BerriAI shipped LiteLLM v1.102.1 on 2026-09-23 (PyPI litellm 1.102.1). Versus 1.102.0 it backports Anthropic/Bedrock/Vertex `safeguards`→`safeguard_results` passthrough so Claude Code Auto mode through a LiteLLM gateway can show Auto mode server: Enabled, and cherry-picks seven TypeSafe Jev changes (passthrough, complexity router, relevance compaction guardrail, auto-router classifier, OpenRouter jev-1.13 pricing, etc.) from the 1.103 RC onto `stable/1.102.x`. Docker images keep cosign verify instructions.

Key Takeaways

  • ✓Shipped: v1.102.1 / PyPI 1.102.1
  • ✓Claude Code: gateway forwards `safeguards` and returns `safeguard_results` (#42538) so Auto mode server can stay Enabled
  • ✓TypeSafe Jev: seven PRs backported to stable/1.102.x (#42595)
  • ✓Deploy: pip install litellm==1.102.1; cosign-verify ghcr.io/berriai/litellm:v1.102.1
  • ✓Docs: Proxy Quick Start + Anthropic provider pages
🔗

Project Links & Resources

Direct Access
Direct access to official project resources and documentation
🐙
GitHub Release LiteLLM v1.102.1
github.com/BerriAI/litellm/releases/tag/v1.102.1
↗
📖
PyPI litellm 1.102.1
pypi.org/project/litellm/1.102.1/
↗
🐙
PR #42538 Anthropic safeguards backport
github.com/BerriAI/litellm/pull/42538
↗
🐙
PR #42595 TypeSafe Jev backport
github.com/BerriAI/litellm/pull/42595
↗
📖
LiteLLM Proxy Quick Start
docs.litellm.ai/docs/proxy/quick_start
↗
📖
LiteLLM Anthropic provider docs
docs.litellm.ai/docs/providers/anthropic
↗
📖
LiteLLM Pass-through docs
docs.litellm.ai/docs/proxy/pass_through
↗
🐙
BerriAI/litellm repository
github.com/BerriAI/litellm
↗
📖
cosign
docs.sigstore.dev/cosign/overview/
↗
🔬

In-Depth Technical Analysis

Core Background & Industry Pain Points Claude Code Auto mode needs the Anthropic /v1/messages safeguards field so the server-side classifier can return safeguard_results. Gateways that strip the field leave Auto mode server: Disabled and force the billed client-side classifier—plus billing notices telling users to contact their gateway provider. TypeSafe Jev (passthrough, complexity routing, relevance compaction) mostly lived on the 1.103 RC, so stable-line operators could not get both fixes on 1.102.x. ### Architecture Highlights & Internals v1.102.1 backports #42152/#42288 via #42538 so safeguards reach Anthropic/Bedrock/Vertex and safeguard_results return to the client. #42595 cherry-picks seven Jev PRs onto stable/1.102.x (passthrough + cost tracking, complexity router, relevance compaction guardrail, auto-router classifier, virtual-key budgets, OpenRouter typesafe/jev-1.13 pricing/pass-through). Release notes keep cosign verify steps for ghcr.io/berriai/litellm:v1.102.1. ### Authoritative Benchmarks & Measured Scores No public latency/throughput scores. Validate with Claude Code /status (Auto mode server: Enabled) and presence of safeguard_results; for Jev, track auto-router hit rate and token deltas around compaction. ### Developer Hands-on Guide pip install 'litellm[proxy]==1.102.1' or pull ghcr.io/berriai/litellm:v1.102.1, follow Proxy Quick Start, point Claude Code ANTHROPIC_BASE_URL at the gateway with --permission-mode auto, and confirm /status. See Anthropic and Pass-through. Package: PyPI 1.102.1.