.2) [Compare Source](BerriAI/litellm@v1.89.2...v1.89.2) ##### Verify Docker Image Signature All LiteLLM Docker images are signed with [cosign](https://docs.sigstore.dev/cosign/overview/). Every release is signed with the same key introduced in [commit `0112e53`](BerriAI/litellm@0112e53). **Verify using the pinned commit hash (recommended):** A commit hash is cryptographically immutable, so this is the strongest way to ensure you are using the original signing key: ```bash cosign verify \ --key https://raw.githubusercontent.com/BerriAI/litellm/0112e53046018d726492c814b3644b7d376029d0/cosign.pub \ ghcr.io/berriai/litellm:v1.89.2 ``` **Verify using the release tag (convenience):** Tags are protected in this repository and resolve to the same key. This option is easier to read but relies on tag protection rules: ```bash cosign verify \ --key https://raw.githubusercontent.com/BerriAI/litellm/v1.89.2/cosign.pub \ ghcr.io/berriai/litellm:v1.89.2 ``` Expected output: ``` The following checks were performed on each of these signatures: - The cosign claims were validated - The signatures were verified against the specified public key ``` *** ##### What's Changed - chore(ui): rebuild ui by [@yuneng-berri](https://github.com/yuneng-berri) in [#30703](BerriAI/litellm#30703) - chore(release): backport [#30380](BerriAI/litellm#30380), [#30503](BerriAI/litellm#30503), [#30558](BerriAI/litellm#30558), [#30130](BerriAI/litellm#30130), [#30588](BerriAI/litellm#30588), [#30495](BerriAI/litellm#30495), [#30690](BerriAI/litellm#30690) to stable/1.89.x and cut 1.89.2 by [@yuneng-berri](https://github.com/yuneng-berri) in [#30681](BerriAI/litellm#30681) **Full Changelog**: <BerriAI/litellm@v1.89.1...v1.89.2> ### [`v1.89.2`](https://github.com/BerriAI/litellm/releases/tag/v1.89.2) [Compare Source](BerriAI/litellm@v1.89.1...v1.89.2) ##### Verify Docker Image Signature All LiteLLM Docker images are signed with [cosign](https://docs.sigstore.dev/cosign/overview/). Every release is signed with the same key introduced in [commit `0112e53`](BerriAI/litellm@0112e53). **Verify using the pinned commit hash (recommended):** A commit hash is cryptographically immutable, so this is the strongest way to ensure you are using the original signing key: ```bash cosign verify \ --key https://raw.githubusercontent.com/BerriAI/litellm/0112e53046018d726492c814b3644b7d376029d0/cosign.pub \ ghcr.io/berriai/litellm:v1.89.2 ``` **Verify using the release tag (convenience):** Tags are protected in this repository and resolve to the same key. This option is easier to read but relies on tag protection rules: ```bash cosign verify \ --key https://raw.githubusercontent.com/BerriAI/litellm/v1.89.2/cosign.pub \ ghcr.io/berriai/litellm:v1.89.2 ``` Expected output: ``` The following checks were performed on each of these signatures: - The cosign claims were validated - The signatures were verified against the specified public key ``` *** ##### What's Changed - chore(ui): rebuild ui by [@yuneng-berri](https://github.com/yuneng-berri) in [#30703](BerriAI/litellm#30703) - chore(release): backport [#30380](BerriAI/litellm#30380), [#30503](BerriAI/litellm#30503), [#30558](BerriAI/litellm#30558), [#30130](BerriAI/litellm#30130), [#30588](BerriAI/litellm#30588), [#30495](BerriAI/litellm#30495), [#30690](BerriAI/litellm#30690) to stable/1.89.x and cut 1.89.2 by [@yuneng-berri](https://github.com/yuneng-berri) in [#30681](BerriAI/litellm#30681) **Full Changelog**: <BerriAI/litellm@v1.89.1...v1.89.2> </details> --- ### Configuration 📅 **Schedule**: (in timezone America/New_York) - Branch creation - At any time (no schedule defined) - Automerge - At any time (no schedule defined) 🚦 **Automerge**: Disabled by config. Please merge this manually once you are satisfied. ♻ **Rebasing**: Whenever PR becomes conflicted, or you tick the rebase/retry checkbox. 🔕 **Ignore**: Close this PR and you won't be reminded about these updates again. --- - [ ] If you want to rebase/retry this PR, check this box --- This PR has been generated by [Mend Renovate](https://github.com/renovatebot/renovate).  Reviewed-on: https://git.greyrock.io/greyrock-labs/home-ops/pulls/327

