Skip to content

feat(ui): deployment affinity toggle for the auto-router - #36302

Merged
tin-berri merged 1 commit into
litellm_internal_stagingfrom
litellm_autorouter_affinity_toggle
Aug 11, 2026
Merged

feat(ui): deployment affinity toggle for the auto-router#36302
tin-berri merged 1 commit into
litellm_internal_stagingfrom
litellm_autorouter_affinity_toggle

Conversation

@tin-berri

@tin-berri tin-berri commented Aug 8, 2026

Copy link
Copy Markdown
Contributor

TLDR

Problem this solves:

How it solves it:

  • Renames the panel to Advanced: Affinity and puts both switches in it, deployment affinity first (default on) with session affinity below it
  • Both payload builders always write deployment_affinity, matching the session_affinity contract; the edit modal hydrates the stored value and manages the key, so a stored false survives an untouched save
  • Presets and the routing-test modal flow through the same single builder, so they inherit the default with no per-preset edits

User Flow

Before: an operator who wants one auto-router load-balanced across a group's deployments has no way to do it from the dashboard

  1. They open http://localhost:4000/ui/?page=llm-playground, then Models & Endpoints, Add Model, Auto Router
  2. They expand Detailed Configuration and every Advanced panel: no deployment affinity control exists anywhere
  3. Their router pins sessions to deployments (the backend default) and the only opt-out is hand-editing deployment_affinity: false into the config YAML

After: the toggle is in the form, prefilled with what the router will actually do

  1. They open the same Add Model, Auto Router flow and expand Detailed Configuration, then Advanced: Affinity
  2. The panel shows "Pin a session to one deployment per model group" already on, and "Pin a session to its first model" off, mirroring both backend defaults
  3. They switch deployment affinity off and create the router; the stored config carries deployment_affinity: false
  4. Editing that router later shows the switch off, and saving without touching it keeps false

Relevant issues

Linear ticket

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

image

UI change, so screenshots to follow this ordered list (proxy on :4000, dashboard dev server on :3000 via npm run dev in ui/litellm-dashboard)

  1. Models & Endpoints, Add Model, select Auto Router, expand Detailed Configuration, open Advanced: Affinity: deployment switch on, session switch off
  2. Toggle deployment affinity off, create the router, then confirm on the model's raw config that complexity_router_config.deployment_affinity is false
  3. Open Edit Auto Router on that router: the deployment switch renders off; click Save Changes without touching anything, re-open, still off
  4. Create a second router from the Claude Code preset without touching the panel and confirm its config carries deployment_affinity: true

Type

🆕 New Feature

Changes

  • ComplexityRouterConfig.tsx: panel renamed Advanced: Affinity, deployment switch added above the session switch, DEFAULT_DEPLOYMENT_AFFINITY = true exported as the single default owner
  • Both payload builders write the key unconditionally; the builder param is non-optional so the compiler forces every call site
  • Edit modal: deployment_affinity added to the managed keys and hydrated in initializeForm, the pair that keeps a stored false from being silently rewritten to the default on an untouched save
  • buildPresetPrefill hydrates the key so a future preset can express a non-default; the preset JSON is unchanged and inherits the default
  • Tests: builder four-case matrices in both builders' suites, untouched-save regression in the modal, default-carry and toggle-off-carry in the add tab, and the shared component's first affinity panel tests. The hydration mutation is killed by the untouched-save test

QA runbook

  1. npm run dev in ui/litellm-dashboard against a proxy running the current staging build
  2. Walk the four screenshot steps above
  3. npx vitest run src/components/add_model/ src/components/edit_auto_router/ src/lib/autorouter_presets.test.ts inside ui/litellm-dashboard: 350 tests

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

Note

Low Risk
Dashboard-only wiring for an existing router flag; behavior changes only when an operator toggles the new control or saves configs through the updated builders.

Overview
Adds dashboard control for deployment_affinity on complexity auto-routers (backend default on), which previously had no UI.

The Advanced: Affinity panel (renamed from Session Affinity) now has two switches: pin a session to one deployment per model group (default on) above pin a session to its first model (default off). Create and edit flows always persist deployment_affinity in complexity_router_config, with hydration and managed-key handling so stored false survives an untouched save.

Bundled Anthropic/OpenAI presets set deployment_affinity: true explicitly; preset prefill uses the same default when the key is omitted.

Reviewed by Cursor Bugbot for commit a29c6d1. Bugbot is set up for automated code reviews on this repo. Configure here.

@tin-berri

Copy link
Copy Markdown
Contributor Author

@greptileai can you reply here with a score and why that score?

@tin-berri

Copy link
Copy Markdown
Contributor Author

bugbot run

@greptile-apps

greptile-apps Bot commented Aug 8, 2026

Copy link
Copy Markdown
Contributor

Score: 4.5/5

What's done well:

  • The change is tightly scoped — one new flag, one shared constant, consistent propagation through both the add and edit flows
  • DEFAULT_DEPLOYMENT_AFFINITY is the single source of truth; every consumer imports it rather than inlining true
  • The "untouched save preserves false" regression test directly kills the hydration/default-override bug that would have been easy to miss
  • MANAGED_COMPLEXITY_ROUTER_KEYS is updated, so the edit modal won't silently strip the key on round-trip
  • Test matrix is thorough: both builders, both UI flows, and the shared component's first affinity panel tests

Minor issues that keep it from 5/5:

  1. Indentation inconsistency in test fixtures — in edit_auto_router_modal.test.ts the two new deployment_affinity: true lines are indented with 6 spaces while the surrounding object uses 2-space indentation. Won't break anything but it's a formatting inconsistency in committed test code.

  2. Third buildUpdatedComplexityRouterConfig deployment affinity test is slightly misleading — the description says "re-asserts the backend's on-by-default when the form value is absent, rather than dropping the key", but what actually happens is the form value is absent so the ?? DEFAULT_DEPLOYMENT_AFFINITY fallback fires, overwriting the stored false with true. That's the intended behavior but the test name doesn't make the "form wins over stored" semantic explicit, which could confuse a future reader.

  3. Session affinity copy update may be inaccurate — the updated description says "Always pins the deployment too, even with the deployment toggle off." If the backend from feat(router): independent, default-on deployment affinity for the auto-router #36146 actually ignores deployment_affinity when session_affinity: true regardless of what the client sends, that copy is correct. But if that's not the backend behavior, the UI is documenting a false constraint. Worth a quick sanity check against the backend implementation.

@cursor cursor Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

✅ 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 c1fc6bf. Configure here.

@greptile-apps

greptile-apps Bot commented Aug 8, 2026

Copy link
Copy Markdown
Contributor

Greptile Summary

Adds dashboard controls for auto-router deployment affinity and groups both affinity settings in one advanced panel

  • Carries deployment affinity through create, preset, routing-test, and edit payload flows
  • Hydrates stored values while matching the backend's default-on behavior
  • Adds component, builder, create-flow, and untouched-edit regression tests

Confidence Score: 5/5

The PR appears safe to merge, with deployment affinity consistently represented across the dashboard's create, preset, test, and edit flows

Stored false values survive hydration and untouched saves, omitted values match the backend default, and all shared builder callers propagate the setting without stripping or overriding it

Important Files Changed

Filename Overview
ui/litellm-dashboard/src/components/add_model/ComplexityRouterConfig.tsx Adds the deployment-affinity switch with backend-aligned defaults and accurately describes its interaction with session affinity
ui/litellm-dashboard/src/components/add_model/add_auto_router_tab.tsx Propagates the form value or backend-aligned default into the shared create and routing-test configuration
ui/litellm-dashboard/src/components/add_model/build_complexity_router_config.ts Makes deployment affinity a required builder input and writes it unconditionally into generated payloads
ui/litellm-dashboard/src/components/edit_auto_router/edit_auto_router_modal.tsx Hydrates, manages, and persists deployment affinity while correctly preserving stored false values
ui/litellm-dashboard/src/lib/autorouter_presets.ts Carries explicit preset values and applies the backend default when presets omit deployment affinity
ui/litellm-dashboard/src/components/edit_auto_router/edit_auto_router_modal.test.tsx Covers default hydration, stored-false preservation, and toggle-off persistence through the modal save flow

Reviews (1): Last reviewed commit: "feat(ui): deployment affinity toggle for..." | Re-trigger Greptile

@tin-berri
tin-berri force-pushed the litellm_autorouter_affinity_toggle branch from c1fc6bf to 6503853 Compare August 8, 2026 21:25
@tin-berri

Copy link
Copy Markdown
Contributor Author

@greptileai can you reply here with a score and why that score?

