LiteLLM v1.102.0: MCP OAuth relay, Mistral TTS, Vertex Lyria, guardrails on streams
libraryLiteLLM 1.102.0 adds opt-in per-server OAuth relay discovery for MCP, applies key/team guardrails to MCP tool calls, ships Mistral /v1/audio/speech text-to-speech, adds Vertex Lyria model support, runs post_call guardrail pipelines (with text rewrites) on streaming responses, adds GovCloud Bedrock pricing, and improves adaptive/auto router behavior. Docker images are now cosign-signed.
An agent can now proxy MCP tool calls through LiteLLM with per-server OAuth relay discovery and key/team guardrails, stream responses through post-call guardrail pipelines, and call Mistral TTS or Vertex Lyria via a unified OpenAI-style API.
Notable agent-facing bits: MCP per-server OAuth relay discovery (PR #39936), MCP guardrails on tool calls (#39629), streaming post_call guardrails (#38788/#39233), Mistral TTS (#38755), Vertex Lyria (#30856), adaptive router fixes. Also cosign-signed Docker images.
pip install litellm==1.102.0from litellm import aspeech
import asyncio
async def main():
# New in 1.102.0: Mistral TTS via /v1/audio/speech
resp = await aspeech(
model="mistral/voxtral-tts",
voice="alloy",
input="Hello from an agent using LiteLLM 1.102.0",
)
with open("out.mp3", "wb") as f:
f.write(resp.content)
asyncio.run(main())