…to v1.89.2 (#206) This PR contains the following updates: | Package | Update | Change | |---|---|---| | [https://github.com/BerriAI/litellm.git](https://github.com/BerriAI/litellm) | patch | `v1.89.1` → `v1.89.2` | --- ### Release Notes <details> <summary>BerriAI/litellm (https://github.com/BerriAI/litellm.git)</summary> ### [`v1.89.2`](https://github.com/BerriAI/litellm/releases/tag/v1.89.2) [Compare Source](BerriAI/litellm@v1.89.1...v1.89.2) #### Verify Docker Image Signature All LiteLLM Docker images are signed with [cosign](https://docs.sigstore.dev/cosign/overview/). Every release is signed with the same key introduced in [commit `0112e53`](BerriAI/litellm@0112e53). **Verify using the pinned commit hash (recommended):** A commit hash is cryptographically immutable, so this is the strongest way to ensure you are using the original signing key: ```bash cosign verify \ --key https://raw.githubusercontent.com/BerriAI/litellm/0112e53046018d726492c814b3644b7d376029d0/cosign.pub \ ghcr.io/berriai/litellm:v1.89.2 ``` **Verify using the release tag (convenience):** Tags are protected in this repository and resolve to the same key. This option is easier to read but relies on tag protection rules: ```bash cosign verify \ --key https://raw.githubusercontent.com/BerriAI/litellm/v1.89.2/cosign.pub \ ghcr.io/berriai/litellm:v1.89.2 ``` Expected output: ``` The following checks were performed on each of these signatures: - The cosign claims were validated - The signatures were verified against the specified public key ``` *** #### What's Changed - chore(ui): rebuild ui by [@yuneng-berri](https://github.com/yuneng-berri) in [#30703](BerriAI/litellm#30703) - chore(release): backport [#30380](BerriAI/litellm#30380), [#30503](BerriAI/litellm#30503), [#30558](BerriAI/litellm#30558), [#30130](BerriAI/litellm#30130), [#30588](BerriAI/litellm#30588), [#30495](BerriAI/litellm#30495), [#30690](BerriAI/litellm#30690) to stable/1.89.x and cut 1.89.2 by [@yuneng-berri](https://github.com/yuneng-berri) in [#30681](BerriAI/litellm#30681) **Full Changelog**: <BerriAI/litellm@v1.89.1...v1.89.2> </details> --- ### Configuration 📅 **Schedule**: (UTC) - Branch creation - At any time (no schedule defined) - Automerge - At any time (no schedule defined) 🚦 **Automerge**: Disabled by config. Please merge this manually once you are satisfied. ♻ **Rebasing**: Whenever PR becomes conflicted, or you tick the rebase/retry checkbox. 🔕 **Ignore**: Close this PR and you won't be reminded about this update again. --- - [ ] If you want to rebase/retry this PR, check this box --- This PR has been generated by [Mend Renovate](https://github.com/renovatebot/renovate).  Co-authored-by: Renovate Bot <renovate@bhamm-lab.com> Reviewed-on: https://codeberg.org/blake-hamm/bhamm-lab/pulls/206