@tin-berri

Copy link
Copy Markdown
Contributor Author

bugbot run

@greptile-apps

greptile-apps Bot commented Aug 8, 2026

Copy link
Copy Markdown
Contributor

Score: 4.8/5

This is a well-executed, focused PR. Here's the breakdown:

What's strong:

  • Correctness is solid. Both payload builders emit deployment_affinity unconditionally (same contract as session_affinity), the edit modal hydrates and adds the key to MANAGED_COMPLEXITY_ROUTER_KEYS, and initializeForm uses a type-checked boolean guard — all three together close the "stored false silently rewritten to default on untouched save" regression.
  • Test coverage is thorough. The four-case matrix for both builders, the untouched-save regression test, the shared component render tests, and the add-tab integration tests are exactly what this change warrants. The test descriptions are precise and map directly to real customer scenarios.
  • Single source of truth. DEFAULT_DEPLOYMENT_AFFINITY = true exported from ComplexityRouterConfig.tsx and consumed everywhere else means the backend default is declared exactly once.
  • Scope discipline. Dashboard-only; no auth, routing, or backend logic touched.

Minor deductions:

  • The third test in buildUpdatedComplexityRouterConfig deployment affinity is named "re-asserts the backend's on-by-default when the form value is absent, rather than dropping the key" but it's actually testing a case where FORM_VALUE has no deployment_affinity key and a stored false gets overwritten by the default. This is intentional behavior (form drives the payload), but an operator who edits an unrelated field could silently lose their false if the UI failed to hydrate correctly — the test name should clarify it's a "form wins" contract, not a pure default-fill. This is mitigated by the edit modal's correct hydration, but the description is misleading.
  • build_complexity_router_config.test.ts adds deploymentAffinity: true to baseParams but doesn't add a false case the way the session affinity suite does — minor coverage gap, not a blocker.

@cursor cursor Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

✅ 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 6503853. Configure here.

@tin-berri

Copy link
Copy Markdown
Contributor Author

@greptileai can you reply here with a score and why that score?

@tin-berri

Copy link
Copy Markdown
Contributor Author

bugbot run

@tin-berri
tin-berri force-pushed the litellm_autorouter_affinity_toggle branch from 6503853 to a29c6d1 Compare August 8, 2026 21:41
@greptile-apps

greptile-apps Bot commented Aug 8, 2026

Copy link
Copy Markdown
Contributor

Score: 5/5

This is a clean, well-scoped dashboard feature with no issues worth blocking on.

Why 5:

  • Correctness: The deployment_affinity key is always written unconditionally (non-optional in the builder param), matching the session_affinity contract. The ?? fallback in both the create and edit paths uses the exported DEFAULT_DEPLOYMENT_AFFINITY = true, so the backend default is the single source of truth in one place.

  • Regression safety: The untouched-save test in EditAutoRouterModal directly exercises the case where deployment_affinity: false survives a round-trip without being clobbered by the default — the exact footgun this kind of hydration is prone to.

  • MANAGED_COMPLEXITY_ROUTER_KEYS: deployment_affinity is added to the set in edit_auto_router_modal.tsx, which is what prevents the key from being silently dropped during a partial save. That's the right place for it.

  • Test coverage: Four-case matrices in both builder suites, three-case modal suite for the new toggle, add-tab default/toggle-off cases, and the shared component's first affinity panel tests. The edge cases (stored false survives untouched save, absent key hydrates to default) are explicitly covered.

  • Scope: Purely dashboard plumbing; no backend changes, no SDK surface touched, no migration concerns.

@cursor cursor Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

✅ 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 a29c6d1. Configure here.

@tin-berri
tin-berri merged commit 6f36bee into litellm_internal_staging Aug 11, 2026
79 checks passed
@tin-berri
tin-berri deleted the litellm_autorouter_affinity_toggle branch August 11, 2026 02:31
doonga pushed a commit to greyrock-labs/home-ops that referenced this pull request Aug 24, 2026
…8.0) (#393)

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)) | minor | `v1.97.0` → `v1.98.0` |

---

### Release Notes

<details>
<summary>BerriAI/litellm (ghcr.io/berriai/litellm)</summary>

### [`v1.98.0`](https://github.com/BerriAI/litellm/releases/tag/v1.98.0)

[Compare Source](https://github.com/BerriAI/litellm/compare/v1.98.0...v1.98.0)

##### 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`](https://github.com/BerriAI/litellm/commit/0112e53046018d726492c814b3644b7d376029d0).

**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:v1.98.0
```

**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/v1.98.0/cosign.pub \
  ghcr.io/berriai/litellm:v1.98.0
```

