3 min read
vllm-project/vllm (v0.34.0rc2) & pydantic/pydantic-ai (v2.68.0) — AI Engineering & Agent Runtimes
Executive Signal Synthesis
Pre-release candidate v0.34.0rc2 updates memory utility defaults and test suites. Test in staging. Migration: Low. In parallel, pydantic/pydantic-ai (v2.68.0): Note text that looks like an application frame in user and. Drop-in upgrade. Migration: None.
Architectural Deep-Dive & Systems Impact
Removed explicit gpu_memory_utilization overrides across extensible test suites.
Refined internal test orchestration and documentation for memory limits.
Note text that looks like an application frame in user and tool content instead of escaping it by @DouweM in #11813
Add OutputTokenLimitExceeded, and fall back to the next model in ModelFallback when a response is cut off at max_tokens by @DouweM in #12407
Actionable Upgrade & Verification Checklist
- 01Update vllm to v0.34.0rc2.
- 02Update pydantic-ai to v2.68.0.
Verification & Upgrade Command Snippet
pip install vllm==0.34.0rc2
npm install pydantic-ai@2.68.0