….2) (#143) This PR contains the following updates: | Package | Update | Change | |---|---|---| | [ghcr.io/berriai/litellm](https://images.chainguard.dev/directory/image/wolfi-base/overview) ([source](https://github.com/BerriAI/litellm)) | patch | `v1.89.1` → `v1.89.2` | --- ### Release Notes <details> <summary>BerriAI/litellm (ghcr.io/berriai/litellm)</summary> ### [`v1.89.2`](https://github.com/BerriAI/litellm/releases/tag/v1.89.2) [Compare Source](BerriAI/litellm@v1.89.2...v1.89.2) ##### Verify Docker Image Signature All LiteLLM Docker images are signed with [cosign](https://docs.sigstore.dev/cosign/overview/). Every release is signed with the same key introduced in [commit `0112e53`](BerriAI/litellm@0112e53). **Verify using the pinned commit hash (recommended):** A commit hash is cryptographically immutable, so this is the strongest way to ensure you are using the original signing key: ```bash cosign verify \ --key https://raw.githubusercontent.com/BerriAI/litellm/0112e53046018d726492c814b3644b7d376029d0/cosign.pub \ ghcr.io/berriai/litellm:v1.89.2 ``` **Verify using the release tag (convenience):** Tags are protected in this repository and resolve to the same key. This option is easier to read but relies on tag protection rules: ```bash cosign verify \ --key https://raw.githubusercontent.com/BerriAI/litellm/v1.89.2/cosign.pub \ ghcr.io/berriai/litellm:v1.89.2 ``` Expected output: ``` The following checks were performed on each of these signatures: - The cosign claims were validated - The signatures were verified against the specified public key ``` *** ##### What's Changed - chore(ui): rebuild ui by [@yuneng-berri](https://github.com/yuneng-berri) in [#30703](BerriAI/litellm#30703) - chore(release): backport [#30380](BerriAI/litellm#30380), [#30503](BerriAI/litellm#30503), [#30558](BerriAI/litellm#30558), [#30130](BerriAI/litellm#30130), [#30588](BerriAI/litellm#30588), [#30495](BerriAI/litellm#30495), [#30690](BerriAI/litellm#30690) to stable/1.89.x and cut 1.89.2 by [@yuneng-berri](https://github.com/yuneng-berri) in [#30681](BerriAI/litellm#30681) **Full Changelog**: <BerriAI/litellm@v1.89.1...v1.89.2> ### [`v1.89.2`](https://github.com/BerriAI/litellm/releases/tag/v1.89.2) [Compare Source](BerriAI/litellm@v1.89.1...v1.89.2) ##### Verify Docker Image Signature All LiteLLM Docker images are signed with [cosign](https://docs.sigstore.dev/cosign/overview/). Every release is signed with the same key introduced in [commit `0112e53`](BerriAI/litellm@0112e53). **Verify using the pinned commit hash (recommended):** A commit hash is cryptographically immutable, so this is the strongest way to ensure you are using the original signing key: ```bash cosign verify \ --key https://raw.githubusercontent.com/BerriAI/litellm/0112e53046018d726492c814b3644b7d376029d0/cosign.pub \ ghcr.io/berriai/litellm:v1.89.2 ``` **Verify using the release tag (convenience):** Tags are protected in this repository and resolve to the same key. This option is easier to read but relies on tag protection rules: ```bash cosign verify \ --key https://raw.githubusercontent.com/BerriAI/litellm/v1.89.2/cosign.pub \ ghcr.io/berriai/litellm:v1.89.2 ``` Expected output: ``` The following checks were performed on each of these signatures: - The cosign claims were validated - The signatures were verified against the specified public key ``` *** ##### What's Changed - chore(ui): rebuild ui by [@yuneng-berri](https://github.com/yuneng-berri) in [#30703](BerriAI/litellm#30703) - chore(release): backport [#30380](BerriAI/litellm#30380), [#30503](BerriAI/litellm#30503), [#30558](BerriAI/litellm#30558), [#30130](BerriAI/litellm#30130), [#30588](BerriAI/litellm#30588), [#30495](BerriAI/litellm#30495), [#30690](BerriAI/litellm#30690) to stable/1.89.x and cut 1.89.2 by [@yuneng-berri](https://github.com/yuneng-berri) in [#30681](BerriAI/litellm#30681) **Full Changelog**: <BerriAI/litellm@v1.89.1...v1.89.2> </details> --- ### Configuration 📅 **Schedule**: (in timezone Europe/London) - Branch creation - At any time (no schedule defined) - Automerge - At any time (no schedule defined) 🚦 **Automerge**: Disabled by config. Please merge this manually once you are satisfied. ♻ **Rebasing**: Whenever PR becomes conflicted, or you tick the rebase/retry checkbox. 🔕 **Ignore**: Close this PR and you won't be reminded about these updates again. --- - [ ] If you want to rebase/retry this PR, check this box --- This PR has been generated by [Mend Renovate](https://github.com/renovatebot/renovate).  Reviewed-on: https://forgejo.hayden.moe/hayden/phoebe/pulls/143