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

- fix(bedrock): drop toolSpec.strict for Claude Sonnet 5 on Converse by [@&#8203;kr0k](https://github.com/kr0k) in [#&#8203;33196](https://github.com/BerriAI/litellm/pull/33196)
- fix(batches): attribute Vertex passthrough batch cost to key/team/tags by [@&#8203;yucheng-berri](https://github.com/yucheng-berri) in [#&#8203;34456](https://github.com/BerriAI/litellm/pull/34456)
- docs: rewrite the CLAUDE.md comment rule with explicit exceptions by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;36301](https://github.com/BerriAI/litellm/pull/36301)
- fix(proxy): scope file list pagination cursors to the caller by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;36093](https://github.com/BerriAI/litellm/pull/36093)
- fix(proxy): skip prisma-dependent hooks when no database is attached by [@&#8203;mateo-berri](https://github.com/mateo-berri) in [#&#8203;36273](https://github.com/BerriAI/litellm/pull/36273)
- fix(proxy): report has\_more false on caller-scoped file list pages by [@&#8203;mateo-berri](https://github.com/mateo-berri) in [#&#8203;36326](https://github.com/BerriAI/litellm/pull/36326)
- fix(proxy): restore management\_v1 query-param validation under fastapi>=0.140.7 by [@&#8203;HuanQian571](https://github.com/HuanQian571) in [#&#8203;35773](https://github.com/BerriAI/litellm/pull/35773)
- fix(proxy): stop /{provider}/v1/files from capturing /openai\_passthrough by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;36092](https://github.com/BerriAI/litellm/pull/36092)
- chore(typing): remove 914 basedpyright Any errors across 16 hotspot files by [@&#8203;mateo-berri](https://github.com/mateo-berri) in [#&#8203;36386](https://github.com/BerriAI/litellm/pull/36386)
- fix(router): keep batch fallbacks inside the model group that owns the file by [@&#8203;mateo-berri](https://github.com/mateo-berri) in [#&#8203;36181](https://github.com/BerriAI/litellm/pull/36181)
- feat(ptu): configure provisioned-throughput flat cost on a model deployment by [@&#8203;yucheng-berri](https://github.com/yucheng-berri) in [#&#8203;35341](https://github.com/BerriAI/litellm/pull/35341)
- docs: clarify the CLAUDE.md comment exceptions are any-of by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;36421](https://github.com/BerriAI/litellm/pull/36421)
- docs: replace the Changes PR template section with Caveats by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;36423](https://github.com/BerriAI/litellm/pull/36423)
- fix(bedrock): enable native structured output for GLM 5 and DeepSeek V3.2 by [@&#8203;alexshtf](https://github.com/alexshtf) in [#&#8203;35669](https://github.com/BerriAI/litellm/pull/35669)
- feat(ptu): daily rollup writes per-model PTU flat cost by active hour by [@&#8203;yucheng-berri](https://github.com/yucheng-berri) in [#&#8203;35343](https://github.com/BerriAI/litellm/pull/35343)
- feat(logging): add opt-in session\_id and trace\_id correlation to JSON log records via contextvars by [@&#8203;deepanshululla](https://github.com/deepanshululla) in [#&#8203;34418](https://github.com/BerriAI/litellm/pull/34418)
- feat(ptu): surface PTU flat cost on the daily activity read path by [@&#8203;yucheng-berri](https://github.com/yucheng-berri) in [#&#8203;35391](https://github.com/BerriAI/litellm/pull/35391)
- feat(router): add per-deployment allowed\_fails\_policy and cooldown\_time override support by [@&#8203;deepanshululla](https://github.com/deepanshululla) in [#&#8203;34416](https://github.com/BerriAI/litellm/pull/34416)
- feat(ptu): add PTU inputs to the model form and flat cost to the Usage page by [@&#8203;yucheng-berri](https://github.com/yucheng-berri) in [#&#8203;35393](https://github.com/BerriAI/litellm/pull/35393)
- fix(cost): price dict-shaped image input token details at the image rate by [@&#8203;vairodp](https://github.com/vairodp) in [#&#8203;33490](https://github.com/BerriAI/litellm/pull/33490)
- fix(model\_prices): refresh deprecation dates, correct xAI pricing and add missing provider models by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;36403](https://github.com/BerriAI/litellm/pull/36403)
- feat(ptu): gate PTU flat-cost attribution behind an opt-in env var by [@&#8203;yucheng-berri](https://github.com/yucheng-berri) in [#&#8203;36138](https://github.com/BerriAI/litellm/pull/36138)
- ci: cache Prisma CLI and engine binaries, split test timeout from setup by [@&#8203;mateo-berri](https://github.com/mateo-berri) in [#&#8203;36417](https://github.com/BerriAI/litellm/pull/36417)
- feat(rate limiting): configurable estimated output tokens per key, team and model by [@&#8203;yassin-berriai](https://github.com/yassin-berriai) in [#&#8203;36143](https://github.com/BerriAI/litellm/pull/36143)
- fix(ui): hide admin-only Logs tabs from roles that cannot call their endpoints by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36333](https://github.com/BerriAI/litellm/pull/36333)
- test(proxy): guard management\_v1 against fastapi names removed in supported releases by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36336](https://github.com/BerriAI/litellm/pull/36336)
- fix(ui): gate policy and prompt lookups on an admin capability by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36335](https://github.com/BerriAI/litellm/pull/36335)
- build(deps): bump pypdf to 6.15.0 to clear osv-scan by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;36350](https://github.com/BerriAI/litellm/pull/36350)
- fix(proxy): isolate guardrail load failures per row by [@&#8203;yucheng-berri](https://github.com/yucheng-berri) in [#&#8203;36432](https://github.com/BerriAI/litellm/pull/36432)
- fix(ui): gate organization and agent usage views behind capabilities by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36334](https://github.com/BerriAI/litellm/pull/36334)
- fix(reset\_budget\_job): atomic budget cascade with chunked reset scans by [@&#8203;ryan-crabbe-berri](https://github.com/ryan-crabbe-berri) in [#&#8203;36287](https://github.com/BerriAI/litellm/pull/36287)
- feat(proxy): add GET /v1/indexes to list vector store indexes by [@&#8203;ryan-crabbe-berri](https://github.com/ryan-crabbe-berri) in [#&#8203;36289](https://github.com/BerriAI/litellm/pull/36289)
- feat(ui): show vector store indexes on the Vector Stores page by [@&#8203;ryan-crabbe-berri](https://github.com/ryan-crabbe-berri) in [#&#8203;36306](https://github.com/BerriAI/litellm/pull/36306)
- fix(proxy): treat SAML as configured in UI SSO detection by [@&#8203;fancybear-dev](https://github.com/fancybear-dev) in [#&#8203;36196](https://github.com/BerriAI/litellm/pull/36196)
- fix(bedrock): reject Anthropic server-side web\_search tool with actionable error by [@&#8203;ryan-crabbe-berri](https://github.com/ryan-crabbe-berri) in [#&#8203;36473](https://github.com/BerriAI/litellm/pull/36473)
- fix(ui): open the classifier prompt editor above the edit auto-router form by [@&#8203;tin-berri](https://github.com/tin-berri) in [#&#8203;36438](https://github.com/BerriAI/litellm/pull/36438)
- fix(arize): trace MCP tool calls instead of crashing on CallToolResult by [@&#8203;yucheng-berri](https://github.com/yucheng-berri) in [#&#8203;36453](https://github.com/BerriAI/litellm/pull/36453)
- refactor(ui): make illegal DataTable prop combinations unrepresentable by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36470](https://github.com/BerriAI/litellm/pull/36470)
- fix(ui): scope Virtual Keys and Logs team lists to the caller by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36472](https://github.com/BerriAI/litellm/pull/36472)
- fix(ui): gate the Old Usage page behind a proxy-admin capability by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36469](https://github.com/BerriAI/litellm/pull/36469)
- docs(terraform): describe the provider release as automatic by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36467](https://github.com/BerriAI/litellm/pull/36467)
- feat(proxy): add per-deployment keepalive\_seconds SSE heartbeat to prevent load-balancer timeout on long streams by [@&#8203;deepanshululla](https://github.com/deepanshululla) in [#&#8203;34423](https://github.com/BerriAI/litellm/pull/34423)
- fix(router): cool down failed fallback deployments and correct cooldown TTL after Redis backfill by [@&#8203;deepanshululla](https://github.com/deepanshululla) in [#&#8203;35104](https://github.com/BerriAI/litellm/pull/35104)
- perf(spend): write each daily spend batch in one upsert statement by [@&#8203;yassin-berriai](https://github.com/yassin-berriai) in [#&#8203;36448](https://github.com/BerriAI/litellm/pull/36448)
- fix(ui): gate four sidebar pages on the roles their endpoints allow by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36475](https://github.com/BerriAI/litellm/pull/36475)
- fix(ui): restore the Logs Deleted Teams tab for organization admins by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36478](https://github.com/BerriAI/litellm/pull/36478)
- fix(websearch): stop leaking interception control fields to providers by [@&#8203;yassin-berriai](https://github.com/yassin-berriai) in [#&#8203;36480](https://github.com/BerriAI/litellm/pull/36480)
- test(e2e): cover the Anthropic web\_search server tool on Bedrock by [@&#8203;yassin-berriai](https://github.com/yassin-berriai) in [#&#8203;36443](https://github.com/BerriAI/litellm/pull/36443)
- fix(router): warn when a deployment's credentials contradict its provider by [@&#8203;yassin-berriai](https://github.com/yassin-berriai) in [#&#8203;36486](https://github.com/BerriAI/litellm/pull/36486)
- fix: net prompt-caching savings against the cache-write premium by [@&#8203;tin-berri](https://github.com/tin-berri) in [#&#8203;36452](https://github.com/BerriAI/litellm/pull/36452)
- feat(ui): deployment affinity toggle for the auto-router by [@&#8203;tin-berri](https://github.com/tin-berri) in [#&#8203;36302](https://github.com/BerriAI/litellm/pull/36302)
- fix(bedrock): use deployment credentials for AWS requests by [@&#8203;daleselaji-dev](https://github.com/daleselaji-dev) in [#&#8203;36160](https://github.com/BerriAI/litellm/pull/36160)
- fix(anthropic): preserve midturn system corrections by [@&#8203;eugene-yao-zocdoc](https://github.com/eugene-yao-zocdoc) in [#&#8203;34290](https://github.com/BerriAI/litellm/pull/34290)
- fix(email): stop duplicate legacy invitation email and fix its onboarding link by [@&#8203;mubashir1osmani](https://github.com/mubashir1osmani) in [#&#8203;36455](https://github.com/BerriAI/litellm/pull/36455)
- feat(ui): show models under each tier in routing benchmark chart by [@&#8203;tin-berri](https://github.com/tin-berri) in [#&#8203;36291](https://github.com/BerriAI/litellm/pull/36291)
