5️⃣ new things @LiteLLM 🚀 Anthropic - [BETA] Citations API support ⚡️ O3-mini - native streaming support 🔒 Security - mask redis sensitive keys on `/cache/ping` 🔧 Azure - Timeout error includes passed timeout + elapsed time ✅ Rerank - infinity rerank endpoint fix

5️⃣ new things @LiteLLM 👋 OIDC - Scope-based model access support 🛠️ UI - fix showing internal user their created keys 🧹 UI - add user id attribution to created team keys 💪 Google AI Studio - gemini-2.0-flash + flash-lite pricing ⚡️ Azure - fix setting max_retries=0

5️⃣ new things @LiteLLM 💸 Vertex AI - Gemini-2.0-flash pricing + model info added 💪 Key RPM - improved rpm tracking during high traffic 🛠️ Azure - pydantic response_format + tool calls input fix ⚡️ Azure - o3 native streaming support 🚀 Nvidia NIM - `dimensions` param added

5️⃣ new things @LiteLLM 🎉 JWT Auth - support OIDC 'role' based proxy access 👋 Logging - Dynamic Message Redaction support 🛠️ Deepseek - fix openai sdk raw response error 🚀 UI - Fix serving UI on custom base path 🧹 Prometheus - fix setting key budget with custom metadata

5️⃣ new things @LiteLLM ✅ [PR] Cohere - add /v2/chat cost tracking 🧹 [PR] OpenAI - bump for reasoning_content 👋 [PR] Prometheus - fix metric label count 💪 [PR] Key Mgt - patch /key/update security 🚀 [PR] User Mgt - add budget to /user/new

5️⃣ new things @LiteLLM: 👋 O3: Added reasoning_effort param 💪 LLM: dev→sys role translation for non-OpenAI ⚡️ OpenRouter: Fixed "include_reasoning" param passing 🚀 Team Mgmt: SSO/IDP onboarding 🎉 Team Mgmt: autosuggest users on team member add

5️⃣ new things @LiteLLM 👋 JWT Auth - control user model access by AD 'groups' 🎉 JWT Auth - new `sso_user_id` field in db 💪 Team Mgmt - prevent duplicate member add 💪 Team Mgmt - raise error on `/team/member_delete` if user not in team 📖 Docs - New `-stable` release

5️⃣ new things @LiteLLM 👋 Guardrails - enforce `default_on` works on all requests 💪 Logging - disable `no-log` clientside param 🔑 Key Rotations - new doc + walkthrough 🚀 UI - new 'public team's' for easy user onboarding ⌛️ [PR] Logging - improved key masking in debug logs

5️⃣ new things @LiteLLM 🪨 Bedrock - new `bedrock/converse_like/` route 💸 XAI - Model Pricing added 🛠️ UI - fix key creation for internal users on UI 💪 Tokenizers - new `disable_hf_tokenizer_download` param for offline usage 🧹 Pydantic - cleanup valid config keys error

5️⃣ new things @LiteLLM 🚀 Guardrails - support caching across multiple event hooks 💪 Text Completion - `completion` logprobs transformation support ⚡️ Databricks - `tools` and `tool_choice` params added 🔥 MERGED - Bedrock Nova doc parsing ✅ MERGED - RBAC Model Access

5️⃣ new things @LiteLLM 🎉 SSO/JWT Auth - Role based model access support 💪 RBAC - new `role_permissions` config 👋 Timeouts - pass dynamic timeouts in headers ✅ Tokenizer - handle hf tokenizer network call fails 🚀 OpenAI - new `metadata` param, preview support

5️⃣ new @LiteLLM 👋 Bedrock - faster async image url converse calls ✅ Bedrock - Nova all files work (csv+) 💪 Anthropic - nested json on proxy fix 🔥 Proxy - high load + hanging request tests ⚡️ Deepseek - stream + reasoning content support

5️⃣ new things @LiteLLM 💸 Bedrock - model pricing updates + improved testing 💪 Cost Tracking - support `base_model` on all endpoints 🛠️ Custom Pricing - fix setting cost per token pricing 🎉 UI - allow user to 'join' available teams 📖 Docs - calling Bedrock via internal proxy

5️⃣ new things @LiteLLM 🐶 Datadog - `/health/services` endpoint support 💪 Bedrock - spend tracking with region name fix 🧹 Replicate - (sync) processing status retry 🛠️ UI - custom branding image fix 🚀 UI - new version update w/ guardrails support

5️⃣ new things @LiteLLM 🎉 Ollama - ssl verify = False support in call ⚡️ Headers - NEW `timeout` + `attempted_retry` headers 💪 End-user - support end-user id passed in as 'int' 🧹 SpendLogs - drop bad log causing db errors 🚀 SDK - New 'provider_specific_header' param

5️⃣ new things @LiteLLM 👋 Deepseek - R1 support w/ 'reasoning_content' ⚡️ Deepseek - streaming 'reasoning_content' support 🛠️ Watsonx - space id env var check fix 💪 VertexAI Anthropic - support dropping `anthropic-beta` headers 🧹 Swagger - fix custom docs rendering issue

5️⃣ new things @LiteLLM 🎉 Model Analytics - free UI model analytics 👋 Bedrock - streaming error exception mapping 💪 Router - fix num_retries=0 not being passed ⌛️ Timeout - timeout value added to openai/vllm error message ✅ Watsonx - stream_timeout support

5️⃣ new things @LiteLLM 🚀 Fallbacks - model auth-check on client-side fallbacks 🏥 Health check - Kill model request if taking too long 👋 UI - Fix Team spend calc. on Key pane 🔒 JWT-Auth - NEW 'enforce_rbac' param 🎉 JWT-Auth - support end user cost tracking on admin token

5️⃣ new things @LiteLLM ⚡️ Bedrock - 10x perf improvement on async image url calls 💪 Anthropic - handle nested pydantic objects 🔥 Router - high-traffic model hanging fixes 🚀 UI - fix team member role check 🎉 UI - allow team admin to delete any team key

5️⃣ new things @LiteLLM 👋 Cooldowns - for single-deployment model groups in high-traffic 💪 LM Studio - none api key support 🛠️ Fallbacks - sdk duplicate key error fix ✅ Vision - ` http://` url support on VertexAI + Bedrock 🧹 Codestral - stream chunk parsing fix