Krrish

5️⃣ new things @LiteLLM 🎉 Gemini - New gemini-embedding-2 model tracking ✅ Bedrock - support for claude 4.6/7 effort levels 🛠️ Guardrails - log guardrail info on streaming post-call 💪 Perf - cache GCP token to avoid blocking event 📖 Docs - new guide logging to Splunk Cloud

5️⃣ new things @LiteLLM 🎉 OpenAI/Azure - gpt-5.5 Day 0 model support ✅ Team Budgets - fix older member budget enforcement 🧹 Releases - new release branch workflow 🛠️ Bedrock - preserve cache_control TTL on claude 4.5+ 🚀 OCR - new OCR Provider (Reducto)

5️⃣ new things @LiteLLM ✅ MCP - working semantic filtering w/ opencode 🛠️ Anthropic - map reasoning_auto_summary to thinking.display 🎉 OVHCloud - tool calling support 🧹 Anthropic - support tool_choice="none" 🚀 Qwen - add image generation support

5️⃣ new things @LiteLLM 🚀 Memory - support for storing agent memory ⚡️ [PERF] VertexAI - reuse cached credentials for partner models 💪 [PERF] VertexAI - single flight credential refresh 🛠️ Fireworks AI - add glm model w/ parallel tool calls 🧹 Chatgpt - gpt-5.4 streaming

5️⃣ new things @LiteLLM 🚀 Anthropic - Opus 4.7 day 0 support ✅ OOM - fix background health check mem leak 🛠️ Mistral - add new OCR params 💪 VertexAI - support multi-region endpoints 🎉 Github Copilot - override default GitHub Copilot authentication

5️⃣ new things @LiteLLM 🚀 Vertex - fix count tokens for anthropic 🛠️ Ollama - bubble up 'length' finish reason ✅ Noma - support key alias for Noma ⚡️ Auto-Router - support /v1/responses in complexity router 🧹 Credentials - sync in-memory credential_list after update

5️⃣ new things @LiteLLM 🚀 Prompt Compression - now available on proxy ✅ Bedrock - skip synthetic tool for response_format calls 🛠️ Openrouter - add pricing for gemini-3.1-flash-lite-preview model 🎉 Prometheus - add 'api_provider' label 🧹 Proxy - disable OpenAPI swagger

5️⃣ new things @LiteLLM 🚀 Prompt Compression - new BM-25 based prompt compression ✅ Gemini - correct indices on batch embedding response 🛠️ Anthropic x XAI - return tool call args on streaming 💪 Datadog - drain batches safely 🧹 Cache - prune in-memory cache

5️⃣ new things @LiteLLM 🐳 Docker - new RC release 💪 Proxy - Support key update via key hash 🛠️ Budgets - fix model max budget enforcement ⚡️ Budget - align Key, Team budget resets to calendar time 🎉 Models - add Wandb kimi-k2.5 and minimax-m2.5 for cost tracking

5️⃣ new things @LiteLLM 💪 vertex_ai - normalize finish_reason enum to OpenAI spec ✅ dashscope - preserve cache_control for prompt caching 🎉 TogetherAI - add together_ai/gpt-oss-120b 🛠️ ModelInfo - expose reasoning effort fields 🚀 PromptGuard - new guardrail integration

5️⃣ new things @LiteLLM 💸 Cache - support Responses API streaming ✅ XAI - add parallel_tool_calls support 💪 CI/CD - use numeric npm min-release age 🎉 Teams - resolve access group names 🛠️ Bedrock - handle document content blocks in Converse API

5️⃣ new things @LiteLLM 🐳 Docker - Signed GHCR releases ✅ MCP - fix tool name prefix check 🛠️ S3 - use prepped URL for sigv4 signed releases 🚀 Presidio - fix anonymize text alignment on input ⚡️ Prometheus - add 7m + 10m latency buckets for long-running LLM Requests

5️⃣ new things @LiteLLM 🐳 Docker - GHCR release tags are now immutable 💪 Anthropic - track web search cost 🛠️ VertexAI - support WIF auth in Azure Container Apps 🎉 Orgs - track org id in prometheus metrics 🚀 Bedrock - track us gov models

📣 Announcing @LiteLLM's April Townhall Agenda: - Product roadmap - Reliability and security updates - Open Q&A with the team Register here: forms.gle/zyyHLhHEyWGo1A…

5️⃣ new things @LiteLLM 🐳 Docker - All Pip+NPM+Docker deps now pinned 💪 Qohash - new Guardrail provider 🛠️ Teams - enforce TPM/RPM limits for members 🎉 Gemini - skip fetch for File API URLs 🧹 Azure - forward api_version on embeddings

5️⃣ new things @LiteLLM 🔒 Key Rotation - distributed lock prevents concurrent rotation 🌐 CORS - fully customizable proxy CORS settings 📊 OTEL - fix callback init + reasoning token tracking ⚡ Streaming - fix deepseek-reasoner routing 🛡️ UI - guard admin tabs for non-admins

5️⃣ new things @LiteLLM 🎉 MCP - granular member permissions 💪 Redis - circuit breaker to fast-fail on outages 🔧 Bedrock - support cache_control_injection_points in tool_config ✅ Gemini - camelCase taskType fix for embeddings 🧹 Proxy - defer logging until guardrails finish

5️⃣ new things @LiteLLM 🧹 Anthropic - support ANTHROPIC_AUTH_TOKEN,ANTHROPIC_BASE_URL 🎉 Guardrails - log execution info on LLM output runs ✅ Prisma - ensure idempotent migrations 🛠️ Prisma - kill orphaned subprocesses 💪 Claude Code Websearch Intercept - support github_copilot

5️⃣ new things @LiteLLM 🚀 Langsmith - show usage data in cost column ✅ Anthropic x Responses API - surface code execution result 🎉 Langfuse - fix langfuse otel trace propogation 💪 Docs - refreshed docs, easier navigation 🛠️ Guardrails - fix post_call model-level guardrails

5️⃣ new things @LiteLLM ✅ Anthropic - support cache control for files/images 🧹 aiohttp - auto-recover shared sessions when closed 🚀 Mistral - preserve diarization in transcriptions ⚡️ Guardrails - new AKTIO integration 🎉 GPT-5.4+ - auto-route tools, reasoning to Responses API