- fix(proxy): inject streaming usage cost on openai passthrough streams by [@&#8203;mateo-berri](https://github.com/mateo-berri) in [#&#8203;36503](https://github.com/BerriAI/litellm/pull/36503)
- docs: require a user flow and live-proxy proof in bug reports by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;36498](https://github.com/BerriAI/litellm/pull/36498)
- fix(proxy): add config\_updated\_at audit timestamp for virtual keys by [@&#8203;ryan-crabbe-berri](https://github.com/ryan-crabbe-berri) in [#&#8203;36488](https://github.com/BerriAI/litellm/pull/36488)
- docs: require a user flow and a stuck-at proof in feature requests by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;36500](https://github.com/BerriAI/litellm/pull/36500)
- feat(router): add required-AND (&) tag prefix and allow\_fail\_open flag by [@&#8203;deepanshululla](https://github.com/deepanshululla) in [#&#8203;36193](https://github.com/BerriAI/litellm/pull/36193)
- feat(proxy): per-key prompt caching toggle via enable\_prompt\_caching by [@&#8203;ryan-crabbe-berri](https://github.com/ryan-crabbe-berri) in [#&#8203;36466](https://github.com/BerriAI/litellm/pull/36466)
- fix(bedrock): send tool-search beta header for Haiku 4.5 on Invoke /v1/messages by [@&#8203;mateo-berri](https://github.com/mateo-berri) in [#&#8203;36502](https://github.com/BerriAI/litellm/pull/36502)
- fix(bedrock): preserve adaptive thinking effort through the /v1/messages bridge by [@&#8203;mateo-berri](https://github.com/mateo-berri) in [#&#8203;36507](https://github.com/BerriAI/litellm/pull/36507)
- ci: retry transient network fetch failures in lint workflow by [@&#8203;mateo-berri](https://github.com/mateo-berri) in [#&#8203;36563](https://github.com/BerriAI/litellm/pull/36563)
- fix(ui): stub useIsOrgAdmin in UsageTab tests so useCan needs no QueryClient by [@&#8203;ryan-crabbe-berri](https://github.com/ryan-crabbe-berri) in [#&#8203;36565](https://github.com/BerriAI/litellm/pull/36565)
- fix(alerting): dedupe scheduled Slack spend reports across pods by [@&#8203;ryan-crabbe-berri](https://github.com/ryan-crabbe-berri) in [#&#8203;36489](https://github.com/BerriAI/litellm/pull/36489)
- chore(typing): clear 1.6k basedpyright Any errors across 56 files by [@&#8203;mateo-berri](https://github.com/mateo-berri) in [#&#8203;36543](https://github.com/BerriAI/litellm/pull/36543)
- fix(bedrock): add text block to converse user messages carrying documents by [@&#8203;mateo-berri](https://github.com/mateo-berri) in [#&#8203;36499](https://github.com/BerriAI/litellm/pull/36499)
- fix(deps): ship boto3 with the base SDK so bedrock works out of the box by [@&#8203;mubashir1osmani](https://github.com/mubashir1osmani) in [#&#8203;36568](https://github.com/BerriAI/litellm/pull/36568)
- fix(model\_prices): add provider-announced deprecation dates for Bedrock, Mistral, Cohere and Gemini models by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;36538](https://github.com/BerriAI/litellm/pull/36538)
- chore: bump litellm-enterprise 0.1.54 -> 0.1.55, litellm-proxy-extras 0.4.84 -> 0.4.85, litellm 1.97.0 -> 1.98.0 by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36577](https://github.com/BerriAI/litellm/pull/36577)
- fix(bedrock\_guardrails): skip ApplyGuardrail when there is no content to scan by [@&#8203;yucheng-berri](https://github.com/yucheng-berri) in [#&#8203;36441](https://github.com/BerriAI/litellm/pull/36441)
- fix(e2e): assert on the gen-AI span that served the stream, not the span count by [@&#8203;yassin-berriai](https://github.com/yassin-berriai) in [#&#8203;36582](https://github.com/BerriAI/litellm/pull/36582)
- test(e2e): harden vendor API coverage by [@&#8203;mubashir1osmani](https://github.com/mubashir1osmani) in [#&#8203;34557](https://github.com/BerriAI/litellm/pull/34557)
- test(e2e): add reproducers for passthrough and model budget gaps by [@&#8203;mubashir1osmani](https://github.com/mubashir1osmani) in [#&#8203;34657](https://github.com/BerriAI/litellm/pull/34657)
- test(e2e): cover google-native generateContent framing and prometheus queue time by [@&#8203;mubashir1osmani](https://github.com/mubashir1osmani) in [#&#8203;34650](https://github.com/BerriAI/litellm/pull/34650)
- chore(ci): promote internal staging to main by [@&#8203;tin-berri](https://github.com/tin-berri) in [#&#8203;36560](https://github.com/BerriAI/litellm/pull/36560)
- feat(router): make routing groups callable as virtual models and list them in /v1/models by [@&#8203;tin-berri](https://github.com/tin-berri) in [#&#8203;36519](https://github.com/BerriAI/litellm/pull/36519)
- fix(xai): bill web\_search from server\_side\_tool\_usage\_details by [@&#8203;geraint0923](https://github.com/geraint0923) in [#&#8203;30817](https://github.com/BerriAI/litellm/pull/30817)
- fix(responses): init completed\_response on bridge streaming iterator ([#&#8203;35411](https://github.com/BerriAI/litellm/issues/35411)) by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;35413](https://github.com/BerriAI/litellm/pull/35413)
- fix(batches): attribute Anthropic passthrough batch cost to the creating key, team and tags by [@&#8203;yucheng-berri](https://github.com/yucheng-berri) in [#&#8203;36468](https://github.com/BerriAI/litellm/pull/36468)
- feat(dashscope): add latest Model Studio models to the cost map by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;36496](https://github.com/BerriAI/litellm/pull/36496)
- fix(proxy): track streamed passthrough Responses cost by [@&#8203;william-xue](https://github.com/william-xue) in [#&#8203;36529](https://github.com/BerriAI/litellm/pull/36529)
- fix(model\_prices): advertise native structured output on every Bedrock DeepSeek V3.2 and GLM 5 id by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36597](https://github.com/BerriAI/litellm/pull/36597)
- test(bedrock): repoint live Claude tests off the retired Claude 3 Sonnet by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36600](https://github.com/BerriAI/litellm/pull/36600)
- fix(anthropic): preserve speed=fast in usage for /v1/messages and pass-through by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;36447](https://github.com/BerriAI/litellm/pull/36447)
- fix(proxy): forward resolved provider and deployment pricing in /cost/estimate by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;35880](https://github.com/BerriAI/litellm/pull/35880)
- feat(proxy): global SSE keepalive ping interval for OpenAI-shaped streaming routes by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;36154](https://github.com/BerriAI/litellm/pull/36154)
- fix(responses): preserve Codex namespace tool calls by [@&#8203;dcadenas](https://github.com/dcadenas) in [#&#8203;32536](https://github.com/BerriAI/litellm/pull/32536)
- fix(nvidia\_nim): preserve image passages and stop sending top\_k to /v1/ranking by [@&#8203;atomic](https://github.com/atomic) in [#&#8203;34177](https://github.com/BerriAI/litellm/pull/34177)
- fix: refactor HTTP handler initialization with client support by [@&#8203;Praveen11558](https://github.com/Praveen11558) in [#&#8203;30952](https://github.com/BerriAI/litellm/pull/30952)
- feat(lint): gate writable TypedDict fields with LIT012 by [@&#8203;mateo-berri](https://github.com/mateo-berri) in [#&#8203;36590](https://github.com/BerriAI/litellm/pull/36590)
- perf(proxy): stagger scheduled background jobs across jobs and pods by [@&#8203;yassin-berriai](https://github.com/yassin-berriai) in [#&#8203;36589](https://github.com/BerriAI/litellm/pull/36589)
- test: remove four mirror test files that exercise none of their module by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;34635](https://github.com/BerriAI/litellm/pull/34635)
- fix(router): stop re-applying router-selecting request tags to the routed tier's deployments by [@&#8203;mateo-berri](https://github.com/mateo-berri) in [#&#8203;36628](https://github.com/BerriAI/litellm/pull/36628)
- test: remove tests that never execute by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36681](https://github.com/BerriAI/litellm/pull/36681)
- fix(ui): align spend and budget columns by [@&#8203;daniel-meismer-zocdoc](https://github.com/daniel-meismer-zocdoc) in [#&#8203;35176](https://github.com/BerriAI/litellm/pull/35176)
- test: rename tests that a later definition shadowed by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36685](https://github.com/BerriAI/litellm/pull/36685)
- fix(passthrough): carry the budget reservation into request metadata by [@&#8203;yassin-berriai](https://github.com/yassin-berriai) in [#&#8203;36592](https://github.com/BerriAI/litellm/pull/36592)
- fix(mcp): bound MCP client requests with a session read timeout by [@&#8203;yassin-berriai](https://github.com/yassin-berriai) in [#&#8203;36675](https://github.com/BerriAI/litellm/pull/36675)
- fix(proxy): log requests rejected for an unparsable body in spend logs by [@&#8203;yassin-berriai](https://github.com/yassin-berriai) in [#&#8203;36673](https://github.com/BerriAI/litellm/pull/36673)
- refactor(ui): migrate cost-optimization to shadcn by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36629](https://github.com/BerriAI/litellm/pull/36629)
- refactor(ui): migrate cost-tracking to shadcn by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36631](https://github.com/BerriAI/litellm/pull/36631)
- refactor(ui): migrate admin-panel to shadcn by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36635](https://github.com/BerriAI/litellm/pull/36635)
- refactor(ui): migrate users dashboard to shadcn by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36642](https://github.com/BerriAI/litellm/pull/36642)
- refactor(ui): migrate prompts to shadcn by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36643](https://github.com/BerriAI/litellm/pull/36643)
- refactor(ui): migrate team settings to shadcn by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36641](https://github.com/BerriAI/litellm/pull/36641)
- refactor(ui): migrate models-and-endpoints to shadcn by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36648](https://github.com/BerriAI/litellm/pull/36648)
- refactor(ui): migrate policy impact popover to shadcn by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36653](https://github.com/BerriAI/litellm/pull/36653)
- fix(proxy): expand config-defined model access groups when resolving team models for /v2/model/info by [@&#8203;ryan-crabbe-berri](https://github.com/ryan-crabbe-berri) in [#&#8203;34211](https://github.com/BerriAI/litellm/pull/34211)
- fix(batches): strip NUL bytes from passthrough batch tags before the managed object write by [@&#8203;yucheng-berri](https://github.com/yucheng-berri) in [#&#8203;36688](https://github.com/BerriAI/litellm/pull/36688)