* fix(proxy): optionally surface public team model name in /v1/models Behind general_settings.use_team_public_model_name (default False). When enabled, /v1/models and /models surface the public team_public_model_name for team-scoped (BYOK) models instead of the internal routing key model_name_{team_id}_{uuid} -- consistent with /v1/model/info and OpenAI-compatible. Off by default so the listing's model ids stay backward-compatible for callers that scripted against the internal name; routing by the internal name is unchanged regardless of the flag. Presentation-layer only: access-group, auth, and routing semantics are unchanged; non-team models are pass-through. * fix(proxy): default team model listings to public names * test(proxy): cover team model listing metadata * test(proxy): cover empty team listing deployments * refactor(proxy): simplify team model listing translation * fix(proxy): resolve public team model name on GET /v1/models/{id} The listing endpoints advertise team_public_model_name, but the retrieve endpoint validated and looked up by the raw id, so a public name 404'd. Resolve the public name back to the internal routing key (scoped to the caller's accessible models so colliding names never cross teams), look up by it, and echo the public name back as the response id. * test(proxy): cover public-name resolution on model retrieve * refactor(proxy): extract team model-name translation into TeamModelNameTranslator Move the team-scoped (BYOK) listing/retrieve name translation out of proxy_server.py into a dedicated common_utils module. Static methods with general_settings injected so the logic is unit-testable without globals and proxy_server.py stays thin. * refactor(proxy): use TeamModelNameTranslator in model_list and model_info * test(proxy): target TeamModelNameTranslator for model-name translation * fix(proxy): type create_model_info_response return as dict[str, object] * fix(proxy): keep internal routing key for team model listing metadata lookup Add listing_entries returning (public response id, internal lookup id) so include_metadata=true resolves fallbacks against the routing key the router indexes by, instead of the translated public name (which never matches). * fix(proxy): build /v1/models metadata from internal key, show public id * test(proxy): cover team listing fallback metadata via internal key * fix(proxy): use builtin dict generics in create_model_info_response (UP006) --------- Co-authored-by: Tushar More <tusharmore8408@gmail.com> Co-authored-by: Ishaan Jaffer <ishaanjaffer0324@gmail.com> (cherry picked from commit 60f4c01)

tushar8408 and others added 6 commits June 16, 2026 10:07

fix(proxy): default team model listings to public names

1a8e11a

test(proxy): cover team model listing metadata

01621d8

test(proxy): cover empty team listing deployments

beae9ac

refactor(proxy): simplify team model listing translation

09d4dc4

Merge remote-tracking branch 'upstream/litellm_internal_staging' into…

253e058

… tushar_byok_models_fix_update

greptile-apps Bot reviewed Jun 17, 2026

View reviewed changes

Comment thread litellm/proxy/proxy_server.py Outdated

ishaan-jaff added 2 commits June 16, 2026 18:27

test(proxy): cover public-name resolution on model retrieve

0bb8166

ishaan-berri force-pushed the litellm_v1_models_team_public_name branch from b2fc6fc to 0bb8166 Compare June 17, 2026 01:27

ishaan-jaff added 3 commits June 16, 2026 18:38

refactor(proxy): use TeamModelNameTranslator in model_list and model_…

bb29245

…info

test(proxy): target TeamModelNameTranslator for model-name translation

2354973

ishaan-jaff added 4 commits June 16, 2026 18:53

fix(proxy): type create_model_info_response return as dict[str, object]