- test(e2e-ui): verify UI mutations against the API instead of trusting the toast by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36632](https://github.com/BerriAI/litellm/pull/36632)
- fix(proxy): serialize model reconciles so concurrent model writes stop evicting each other by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36687](https://github.com/BerriAI/litellm/pull/36687)
- chore(e2e): port the compat-matrix cron publisher to tests/e2e/claude\_code by [@&#8203;mateo-berri](https://github.com/mateo-berri) in [#&#8203;36465](https://github.com/BerriAI/litellm/pull/36465)
- fix(router): never price a strategy-router alias by [@&#8203;tin-berri](https://github.com/tin-berri) in [#&#8203;36691](https://github.com/BerriAI/litellm/pull/36691)
- feat(model\_prices): add NVIDIA Nemotron 3.5 Lightning on OpenRouter and DeepInfra by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;36696](https://github.com/BerriAI/litellm/pull/36696)
- feat(terraform/aws): make VPC, Aurora, and Redis optional by [@&#8203;yassin-berriai](https://github.com/yassin-berriai) in [#&#8203;36676](https://github.com/BerriAI/litellm/pull/36676)
- feat(ui): warn in the Admin UI when no Redis is configured by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;36495](https://github.com/BerriAI/litellm/pull/36495)
- fix(ui): show and edit key-level router settings on a virtual key by [@&#8203;yassin-berriai](https://github.com/yassin-berriai) in [#&#8203;36674](https://github.com/BerriAI/litellm/pull/36674)
- fix(router): forward auto-router alias params from the marker entry, not the first same-name deployment by [@&#8203;mateo-berri](https://github.com/mateo-berri) in [#&#8203;36626](https://github.com/BerriAI/litellm/pull/36626)
- fix(bedrock\_mantle): 1M context window and long-context pricing for GPT-5.6 Sol/Terra/Luna by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;36698](https://github.com/BerriAI/litellm/pull/36698)
- fix(model\_prices): sync the Groq registry with Groq's docs by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;36664](https://github.com/BerriAI/litellm/pull/36664)
- fix(router): let untagged requests bypass a tagged pre-routing strategy on shared model names by [@&#8203;mateo-berri](https://github.com/mateo-berri) in [#&#8203;36627](https://github.com/BerriAI/litellm/pull/36627)
- fix(spend): stop losing spend log rows when a flush is cancelled by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;34826](https://github.com/BerriAI/litellm/pull/34826)
- docs(claude): drop the @&#8203; prefix from the PR template path by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;36726](https://github.com/BerriAI/litellm/pull/36726)
- fix(langfuse): emit otel trace version and release on the keys langfuse v4 reads by [@&#8203;yucheng-berri](https://github.com/yucheng-berri) in [#&#8203;36702](https://github.com/BerriAI/litellm/pull/36702)
- test(interactions): follow Google spec drift replacing Turn with typed steps by [@&#8203;mateo-berri](https://github.com/mateo-berri) in [#&#8203;36730](https://github.com/BerriAI/litellm/pull/36730)
- refactor(ui): migrate team detail controls to shadcn by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36695](https://github.com/BerriAI/litellm/pull/36695)
- refactor(ui): migrate guardrail and duration controls to shadcn by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36693](https://github.com/BerriAI/litellm/pull/36693)
- refactor(ui): migrate guardrails-monitor, projects, logs to shadcn by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;34606](https://github.com/BerriAI/litellm/pull/34606)
- refactor(ui): migrate search and user controls to shadcn by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36694](https://github.com/BerriAI/litellm/pull/36694)
- fix(guardrails): scan and re-emit raw Anthropic SSE streams in the bedrock post-call hook by [@&#8203;yucheng-berri](https://github.com/yucheng-berri) in [#&#8203;36598](https://github.com/BerriAI/litellm/pull/36598)
- fix(helm): render nodeSelector on the migrations job by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36747](https://github.com/BerriAI/litellm/pull/36747)
- fix(langfuse): coerce header-sourced mask and trace-update steering values by [@&#8203;yucheng-berri](https://github.com/yucheng-berri) in [#&#8203;36740](https://github.com/BerriAI/litellm/pull/36740)
- refactor(ui): migrate usage tables to shared DataTable by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36707](https://github.com/BerriAI/litellm/pull/36707)
- refactor(ui): migrate guardrails monitor table to shared DataTable by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36709](https://github.com/BerriAI/litellm/pull/36709)
- refactor(ui): migrate guardrails content tables to shared DataTable by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36708](https://github.com/BerriAI/litellm/pull/36708)
- feat(gemini): day-0 pricing for gemini-3.7-flash by [@&#8203;mateo-berri](https://github.com/mateo-berri) in [#&#8203;36792](https://github.com/BerriAI/litellm/pull/36792)
- ci: promote staging to main by [@&#8203;mateo-berri](https://github.com/mateo-berri) in [#&#8203;36725](https://github.com/BerriAI/litellm/pull/36725)
- build(deps): bump nanoid to 3.3.18 to clear osv-scan by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;36787](https://github.com/BerriAI/litellm/pull/36787)
- fix(router): stop scoring system prompt text for code/technical complexity by [@&#8203;tin-berri](https://github.com/tin-berri) in [#&#8203;36721](https://github.com/BerriAI/litellm/pull/36721)
- feat(complexity\_router): calibrate the classifier rubric with worked examples, selectable per router by [@&#8203;tin-berri](https://github.com/tin-berri) in [#&#8203;36578](https://github.com/BerriAI/litellm/pull/36578)
- fix(interactions): map step and turn history to Responses API roles and content types by [@&#8203;mateo-berri](https://github.com/mateo-berri) in [#&#8203;36733](https://github.com/BerriAI/litellm/pull/36733)
- fix(ui): restore playground model filtering by endpoint by [@&#8203;mubashir1osmani](https://github.com/mubashir1osmani) in [#&#8203;36130](https://github.com/BerriAI/litellm/pull/36130)
- fix(proxy/batches): stop forwarding custom\_llm\_provider twice in list and cancel by [@&#8203;anxkhn](https://github.com/anxkhn) in [#&#8203;32813](https://github.com/BerriAI/litellm/pull/32813)
- refactor(ui): migrate TokenFlow and JsonViewer to shadcn by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36735](https://github.com/BerriAI/litellm/pull/36735)
- feat: pre-adoption shadow eval for the auto-router (blind pairwise judge, derived state) by [@&#8203;tin-berri](https://github.com/tin-berri) in [#&#8203;36587](https://github.com/BerriAI/litellm/pull/36587)
- refactor(ui): migrate SimpleMessageBlock and SimpleToolCallBlock to shadcn by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36737](https://github.com/BerriAI/litellm/pull/36737)
- refactor(ui): migrate HistoryTree and CollapsibleMessage to shadcn by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36738](https://github.com/BerriAI/litellm/pull/36738)
- refactor: replace Any with precise types across responses, proxy, and llms modules by [@&#8203;mateo-berri](https://github.com/mateo-berri) in [#&#8203;36763](https://github.com/BerriAI/litellm/pull/36763)
- refactor(ui): migrate TruncatedValue and OutputCard to shadcn by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36739](https://github.com/BerriAI/litellm/pull/36739)
- refactor(ui): migrate SectionHeader and ToolsSection to shadcn by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36793](https://github.com/BerriAI/litellm/pull/36793)
- feat(ui): migrate playground chat controls to shadcn by [@&#8203;mubashir1osmani](https://github.com/mubashir1osmani) in [#&#8203;36129](https://github.com/BerriAI/litellm/pull/36129)
- feat(xai): day-0 pricing for grok-4.6 by [@&#8203;mateo-berri](https://github.com/mateo-berri) in [#&#8203;36805](https://github.com/BerriAI/litellm/pull/36805)
- feat(ui): highlight Auto Router in the navbar announcement by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;36315](https://github.com/BerriAI/litellm/pull/36315)
- test(e2e): assert the model allow-list permits, not only denies by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36823](https://github.com/BerriAI/litellm/pull/36823)
- fix(proxy): tolerate a concurrent creator when creating spend views by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36824](https://github.com/BerriAI/litellm/pull/36824)
- fix(proxy): honor explicit null budget\_duration on team and key create + clearable UI dropdowns by [@&#8203;ryan-crabbe-berri](https://github.com/ryan-crabbe-berri) in [#&#8203;36699](https://github.com/BerriAI/litellm/pull/36699)
- feat(model\_prices): add meta/muse-spark-1.2 and its contributor tier by [@&#8203;mateo-berri](https://github.com/mateo-berri) in [#&#8203;36717](https://github.com/BerriAI/litellm/pull/36717)
- fix(auth): carry team grants in lite login session tokens by [@&#8203;yassin-berriai](https://github.com/yassin-berriai) in [#&#8203;36826](https://github.com/BerriAI/litellm/pull/36826)
- feat(ui): show provider prompt cache tokens in chat response metrics by [@&#8203;yassin-berriai](https://github.com/yassin-berriai) in [#&#8203;36827](https://github.com/BerriAI/litellm/pull/36827)
- fix(auth): stop the team fallback from widening model access by [@&#8203;yassin-berriai](https://github.com/yassin-berriai) in [#&#8203;36837](https://github.com/BerriAI/litellm/pull/36837)
- fix(proxy/team): resolve member\_delete cleanup by user id, not the addressed email by [@&#8203;yassin-berriai](https://github.com/yassin-berriai) in [#&#8203;36839](https://github.com/BerriAI/litellm/pull/36839)
- fix(cli): launch agents as a child process on Windows by [@&#8203;yassin-berriai](https://github.com/yassin-berriai) in [#&#8203;36822](https://github.com/BerriAI/litellm/pull/36822)
- feat(ui): shadow evals tab beside auto-router usage by [@&#8203;tin-berri](https://github.com/tin-berri) in [#&#8203;36588](https://github.com/BerriAI/litellm/pull/36588)
- feat(cli): make the hidden `lite` command list configurable by [@&#8203;yassin-berriai](https://github.com/yassin-berriai) in [#&#8203;36816](https://github.com/BerriAI/litellm/pull/36816)
- feat(azure\_ai): add Fireworks FW model pricing on Azure AI Foundry by [@&#8203;emerzon](https://github.com/emerzon) in [#&#8203;35613](https://github.com/BerriAI/litellm/pull/35613)
- fix: enable xhigh reasoning support for gpt-5.4-mini models by [@&#8203;emerzon](https://github.com/emerzon) in [#&#8203;26909](https://github.com/BerriAI/litellm/pull/26909)
- feat(azure-ai): add Grok 4.3 model metadata by [@&#8203;emerzon](https://github.com/emerzon) in [#&#8203;27932](https://github.com/BerriAI/litellm/pull/27932)
- feat(ui): render request metrics on the /ui/chat surface by [@&#8203;yassin-berriai](https://github.com/yassin-berriai) in [#&#8203;36845](https://github.com/BerriAI/litellm/pull/36845)
- fix(ui): stop a deselected MCP server keeping its grant on a virtual key by [@&#8203;yassin-berriai](https://github.com/yassin-berriai) in [#&#8203;36840](https://github.com/BerriAI/litellm/pull/36840)
- fix(team): sweep dangling team references and cache on team delete by [@&#8203;yassin-berriai](https://github.com/yassin-berriai) in [#&#8203;36819](https://github.com/BerriAI/litellm/pull/36819)
- fix(mcp): resolve admin OAuth sessions from any worker via DB-backed drafts by [@&#8203;yassin-berriai](https://github.com/yassin-berriai) in [#&#8203;36844](https://github.com/BerriAI/litellm/pull/36844)
- refactor(ui): migrate usage to shadcn by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36834](https://github.com/BerriAI/litellm/pull/36834)
- refactor(ui): migrate guardrails-monitor to shadcn by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36838](https://github.com/BerriAI/litellm/pull/36838)
- refactor(ui): migrate playground to shadcn by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36847](https://github.com/BerriAI/litellm/pull/36847)
- refactor(ui): migrate guardrails to shadcn by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36832](https://github.com/BerriAI/litellm/pull/36832)
- fix(batches): stop uncostable batches from starving the cost poll page by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;36714](https://github.com/BerriAI/litellm/pull/36714)
- perf(spend-logs): bound retention cleanup so one run cannot saturate the database by [@&#8203;yassin-berriai](https://github.com/yassin-berriai) in [#&#8203;36594](https://github.com/BerriAI/litellm/pull/36594)
- fix(proxy): fail config load when a callbacks entry is not dispatchable by [@&#8203;yassin-berriai](https://github.com/yassin-berriai) in [#&#8203;36858](https://github.com/BerriAI/litellm/pull/36858)
- fix(bedrock): hoist custom.defer\_loading before dropping custom on invoke tools by [@&#8203;yassin-berriai](https://github.com/yassin-berriai) in [#&#8203;36855](https://github.com/BerriAI/litellm/pull/36855)
- fix(access groups): sync assigned\_key\_ids from the key write paths by [@&#8203;yassin-berriai](https://github.com/yassin-berriai) in [#&#8203;36843](https://github.com/BerriAI/litellm/pull/36843)
- fix(mcp): expose client HTTP headers to logging callbacks and hooks by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;36724](https://github.com/BerriAI/litellm/pull/36724)
- fix(ptu): stop per-token billing on a PTU-configured deployment by [@&#8203;yucheng-berri](https://github.com/yucheng-berri) in [#&#8203;36829](https://github.com/BerriAI/litellm/pull/36829)
- fix(ui): add nvidia riva to the model provider list by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;36769](https://github.com/BerriAI/litellm/pull/36769)
- fix(scripts): end make check with a ran/skipped summary and verdict by [@&#8203;mateo-berri](https://github.com/mateo-berri) in [#&#8203;36864](https://github.com/BerriAI/litellm/pull/36864)
- fix(proxy): track spend for OpenAI passthrough /v1/embeddings by [@&#8203;lostmartian](https://github.com/lostmartian) in [#&#8203;36660](https://github.com/BerriAI/litellm/pull/36660)
- test(proxy): stop monkeypatch.undo re-planting fixture-mocked prisma\_client by [@&#8203;mateo-berri](https://github.com/mateo-berri) in [#&#8203;36872](https://github.com/BerriAI/litellm/pull/36872)
- fix(access groups): sync assigned\_team\_ids from the team write paths by [@&#8203;yassin-berriai](https://github.com/yassin-berriai) in [#&#8203;36825](https://github.com/BerriAI/litellm/pull/36825)
- ci: drop the CircleCI ui\_build and ui\_unit\_tests jobs by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36893](https://github.com/BerriAI/litellm/pull/36893)
- fix(langfuse)!: source the emitted metadata blob from StandardLoggingPayload by [@&#8203;yucheng-berri](https://github.com/yucheng-berri) in [#&#8203;36744](https://github.com/BerriAI/litellm/pull/36744)
- refactor(ui): migrate Navbar off antd to shadcn by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36902](https://github.com/BerriAI/litellm/pull/36902)
- refactor(ui): migrate log details drawer off antd to shadcn by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36904](https://github.com/BerriAI/litellm/pull/36904)
- refactor(ui): migrate AI Hub off antd and tremor to shadcn by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36908](https://github.com/BerriAI/litellm/pull/36908)
- refactor(ui): move the shared dropdowns and selectors onto shadcn primitives by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36924](https://github.com/BerriAI/litellm/pull/36924)
- refactor(ui): move the root-level dashboard components onto shadcn primitives by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36927](https://github.com/BerriAI/litellm/pull/36927)
- refactor(ui): move the settings page and bulk user invite onto shadcn primitives by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36936](https://github.com/BerriAI/litellm/pull/36936)
- refactor(ui): move the cost tracking components onto shadcn primitives by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36955](https://github.com/BerriAI/litellm/pull/36955)
- ci: drop the duplicate proxy\_unit\_tests letter-shard workflow by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36866](https://github.com/BerriAI/litellm/pull/36866)
- refactor(ui): migrate shared common\_components off antd and tremor by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36910](https://github.com/BerriAI/litellm/pull/36910)
- refactor(ui): migrate key info and permissions views off antd and tremor by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36913](https://github.com/BerriAI/litellm/pull/36913)
- feat(proxy): serve Anthropic-native /v1/models for Claude Code gateway discovery by [@&#8203;Ar-maan05](https://github.com/Ar-maan05) in [#&#8203;35455](https://github.com/BerriAI/litellm/pull/35455)
- refactor(ui): migrate router settings and shared badges off antd and tremor by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36915](https://github.com/BerriAI/litellm/pull/36915)
- refactor(ui): move the model hub and model select onto shadcn primitives by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36918](https://github.com/BerriAI/litellm/pull/36918)
- fix(ui): keep the cost tracking removal confirmation open until it settles by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36960](https://github.com/BerriAI/litellm/pull/36960)
- refactor(ui): declare DateRangePickerValue locally instead of importing it from tremor by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36962](https://github.com/BerriAI/litellm/pull/36962)
- fix(main): an explicit provider outranks a known OpenAI model name by [@&#8203;FahimaGold](https://github.com/FahimaGold) in [#&#8203;36800](https://github.com/BerriAI/litellm/pull/36800)
- fix(exception\_mapping): bare 429 in an error body no longer outranks the status code by [@&#8203;FahimaGold](https://github.com/FahimaGold) in [#&#8203;36705](https://github.com/BerriAI/litellm/pull/36705)
- refactor(ui): move MCP permission panels onto shadcn primitives by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36964](https://github.com/BerriAI/litellm/pull/36964)
- refactor(ui): migrate ten small dashboard files off antd and tremor by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36966](https://github.com/BerriAI/litellm/pull/36966)
- fix(proxy): force prisma recreate on postgres cached-plan error by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;36428](https://github.com/BerriAI/litellm/pull/36428)
- fix(transcription): stop a zero output rate from zeroing transcription cost by [@&#8203;hMED22](https://github.com/hMED22) in [#&#8203;36914](https://github.com/BerriAI/litellm/pull/36914)
- fix(langfuse): restrict trace steering keys to real langfuse trace fields by [@&#8203;yucheng-berri](https://github.com/yucheng-berri) in [#&#8203;36862](https://github.com/BerriAI/litellm/pull/36862)
- Revert "fix(auth): stop the team fallback from widening model access" ([#&#8203;36837](https://github.com/BerriAI/litellm/issues/36837)) by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36982](https://github.com/BerriAI/litellm/pull/36982)
- fix(ui): show zeroed auto-router usage stats when a window has no sessions by [@&#8203;tin-berri](https://github.com/tin-berri) in [#&#8203;36868](https://github.com/BerriAI/litellm/pull/36868)
- fix(mcp): keep admin-entered oauth endpoints in management reads by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;36888](https://github.com/BerriAI/litellm/pull/36888)
- fix(ui): distinguish hosted and local vLLM in the provider dropdown by [@&#8203;mateo-berri](https://github.com/mateo-berri) in [#&#8203;36974](https://github.com/BerriAI/litellm/pull/36974)
- fix(openai,azure): return a length-truncated 200 when the output budget fits no token by [@&#8203;yassin-berriai](https://github.com/yassin-berriai) in [#&#8203;36859](https://github.com/BerriAI/litellm/pull/36859)
- fix(proxy): always emit the Anthropic /v1/models token limits, null when unknown by [@&#8203;yassin-berriai](https://github.com/yassin-berriai) in [#&#8203;36961](https://github.com/BerriAI/litellm/pull/36961)
- feat(helm): add startupProbe and hpa.behavior to the componentized chart by [@&#8203;Louis-Vauterin](https://github.com/Louis-Vauterin) in [#&#8203;36382](https://github.com/BerriAI/litellm/pull/36382)
- fix(proxy): serve aggregate MCP endpoint on bare /mcp instead of 307-redirecting by [@&#8203;tin-berri](https://github.com/tin-berri) in [#&#8203;34845](https://github.com/BerriAI/litellm/pull/34845)
- feat(shadow\_eval): add reverse-direction shadow eval jobs by [@&#8203;tin-berri](https://github.com/tin-berri) in [#&#8203;36865](https://github.com/BerriAI/litellm/pull/36865)
- fix(proxy): requeue Redis spend buffer transactions when the DB commit fails by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;33881](https://github.com/BerriAI/litellm/pull/33881)
- feat(search): add Nimble as a search provider by [@&#8203;ilchemla](https://github.com/ilchemla) in [#&#8203;36347](https://github.com/BerriAI/litellm/pull/36347)
- fix(mcp): drop caller host and configured upstream headers from logged metadata by [@&#8203;yucheng-berri](https://github.com/yucheng-berri) in [#&#8203;36901](https://github.com/BerriAI/litellm/pull/36901)