75a5ec8

fix(proxy): build /v1/models metadata from internal key, show public id

6fc4e3a

test(proxy): cover team listing fallback metadata via internal key

801a158

veria-ai Bot reviewed Jun 17, 2026

View reviewed changes

Comment thread litellm/proxy/common_utils/model_listing_utils.py Outdated

BerriAI deleted a comment from veria-ai Bot Jun 17, 2026

fix(proxy): use builtin dict generics in create_model_info_response (…

1d3b830

…UP006)

ishaan-berri enabled auto-merge (squash) June 17, 2026 02:32

mateo-berri added 2 commits June 17, 2026 04:11

cursor Bot reviewed Jun 17, 2026

View reviewed changes

Comment thread litellm/proxy/proxy_server.py

Comment thread litellm/proxy/proxy_server.py

cursor Bot reviewed Jun 17, 2026

View reviewed changes

Comment thread litellm/proxy/common_utils/model_listing_utils.py

cursoragent and others added 2 commits June 17, 2026 14:13

cursor Bot reviewed Jun 17, 2026

View reviewed changes

Comment thread litellm/proxy/proxy_server.py

cursor Bot reviewed Jun 17, 2026

View reviewed changes

mateo-berri approved these changes Jun 17, 2026

View reviewed changes

ishaan-berri merged commit 60f4c01 into litellm_internal_staging Jun 17, 2026
157 of 159 checks passed

ishaan-berri deleted the litellm_v1_models_team_public_name branch June 17, 2026 16:17

yuneng-berri mentioned this pull request Jun 17, 2026

chore(release): backport #30380, #30503, #30558, #30130, #30588, #30495, #30690 to stable/1.89.x and cut 1.89.2 #30681

Merged

4 tasks

yuneng-berri added a commit that referenced this pull request Jun 18, 2026

Merge pull request #30681 from BerriAI/litellm_backport_1_89_x_0617

94dae27

chore(release): backport #30380, #30503, #30558, #30130, #30588, #30495, #30690 to stable/1.89.x and cut 1.89.2

Uh oh!

Conversation

ishaan-berri commented Jun 17, 2026 • edited by cursor Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What changed

Relevant issues

Pre-Submission checklist

Type

Changes

Screenshots / Proof of Fix

Uh oh!

greptile-apps Bot commented Jun 17, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Greptile Summary

Confidence Score: 5/5

Important Files Changed

Uh oh!

Uh oh!

ishaan-berri commented Jun 17, 2026

Uh oh!

ishaan-berri commented Jun 17, 2026

Uh oh!

codecov Bot commented Jun 17, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

ishaan-berri commented Jun 17, 2026

Uh oh!

ishaan-berri commented Jun 17, 2026

Uh oh!

Uh oh!

veria-ai Bot commented Jun 17, 2026

PR overview

Security review

Uh oh!

mateo-berri commented Jun 17, 2026

Uh oh!

mateo-berri commented Jun 17, 2026

Uh oh!

cursor Bot left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

mateo-berri commented Jun 17, 2026

Uh oh!

mateo-berri commented Jun 17, 2026

Uh oh!

cursor Bot left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

mateo-berri commented Jun 17, 2026

Uh oh!

mateo-berri commented Jun 17, 2026

Uh oh!

cursor Bot left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

mateo-berri commented Jun 17, 2026

Uh oh!

mateo-berri commented Jun 17, 2026

Uh oh!

cursor Bot left a comment

Choose a reason for hiding this comment

Uh oh!

mateo-berri commented Jun 17, 2026

Uh oh!

mateo-berri commented Jun 17, 2026

Uh oh!

cursor Bot left a comment

Choose a reason for hiding this comment

Uh oh!

mateo-berri left a comment

Choose a reason for hiding this comment

Uh oh!

ishaan-berri commented Jun 17, 2026 •

edited by cursor Bot

Loading

greptile-apps Bot commented Jun 17, 2026 •

edited

Loading

codecov Bot commented Jun 17, 2026 •

edited

Loading

cursor Bot left a comment •

edited

Loading

cursor Bot left a comment •

edited

Loading

cursor Bot left a comment •

edited

Loading