- fix(azure\_ai): recognize real Search doc endpoints so teams can read/write via passthrough by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;36798](https://github.com/BerriAI/litellm/pull/36798)
- fix(anthropic): aggregate 5m/1h cache-write split across iterations path by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;34860](https://github.com/BerriAI/litellm/pull/34860)
- fix(anthropic cost): apply regional geo uplift to cached tokens by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;34850](https://github.com/BerriAI/litellm/pull/34850)
- fix(ui): match the MCP servers count badge to its sibling permission badges by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36984](https://github.com/BerriAI/litellm/pull/36984)
- fix(batches): mark terminal batch with no output file as processed in CheckBatchCost by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;35360](https://github.com/BerriAI/litellm/pull/35360)
- fix(caching): cache anthropic /v1/messages responses, including streaming by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;34581](https://github.com/BerriAI/litellm/pull/34581)
- fix(anthropic\_messages): make tool\_result images visible to OpenAI-compatible providers by [@&#8203;hMED22](https://github.com/hMED22) in [#&#8203;34462](https://github.com/BerriAI/litellm/pull/34462)
- feat(fireworks\_ai): translate NIM/vLLM extra params to Fireworks-native args by [@&#8203;milesadkins](https://github.com/milesadkins) in [#&#8203;35969](https://github.com/BerriAI/litellm/pull/35969)
- fix(ui): stop the models tab strip from scrolling vertically by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36993](https://github.com/BerriAI/litellm/pull/36993)
- fix(ui): anchor chips-combobox popups to the field instead of the inner input by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36995](https://github.com/BerriAI/litellm/pull/36995)
- feat(proxy): per-component response cost headers by [@&#8203;erensh27](https://github.com/erensh27) in [#&#8203;36965](https://github.com/BerriAI/litellm/pull/36965)
- fix(cost): track OpenAI/Azure web search tool cost per call by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;35286](https://github.com/BerriAI/litellm/pull/35286)
- fix(bedrock): resolve aliases in batch file records by [@&#8203;daleselaji-dev](https://github.com/daleselaji-dev) in [#&#8203;36159](https://github.com/BerriAI/litellm/pull/36159)
- fix: report real token usage on guardrail-blocked /v1/responses replies by [@&#8203;guptaishaan](https://github.com/guptaishaan) in [#&#8203;36907](https://github.com/BerriAI/litellm/pull/36907)
- fix(proxy): requeue spend logs when the DB write fails with a transport error by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;36716](https://github.com/BerriAI/litellm/pull/36716)
- fix(cost): tiered pricing supports cache creation cost and is all-or-nothing by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;36720](https://github.com/BerriAI/litellm/pull/36720)
- fix(vertex\_ai): translate /v1/embeddings batch rows to the Gemini embedding shape by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;35092](https://github.com/BerriAI/litellm/pull/35092)
- docs(claude): require ReadOnly on every TypedDict field (LIT012) by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;37005](https://github.com/BerriAI/litellm/pull/37005)
- refactor(ui): migrate access group create modal to RHF + zod + shadcn by [@&#8203;ryan-crabbe-berri](https://github.com/ryan-crabbe-berri) in [#&#8203;37033](https://github.com/BerriAI/litellm/pull/37033)
- refactor(ui): re-sync badge and skeleton onto the base-vega shadcn style by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;36991](https://github.com/BerriAI/litellm/pull/36991)
- feat(ui): link user detail team names to team pages by [@&#8203;ryan-crabbe-berri](https://github.com/ryan-crabbe-berri) in [#&#8203;37022](https://github.com/BerriAI/litellm/pull/37022)
- fix(model\_prices): correct DeepSeek V4 max output tokens by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;36925](https://github.com/BerriAI/litellm/pull/36925)
- fix(ui): rename models table Status column to Source by [@&#8203;ryan-crabbe-berri](https://github.com/ryan-crabbe-berri) in [#&#8203;37021](https://github.com/BerriAI/litellm/pull/37021)
- chore: bump litellm-enterprise 0.1.55 -> 0.1.56, litellm-proxy-extras 0.4.85 -> 0.4.86 by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;37045](https://github.com/BerriAI/litellm/pull/37045)
- feat(proxy): gate the Global Control Plane worker registry on an enterprise license by [@&#8203;yassin-berriai](https://github.com/yassin-berriai) in [#&#8203;36996](https://github.com/BerriAI/litellm/pull/36996)
- fix(model\_prices): add gemini 3.1 flash tts preview and legacy OpenAI shutdown dates by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;36788](https://github.com/BerriAI/litellm/pull/36788)
- fix(panw\_prisma\_airs): surface scan\_id on allowed requests by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;37037](https://github.com/BerriAI/litellm/pull/37037)
- fix(model\_map): flag native structured outputs on Anthropic-direct claude-sonnet-5 and claude-haiku-4-5 by [@&#8203;anmolg1997](https://github.com/anmolg1997) in [#&#8203;35930](https://github.com/BerriAI/litellm/pull/35930)
- fix(router): stop get\_router\_model\_info from wiping cached pricing by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;36985](https://github.com/BerriAI/litellm/pull/36985)
- fix(redis): unwrap decorated \_\_init\_\_s when deriving the from\_url kwargs allowlist by [@&#8203;anmolg1997](https://github.com/anmolg1997) in [#&#8203;36654](https://github.com/BerriAI/litellm/pull/36654)
- fix(proxy): reserve the larger declared output budget for TPM limits by [@&#8203;yassin-berriai](https://github.com/yassin-berriai) in [#&#8203;37001](https://github.com/BerriAI/litellm/pull/37001)
- fix(databricks): surface provider usage, including prompt-cache counts, in streaming chunks by [@&#8203;pokepoke81](https://github.com/pokepoke81) in [#&#8203;36943](https://github.com/BerriAI/litellm/pull/36943)
- fix(spend): give a batch's cost row a primary key of its own by [@&#8203;marty-sullivan](https://github.com/marty-sullivan) in [#&#8203;36876](https://github.com/BerriAI/litellm/pull/36876)
- feat: shadow eval samples /v1/messages and /v1/responses traffic by [@&#8203;tin-berri](https://github.com/tin-berri) in [#&#8203;36830](https://github.com/BerriAI/litellm/pull/36830)
- fix(ptu): stop a PTU deployment billing for grounded search by [@&#8203;yucheng-berri](https://github.com/yucheng-berri) in [#&#8203;37043](https://github.com/BerriAI/litellm/pull/37043)
- fix(fireworks\_ai): support router slugs via routers/ prefix by [@&#8203;heathriel](https://github.com/heathriel) in [#&#8203;34257](https://github.com/BerriAI/litellm/pull/34257)
- fix(bedrock): register managed-batch litellm\_params so they stop leaking to the provider (internal copy of [#&#8203;36633](https://github.com/BerriAI/litellm/issues/36633)) by [@&#8203;mateo-berri](https://github.com/mateo-berri) in [#&#8203;37048](https://github.com/BerriAI/litellm/pull/37048)
- fix(bedrock): resolve the managed-batch output bucket on every path that reads it by [@&#8203;mateo-berri](https://github.com/mateo-berri) in [#&#8203;37047](https://github.com/BerriAI/litellm/pull/37047)
- fix(bedrock): resolve the managed-batch output bucket on every path that reads it by [@&#8203;marty-sullivan](https://github.com/marty-sullivan) in [#&#8203;36634](https://github.com/BerriAI/litellm/pull/36634)
- feat(scripts): queue heavy gates behind a machine-wide slot lock by [@&#8203;mateo-berri](https://github.com/mateo-berri) in [#&#8203;36988](https://github.com/BerriAI/litellm/pull/36988)
- feat(mcp): scope gateway session bearers to the RFC 8707 resource by [@&#8203;tin-berri](https://github.com/tin-berri) in [#&#8203;35045](https://github.com/BerriAI/litellm/pull/35045)
- feat(ui): direction picker and reverse-mode display for shadow evals by [@&#8203;tin-berri](https://github.com/tin-berri) in [#&#8203;36994](https://github.com/BerriAI/litellm/pull/36994)
- fix(guardrails): return the full PANW AIRS scan response on blocked requests by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;37036](https://github.com/BerriAI/litellm/pull/37036)
- fix(passthrough): stop forwarding client Accept-Encoding upstream by [@&#8203;mateo-berri](https://github.com/mateo-berri) in [#&#8203;37058](https://github.com/BerriAI/litellm/pull/37058)
- fix(batches): account a managed batch's cost exactly once by [@&#8203;mateo-berri](https://github.com/mateo-berri) in [#&#8203;37050](https://github.com/BerriAI/litellm/pull/37050)
- fix(panw\_prisma\_airs): scan tool call args as plain text, not a tool\_event by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;37038](https://github.com/BerriAI/litellm/pull/37038)
- feat(lint): exempt TypedDict-annotated dict literals from LIT002 by [@&#8203;mateo-berri](https://github.com/mateo-berri) in [#&#8203;36869](https://github.com/BerriAI/litellm/pull/36869)
- docs(claude): tell agents to let heavy gates queue for machine-wide slots by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;37057](https://github.com/BerriAI/litellm/pull/37057)
- test: unstick the suites CircleCI is failing on by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;37059](https://github.com/BerriAI/litellm/pull/37059)
- docs(github): proof-of-fix section shows only the latest run as Before/After with nested cases by [@&#8203;mateo-berri](https://github.com/mateo-berri) in [#&#8203;37063](https://github.com/BerriAI/litellm/pull/37063)
- test(e2e): assert provider error shape instead of pinned prose by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;37065](https://github.com/BerriAI/litellm/pull/37065)
- fix(ui): de-duplicate the reset budget option and polish shadcn surfaces by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;37010](https://github.com/BerriAI/litellm/pull/37010)
- chore: rebuild Admin UI bundle from litellm\_internal\_staging by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;37066](https://github.com/BerriAI/litellm/pull/37066)
- test(e2e/ui): assert the log drawer chevrons by their lucide classes by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;37069](https://github.com/BerriAI/litellm/pull/37069)
- chore(ci): promote internal staging to main by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;37042](https://github.com/BerriAI/litellm/pull/37042)
- fix(ui): keep completion-mode models in the playground chat dropdown (backport to rc/1.98.0) by [@&#8203;yuneng-berri](https://github.com/yuneng-berri) in [#&#8203;37955](https://github.com/BerriAI/litellm/pull/37955)

##### New Contributors

- [@&#8203;kr0k](https://github.com/kr0k) made their first contribution in [#&#8203;33196](https://github.com/BerriAI/litellm/pull/33196)
- [@&#8203;HuanQian571](https://github.com/HuanQian571) made their first contribution in [#&#8203;35773](https://github.com/BerriAI/litellm/pull/35773)
- [@&#8203;alexshtf](https://github.com/alexshtf) made their first contribution in [#&#8203;35669](https://github.com/BerriAI/litellm/pull/35669)
- [@&#8203;vairodp](https://github.com/vairodp) made their first contribution in [#&#8203;33490](https://github.com/BerriAI/litellm/pull/33490)
- [@&#8203;fancybear-dev](https://github.com/fancybear-dev) made their first contribution in [#&#8203;36196](https://github.com/BerriAI/litellm/pull/36196)
- [@&#8203;daleselaji-dev](https://github.com/daleselaji-dev) made their first contribution in [#&#8203;36160](https://github.com/BerriAI/litellm/pull/36160)
- [@&#8203;eugene-yao-zocdoc](https://github.com/eugene-yao-zocdoc) made their first contribution in [#&#8203;34290](https://github.com/BerriAI/litellm/pull/34290)
- [@&#8203;geraint0923](https://github.com/geraint0923) made their first contribution in [#&#8203;30817](https://github.com/BerriAI/litellm/pull/30817)
- [@&#8203;william-xue](https://github.com/william-xue) made their first contribution in [#&#8203;36529](https://github.com/BerriAI/litellm/pull/36529)
- [@&#8203;dcadenas](https://github.com/dcadenas) made their first contribution in [#&#8203;32536](https://github.com/BerriAI/litellm/pull/32536)
- [@&#8203;atomic](https://github.com/atomic) made their first contribution in [#&#8203;34177](https://github.com/BerriAI/litellm/pull/34177)
- [@&#8203;Praveen11558](https://github.com/Praveen11558) made their first contribution in [#&#8203;30952](https://github.com/BerriAI/litellm/pull/30952)
- [@&#8203;anxkhn](https://github.com/anxkhn) made their first contribution in [#&#8203;32813](https://github.com/BerriAI/litellm/pull/32813)
- [@&#8203;lostmartian](https://github.com/lostmartian) made their first contribution in [#&#8203;36660](https://github.com/BerriAI/litellm/pull/36660)
- [@&#8203;FahimaGold](https://github.com/FahimaGold) made their first contribution in [#&#8203;36800](https://github.com/BerriAI/litellm/pull/36800)
- [@&#8203;Louis-Vauterin](https://github.com/Louis-Vauterin) made their first contribution in [#&#8203;36382](https://github.com/BerriAI/litellm/pull/36382)
- [@&#8203;ilchemla](https://github.com/ilchemla) made their first contribution in [#&#8203;36347](https://github.com/BerriAI/litellm/pull/36347)
- [@&#8203;milesadkins](https://github.com/milesadkins) made their first contribution in [#&#8203;35969](https://github.com/BerriAI/litellm/pull/35969)
- [@&#8203;erensh27](https://github.com/erensh27) made their first contribution in [#&#8203;36965](https://github.com/BerriAI/litellm/pull/36965)
- [@&#8203;guptaishaan](https://github.com/guptaishaan) made their first contribution in [#&#8203;36907](https://github.com/BerriAI/litellm/pull/36907)
- [@&#8203;pokepoke81](https://github.com/pokepoke81) made their first contribution in [#&#8203;36943](https://github.com/BerriAI/litellm/pull/36943)
- [@&#8203;heathriel](https://github.com/heathriel) made their first contribution in [#&#8203;34257](https://github.com/BerriAI/litellm/pull/34257)

**Full Changelog**: <https://github.com/BerriAI/litellm/compare/v1.97.0...v1.98.0>

### [`v1.98.0`](https://github.com/BerriAI/litellm/releases/tag/v1.98.0)

[Compare Source](https://github.com/BerriAI/litellm/compare/v1.97.0...v1.98.0)

##### 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`](https://github.com/BerriAI/litellm/commit/0112e53046018d726492c814b3644b7d376029d0).

**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:v1.98.0
```

**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/v1.98.0/cosign.pub \
  ghcr.io/berriai/litellm:v1.98.0
```

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

- fix(bedrock): drop toolSpec.strict for Claude Sonnet 5 on Converse by [@&#8203;kr0k](https://github.com/kr0k) in [#&#8203;33196](https://github.com/BerriAI/litellm/pull/33196)
- fix(batches): attribute Vertex passthrough batch cost to key/team/tags by [@&#8203;yucheng-berri](https://github.com/yucheng-berri) in [#&#8203;34456](https://github.com/BerriAI/litellm/pull/34456)
- docs: rewrite the CLAUDE.md comment rule with explicit exceptions by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;36301](https://github.com/BerriAI/litellm/pull/36301)
- fix(proxy): scope file list pagination cursors to the caller by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;36093](https://github.com/BerriAI/litellm/pull/36093)
- fix(proxy): skip prisma-dependent hooks when no database is attached by [@&#8203;mateo-berri](https://github.com/mateo-berri) in [#&#8203;36273](https://github.com/BerriAI/litellm/pull/36273)
- fix(proxy): report has\_more false on caller-scoped file list pages by [@&#8203;mateo-berri](https://github.com/mateo-berri) in [#&#8203;36326](https://github.com/BerriAI/litellm/pull/36326)
- fix(proxy): restore management\_v1 query-param validation under fastapi>=0.140.7 by [@&#8203;HuanQian571](https://github.com/HuanQian571) in [#&#8203;35773](https://github.com/BerriAI/litellm/pull/35773)
- fix(proxy): stop /{provider}/v1/files from capturing /openai\_passthrough by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;36092](https://github.com/BerriAI/litellm/pull/36092)
- chore(typing): remove 914 basedpyright Any errors across 16 hotspot files by [@&#8203;mateo-berri](https://github.com/mateo-berri) in [#&#8203;36386](https://github.com/BerriAI/litellm/pull/36386)
- fix(router): keep batch fallbacks inside the model group that owns the file by [@&#8203;mateo-berri](https://github.com/mateo-berri) in [#&#8203;36181](https://github.com/BerriAI/litellm/pull/36181)
- feat(ptu): configure provisioned-throughput flat cost on a model deployment by [@&#8203;yucheng-berri](https://github.com/yucheng-berri) in [#&#8203;35341](https://github.com/BerriAI/litellm/pull/35341)
- docs: clarify the CLAUDE.md comment exceptions are any-of by [@&#8203;devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#&#8203;36421](https://github.com/BerriAI/litellm/pull/36421)
- docs: replace the Changes …
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